Enterprise AI advantage will not be won by the smartest model. Nor will it be won by having the best integration connectors.

Architectures are evolving fast, but what is clear is that many teams are misallocating their engineering capital because they cannot tell the difference between differentiation, control, and pure plumbing.

If you want to move beyond expensive AI science experiments and build a defensible operational moat, you have to ruthlessly map your architecture to where the value actually sits.

1. The Infrastructure Tier: Zero strategic advantage

This is the foundation layer. Ask a developer who is using AI code assistants and they will likely have strong views on which particular model they like to use. But the reality is that it matters less than most people think (it is more about familiarity) and for certain it will matter less in the future as model capabilities converge. There is technical leverage in the infrastructure tier and it matters for speed of deployment. I like how Google Gemini provides a rich, higher-level API with strong tool integrations, moving more of my orchestration and integration layer to their platforms. But everyone can rent the same capability, the same MCP integrations, and, in most cases, buy the same user interfaces.

Many AI roadmaps are currently obsessed with two things: enabling access to the frontier models, whether the latest or the easiest to write governance policies for, and writing connectors to plug them into internal databases or tools.

This is pure plumbing. Between aggressive model feature-wars and open-source standardization, the baseline tech stack is a commodity utility. Even the Model Context Protocol (MCP), while a brilliant, pragmatic standard, is just a upgraded USB connector. Connecting a model to enterprise data is a baseline requirement for survival, not a competitive edge. If your competitor can rent the same model and install the same socket, your technology moat is exactly zero.

The user interface layer does provide a small opportunity for differentiation, but most organizations are not pushing that hard but instead relying on vendor-provided applications, which brings its own problems as I discussed in The AI Dependency Trap.

This doesn’t mean you shouldn’t do the plumbing: you absolutely should, and you should do it well, and you should do it now. Just don’t kid yourself that this is where you build strategic value.

2. The Deployment Tier: Your licence to operate

This is the control layer. Enterprises will decide if they access AI models direct through the API to the vendor or inference engine, through a gateway, or through provisioned capacity such as AWS Bedrock Provisioned Throughput or Azure Foundry PTU. There is important technical value to be gained here, especially in agility to change and cost to support. In an agentic AI world identity and access management is a particularly hard problem that is well beyond what current solutions provide and is largely unsolved. The dynamic policy engines that I briefly mentioned in the previous post are still emerging. Also included in this layer, but not shown in the diagram, is your logging and audit infrastructure.

I regularly see organizations paying for thousands of Microsoft 365 Copilot subscriptions, only for their “governance” teams to block the tool from accessing company files. They are paying for a brain but refusing to give it sight.

If companies are paralyzed by letting a passive assistant read a slide deck, they have zero chance of deploying autonomous, background agents. If you can’t provide access in the walled garden of your M365 subscription, then your governance and tech leadership is broken.

The wider control challenge is much harder. As I wrote in Your AI Governance Was Built for Assistants, Not Agents, legacy IT wrappers that check a user’s ID at login completely fail here. To deploy in an agentic world, you need an active, programmatic gate embedded in the infrastructure to verify the model’s intent before it executes a tool loop, especially when hidden background triggers begin running silent processes outside human vision.

Controls don’t generate value, but they dictate whether you can turn the engine on. And if you can get them right before your competitors then you do have a temporal advantage, though it is unlikely to be sustained.

3. The Differentiation Tier: Your strategic differentiation

Strip away the plumbing and the controls, and you are left with the only two boxes that are uniquely proprietary to your business: Your Knowledge and Your Processes, deployed in unique ways.

• Enterprise Knowledge & Context (The Memory): What your company uniquely knows. Your proprietary data graphs, historical institutional context, and unstructured intellectual property.
• Application Logic & Agentic Orchestration (The Method): How your company uniquely operates, the processes that support your strategic differentiation.

Not all of your processes need to be unique. You might be a business that competes primarily on superior access to capital, entrenched regulatory moats, or classic economic rent-seeking.

But even if you win on asset scale or regulatory positioning, you still have to execute. If your AI strategy relies on generic, out-of-the-box prompts running on standard interfaces, you aren’t automating your edge; you are standardizing your mediocrity. The differentiation happens when your orchestration layer natively encodes the exact, specific procedural nuances that make your firm defensive.

This is also where the smartest vendors innovate. The foundational providers integrate more of the orchestration through reasoning models and advanced tool integrations, with some divergence in how this is architected across vendors. Higher in the value chain, vendors are providing entire specialist stacks for use cases like legal sectors where, again, their key value is in the curated knowledge to provide reliable information and the logic and orchestration that provides the checks and controls.

The Existential Challenge: If a vanilla, off-the-shelf agent running on a base model can replicate your workflow, your process is not a differentiator. It is legacy bureaucracy waiting to be automated out of existence by a leaner competitor.

Lessons from the Engine Room

Our recent work with the Lloyd’s Market Association (LMA) on AI and ML in Actuarial and Risk highlighted a critical market shift. As firms move from early experimentation to early-stage deployment, traditional compliance frameworks are hitting a wall. The overriding takeaway is clear: You cannot govern intelligent systems with static IT checklists. AI governance is fundamentally a process design challenge.

When we work in highly regulated markets—like actuarial reserving—the hard part is never connecting the LLM. That takes an hour. The hard part is embedding the controls directly into the execution path:

1. Where does the system sit?
2. What proprietary context can it safely access?
3. How do we verify intent before it triggers an action?

In a reserving transformation I led for a UK specialist insurer, we mapped out a transition from legacy Excel models to a code-first Python and Snowflake environment for actuarial teams. This was machine learning, the older cousin of AI. The breakthrough wasn’t “smarter math.” It was putting data segmentation logic directly into the actuaries’ hands and using Git-based workflows to make model changes completely transparent and rigorous.

We didn’t just automate a task; we built an immutable, auditable system of controls as code. We codified a unique process to protect unique knowledge.

The Executive Playbook

If you are directing capital for an AI-powered enterprise, your mandate is simple:

1. Minimize the plumbing: Rent the models, install the standard MCP sockets, and treat the infrastructure as an interchangeable utility.
2. Implement the controls: Build a dynamic policy engine so your risk team gives you the license to deploy.
3. Over-index on the value: Put every single dollar of engineering capital into codifying your proprietary knowledge graphs and designing your unique workflow orchestration.

A firm with a moderately smart model deeply embedded in an unassailable, highly specific workflow has the foundations to completely obliterate a competitor with a brilliant model trapped in an inbox.

About me: I help you turn complex data into clear decisions and commercial outcomes. I don’t just design architecture or manage governance; I build the engines that turbo-charge innovation and P&L growth. When most data leaders play defence, I play offence, bringing a CERN physicist’s first-principles edge to your competitive strategy.

To understand in detail how we implement the “Memory” layer of this stack, I built Vault Intelligence, an open-source framework for Obsidian dedicated to local knowledge systems. Moving beyond basic semantic search, it relies on two core architectural innovations:

1. Graph maintenance: The Gardener is an automated, human-in-the-loop process that prunes and maintains a structural knowledge graph of your institutional or personal knowledge to help create a carefully curated repository.
2. Context maximization: The Shadow Graph is an orchestration technique that optimizes and maximizes the fill of the model’s context window to feed the LLM a hyper-dense, hyper-relevant stream of institutional memory.

This works up to perhaps tens of thousands of documents. Next step is infinite scalability: not a conceptually difficult step but I always find that physical implementation brings out challenges that I would not otherwise consider.

Are your AI projects currently trapped in prototyping hell or governance paralysis, or are you successfully embedding them into your core workflows? Let me know in the comments. 👇

Follow me on LinkedIn for more insights.

You think you are solving AI governance. In reality, you may be capping your AI ROI.

We recently worked with the Lloyd’s Market Association on research into AI Governance and Risk. The finding that stayed with me was not that the market has moved beyond experimentation. It was that much of today’s governance is still built for AI assistance, not AI delegation.

Roughly 72% of firms now have AI frameworks in place. Over 60% require human-in-the-loop review for AI outputs. Those are sensible controls for copilots. They are not, by themselves, a delegation model.

That distinction matters because the risk shifts from content to consequences. Assistive AI helps people work faster. Delegated or Agentic AI acts within a workflow. It retrieves information, calls tools, triggers processes, and changes records. It takes action.

The governance problem is different. It is no longer just: “Was the output reviewed?” It becomes: “Was this system authorized to take this specific action, in this specific context, for this specific purpose, at this specific moment?”

That is where many AI ROI cases will hit the wall.

1. The Productivity Trap

Assistive AI (Copilots) makes individuals faster at existing tasks. It’s “Typewriter 2.0.” You get an efficiency gain, but your business model stays exactly the same. The bottlenecks just move around.

The larger ROI prize will not come from making existing tasks marginally faster. It will come from redesigning workflows so that AI can safely take on bounded actions.

The next step is therefore not a better chatbot. It is a system that can plan a workflow, call tools, trigger actions, and provide an auditable record of what it did and why.

2. The Permissions Problem Nobody Wants to Own

The subtle technical hurdle isn’t about traditional security; it’s about Contextual Authorization.

Traditional identity and access management was built around relatively stable roles: underwriter, claims handler, finance user, administrator.

Agentic AI breaks that model. An agent is not simply a user. It is a non-human actor operating on behalf of a user, within a task, using tools, data, and inferred intent.

Static permissions are too blunt for that reality.

The question is not simply: “Can this agent access the payments system?”

The better question is: “Does this agent have a valid, policy-aligned intent that means it is authorized at this moment in time to trigger this payment, for this claim, under these policy terms, within this authority limit, with this evidence trail?”

If the answer is broad standing access, you have created a control weakness. If the answer is human approval at every step, you have removed the operating advantage.

That is the authorization gap. Current Identity and Access Management (IAM) systems can only see the login, not the intent and the context.

3. The Logic Drift

When you delegate, you aren’t just delegating a task; you are delegating interpretation.

The risk here isn’t just a “hallucination”. It is Logic Drift, where an agent’s interpretation of compliance shifts after a model update, or, more subtly, over time. If your governance relies on reviewing the output after the fact, you are already too late. You are operating outside your authority in real-time.

The Provocation: You are Optimizing for What-Is

By sticking to human-review-only frameworks, you are optimizing what you do now, not your future capability. You are building a framework to survive an audit, not to compete in a different operating model.

When a competitor learns how to govern the intent of an agent in real time, rather than reviewing the output after the fact, the gap will not just be technological. It will be operational.

You cannot compete on manual review cycles against a firm that has made delegated action governable, observable, and safe.

Meanwhile, if official governance cannot distinguish between safe and unsafe delegation, business units will eventually find informal workarounds and those workarounds will be harder to observe, audit, and control.

You are then disadvantaged twice: too slow to capture the upside, and too weakly instrumented to control the risk.

The So-What: How to Govern Delegation

The practical answer is not to abandon human oversight. It is to move human judgement to the right point in the system.

Firms will need to govern delegation through architecture, not just policy and model controls. That means:

• Contextual controls: authorization based on task, role, purpose, data, limit, workflow state, and evidence.
• Action-level auditability: a record not only of what the model produced, but what it did, why it did it, and under whose authority.
• Pre-action guardrails: real-time controls that prevent unauthorized action before it occurs, not merely review it afterwards.
• Clear escalation rules: humans intervene where judgement, ambiguity, or authority limits require it, not as a universal brake on every action.

Most controls, most of the time, should be code that can run in real-time and at scale. Governance as code.

The Bottom Line

If your AI strategy depends on a human approving every meaningful action, you may still have a useful productivity tool. But you do not yet have a delegation model. And without a delegation model you will not achieve transformational benefits from the new technology.

The problem is not human involvement. The problem is indiscriminate human involvement.

Human judgement should be reserved for the points where judgement is genuinely needed. If every action requires manual approval, delegation fails. If no action requires manual escalation, governance fails.

Most firms are building AI governance for reviewable outputs. The next ROI frontier requires governance for authorized action.

See https://lmalloyds.com/ai-and-ml-in-actuarial-and-risk/ for the report.

About me: I help you turn complex data into clear decisions and commercial outcomes. I don’t just design architecture or manage governance; I build the engines that turbo-charge innovation and P&L growth. When most data leaders play defence, I play offence, bringing a CERN physicist’s first-principles edge to your competitive strategy.

Follow me on LinkedIn for more insights.

We worked with the Lloyds Market Association (LMA) to understand the state of AI adoption and to probe how risk and governance has evolved. While the insurance market at Lloyds is a specialized business, I believe the results and insights are generally useful and applicable.

Much of our discussions with Chief Risk Officers (CROs) were naturally about the in-house implementations, the value they have delivered, and the risk analysis and controls that the organizations had adopted. Ingestion of unstructured data for use in structured systems for underwriting and claims was a common use case with clear and measurable business value. Most had IT processes for procuring and controlling the tools. Requiring human oversight, validation, and decision was almost universally adopted as the main operational risk control.

There are some issues with this, but they are not for this post. Today, I want to focus on the hidden risk that many have ignored.

The hidden risk

Third-party AI is where governance becomes harder, because the firm does not control the system, the roadmap, or the commercial terms.

Firms are responding to this risk in a sensible way at the front end. Over 60% have specific AI contractual clauses and the same number use AI-specific due diligence during procurement, and over half have in place ongoing monitoring of third-party performance, security, and compliance.

These are all sensible defences that all firms in all industries can and should adopt.

But the more interesting question is whether they are enough once a vendor is strategically embedded in a core operating process. Because if it takes years to change an IT solution that supports underwriting, claims, finance, or some other key workflow, then the real issue is not simply whether the vendor uses AI. The real issue is what happens when that vendor introduces AI into the product, changes the terms, and you do not like the updated clauses.

At that point, the governance question is not just ‘can we object?’. It is ‘what is our practical alternative?’. And in many cases the honest answer may be: very little.

That is why I think the sharpest third-party risk here is dependency. A firm’s formal right to object may be much stronger than its practical ability to exit.

So the gap is not that firms have ignored vendor risk. It is that many have focused on diligence and clauses at the front end, without yet solving the harder issue of bargaining power, resilience, and realistic alternatives once the dependency already exists.

You may object that this is not different from every other vendor decision and argue this is not a new risk.

I think you’d be wrong.

The $15 trillion reason

About 15 trillion dollars of market value is dependent on AI becoming a foundational tool across the economy, or some similar vision of AI “success”. If you are a software company, then approximately¹ 100% of your stock market valuation depends on you having a compelling “AI story”. Therefore, about 100% of your bonus if you are a manager there.

Your software vendors will introduce more and more AI in their tools. $15 trillion says so. They will do so at an unprecedented pace and they will need to do it whether you want it or not because their investors demand it. $15 trillion says so.

This is new, and this is why it is different from vendor risks in the past. You are not looking at one vendor perhaps introducing something over time. You are looking at all vendors definitely introducing AI, all at once.

Reclaiming bargaining power

If the gap is bargaining power, what can you practically do to get it back?

Begin by looking at your data strategy. The real lock-in happens when your data is trapped in a vendor’s proprietary workflow and formats. To maintain strategic resilience, prioritize:

• Data sovereignty: Ensure you retain full ownership and access to the raw data processed by third-party AI.
• Vendor-agnostic architecture: Build your systems so that the data layer is decoupled from the application layer.
• Practical exit testing: Don’t just check the “exit clause” in the contract; test the technical feasibility of moving your data to a competitor.

Data portability is one critical step to ensuring you have practical alternatives. Without portability, you don’t have governance; you have a subscription to someone else’s risk appetite.

Third-party AI governance is not just a procurement exercise. It is a strategic resilience issue. By the time a critical vendor presents AI-enabled terms on a take-it-or-leave-it basis, you are already negotiating from a weakened position. And this will happen, guaranteed, and soon. Fifteen trillion dollars say it will. Now is the time to wake up and prepare.

→ What’s your take? How prepared are you? Reach out to discuss not just AI governance but also how to drive EBITDA impact from AI innovation.

See: https://lmalloyds.com/ai-and-ml-in-actuarial-and-risk/ for the research summary.

But this race is being run on a dangerously flawed metric. We are rewarding vendors for a narrow, superficial definition of “performance” – raw accuracy and photorealistic outputs – while ignoring the true measure of enterprise-grade AI: resilience.

This incentive structure is creating a hidden crisis. To win deals based on dazzling demos, vendors are building ever-more complex ‘hybrid’ models that fuse multiple AI architectures. This complexity drives the impressive results, but it also creates an opaque and unpredictable system where risks don’t just add up; they multiply.

This isn’t a theoretical problem. It’s a real-world, $440 million problem.

On 1 August 2012, the financial firm Knight Capital lost $440 million in 45 minutes and was driven to bankruptcy. The cause was not one faulty algorithm. It was the unintended, destructive interaction between a new trading system and a piece of old, supposedly defunct code. The components, safe in isolation, created a devastating feedback loop when combined. As the US Securities and Exchange Commission (SEC) later confirmed, this was a failure of systems interacting in an unforeseen way, running “without any substantial human intervention”1.

This is the blueprint for a future hybrid AI disaster. We are incentivising the creation of systems where one component could be silently failing, while another expertly masks the error, producing outputs that are plausible, confident, and catastrophically wrong.

The Governance Gap: When the Vendor’s “Best” is Your Biggest Risk ⚖️

Most leaders rightly protest that they have no control over a vendor’s proprietary architecture. This is true, but it’s also the heart of the problem. Vendor opacity does not absolve an organization of its duty of care; it dramatically increases the required level of diligence.

The legal and regulatory landscape is already struggling to keep pace. The UK’s Law Commission, in a recent discussion paper, highlights the immense difficulty of assigning liability within a complex “AI supply chain”. The paper notes that the “opacity” of these systems poses a fundamental challenge to establishing accountability, making it “unclear which parties owe the victim a duty of care”.

When the legal framework itself is admitting uncertainty, relying on a vendor’s performance metrics is no longer a viable governance strategy. If you cannot inspect the engine, you must become ruthless in your oversight of the vehicle’s safety features and operational limits.

A New Playbook for AI Leadership 🎯

A CDO’s power lies not in second-guessing code, but in redefining the terms of engagement. It requires a shift from chasing performance to demanding resilience.

1. Redefine Procurement Diligence: From ‘Capability’ to ‘Fragility’.

   The conversation with vendors must be fundamentally re-engineered. Change your metrics to change their behaviour.

   • Old Question: “What is your model’s accuracy?”

   • New Question: “Describe your system’s three most likely failure modes, particularly those involving cascading errors between internal components. How do you test for them?”

   • New Question: “What is your protocol for a root cause analysis when the model produces a nonsensical but plausible output? How do you prove it wasn’t an emergent failure from system complexity?”

2. Mandate Contractual Assurance and Audibility.

   The legal ambiguity surrounding AI is a powerful lever. Use it to build assurance into the contractual framework.

   • Demand Diagnostic Access: Insist on the right to access tools and logs that can trace a bad outcome to a specific input or component, even if the “why” remains opaque.

   • Introduce Independent Audits: Propose clauses requiring periodic, independent “adversarial audits,” where a third party actively tries to break the model. This is becoming a recognised practice, with frameworks like the NIST AI Risk Management Framework promoting continuous testing and independent reviews.

3. Build an Internal ‘Business Red Team’.

   Your organisation’s best defence is its own people. A “Business Red Team” is not a technical function; it is a formalised process where staff from the business – customer service, finance, marketing – are tasked with being the “user from hell.”

   • Their mandate is to probe the AI system with the messy, illogical, and adversarial behaviour of real-world humans. Their findings are not bug reports; they are a vital, qualitative dataset on the system’s true resilience.

The age of accepting AI systems as impenetrable black boxes is over. The true role of a data leader is to be the chief assurance officer for these powerful, complex, and fragile new capabilities. We do not need to build the engine, but we are absolutely accountable for the crash test results.

The AI arms race is moving faster than our governance frameworks.

☛ How are you and your teams redefining ‘performance’ to prioritize resilience over raw power? I’m keen to hear your perspectives in the comments. ☚

About me: I help organizations turn complex data into clear decisions and commercial outcomes. My focus is on enabling better decision-making and unlocking new value through data-driven innovation — especially where the stakes are high and the problems are difficult and poorly defined.

Follow me on LinkedIn for more insights.

References

1. Yazdani, S., Singh, A., Saxena, N., et al. (2025). Generative AI in depth: A survey of recent advances, model variants, and real-world applications. Journal of Big Data, 12(230).

2. U.S. Securities and Exchange Commission. (2013, October 16). Order Instituting Administrative and Cease-and-Desist Proceedings… against Knight Capital Americas LLC. (Release No. 70694). Available at: https://www.sec.gov/files/litigation/admin/2013/34-70694.pdf

3. Law Commission. (2025, August). AI and the Law: A Discussion Paper. Available from: https://lawcom.gov.uk/publication/artificial-intelligence-and-the-law-a-discussion-paper/

4. National Institute of Standards and Technology. (2023). AI Risk Management Framework (AI RMF 1.0). U.S. Department of Commerce. Available at: https://www.nist.gov/itl/ai-risk-management-framework

🚆 It started with a cancelled train and a long wait in the (presumably) dying British institution of Costa Coffee. It quickly got better.

🎨 First day of main event started strong with an insightful talk by Eric Drass. His art is maybe not for my walls, but he has thought long and hard about AI and society and is well worth a listen. (And he is an old perl hacker, so he’s OK by me…)

✨ Great line-up of speakers all of the day, finishing strong with Craig West talking about AI agents before a lovely evening event at the Royal Pavilion. More on that another day.

Themes that emerged for me included:

1. The Convergence of R and Python: A Polyglot Future

   A feature of the conference is the seamless integration of R and Python, not as competing languages, but as complementary tools in the data scientist’s arsenal. Many presentations explicitly mention the use of both languages to solve complex problems, reflecting a growing trend towards a ‘polyglot’ approach in data science teams.

   This pragmatic, tool-agnostic approach suggests that the era of language tribalism is waning, replaced by a focus on using the best tool for the specific task at hand.

2. The Pragmatic Application of Generative AI: From Hype to Utility

   While the excitement around generative AI and Large Language Models (LLMs) is palpable, the conference agenda demonstrates a clear shift from theoretical discussions to practical, real-world applications. The focus is on how these technologies can be harnessed to deliver tangible business value and enhance existing workflows.

   The emphasis here is on the ‘how’ rather than the ‘what’ of generative AI, signalling a maturation of this technology within the enterprise.

3. Data Science for Social and Public Good: A Focus on Impact

   A significant number of presentations are from organisations in the public and non-profit sectors, highlighting a strong undercurrent of using data science for social good. These talks demonstrate the power of data to address some of society’s most pressing challenges.

   This theme provides a vital counterpoint to the purely commercial applications of data science and underscores the profession’s potential to make a positive societal impact.

4. The ‘Last Mile’ of Data Science: Production, Validation, and Adoption

   The conference places a strong emphasis on the often-overlooked ‘last mile’ of data science: the challenges of deploying, validating, and ensuring the adoption of data products. This reflects a growing recognition that a model is only as good as its implementation and its impact on decision-making.

   This focus on the operational aspects of data science signals a maturing of the profession, moving beyond a sole focus on model building to a more holistic view of the data science lifecycle.

5. Bridging the Gap: From Data to Decision

   A final, overarching theme is the relentless focus on ensuring that data science leads to better decisions and commercial outcomes. This is the thread that connects all the other themes and is at the heart of my own professional focus.

A final, overarching theme is the relentless focus on ensuring that data science leads to better decisions and commercial outcomes. This is the thread that connects all the other themes and is at the heart my own professional focus.

The conference as a whole, with its mix of technical deep dives, case studies, and strategic discussions, is designed to equip practitioners with the skills and knowledge they need to bridge the gap between data and decisions, and to unlock new value through data-driven innovation.

While some talks touch on responsible AI, there was no dedicated track on the ethical implications of data science, such as bias, fairness, and transparency.

I am happy with the approach to safe and responsible AI we have adopted at Barnett Waddingham but it requires constant vigilance as tools develop. Conferences like this are invaluable.

🤔 How are you staying on top of developments in your area? As a leader, how do you connect with practitioners? 🔗

About me: I help organizations turn complex data into clear decisions and commercial outcomes. My focus is on enabling better decision-making and unlocking new value through data-driven innovation — especially where the stakes are high and the problems are difficult and poorly defined.

Follow me here and on LinkedIn for more insights.

The Problem: Nobody wants your dashboard 🧭

I speak with enterprise leaders and teams who are busy upgrading their organization’s data management and building dashboards. We live in a data economy, after all. They are integrating data sources, building data data platforms using Databricks, Snowflake, and similar architectures, and the best of them bring clarity to data provenance and are developing robust data dictionaries. These dictionaries are then used by developers who are producing dashboards and other data products for the organization.

Meanwhile, everyone in the enterprise are busy using conversational AI tools like ChatGPT, Copilot, Gemini, and so on.

Bad news: nobody wants a dashboard anymore. They never did. They want answers or inspiration. Dashboards were the best tools we had.

The Action: Adopt the right AI-First approach ✨

We absolutely still need to build high-quality, well-tested, robust data products for our organizations. The popular idea that AI will make all hard analysis go away is just wrong. Please continue to build great models and compelling visualizations.

But the AI as the preferred interface – that idea I can get behind. Just look around on your colleagues’ screens. Does more than half of them have an AI tool open as an app or in search?

Recall the “mobile-first” paradigm in user interface architecture, where we recognized the futility of forcing desktop interfaces onto mobile devices and instead embraced designing specifically for the mobile as the primary interaction point.

Now, we are entering the era of the AI-first paradigm.

Conversation may not be the only user interface, but it is a formidable one that is here to stay. The AI as an orchestration engine (‘reasoning model’) is powerful and is already transforming our productivity and helping us discover new depths of insights.

The enabling technology for AI orchestration is the Model Context Protocol (MCP). This allows your AI engine (“MCP Host”) to use services (“MCP Server”) to gather additional information to use as context for the query. This can be anything: databases, file systems, models, forecasts, external data, ….

Your data product must become an MCP Server. It can also be a dashboard, an Excel plugin, a database, a complete application, or something else. But it must be an MCP server. Its capability as an MCP Server will define your new enterprise integration layer.

This is the baseline expectation. Without it, your information architecture will be obsolete in 18 months because it will not support the orchestration layer nor your users’ preferred interface.

The Impact: New skills, new architecture ⚡

Some things stay the same, some things change. Your data dictionaries will now be available through MCP so they need to serve not just developers but also end users by describing what the data means in the language your colleagues and customers use and understand.

Your developers had data and user interface (UI) skills. You still need data and modelling, and you may still need UI skills. But you absolutely need API skills so your teams can create compelling, useful, and robust MCP servers.

A crucial evolution for your enterprise will be the development of an MCP Server that acts as a comprehensive registry or repository for other MCP Servers. This is likely the second MCP Server you want to build. As the enterprise is moving beyond traditional dashboards and leveraging AI as an orchestration engine, a robust discovery mechanism for data and context is absolutely crucial.

This approach creates a flexible, extensible data fabric where new “models of context” can be added by various teams, yet remain discoverable and usable across the entire enterprise, without creating silos or requiring manual, ad-hoc connections.

The Bottom Line: Start today 📌

You need to start today. AI adoption is the fastest technology adoption we have ever seen. Ignore the hype about AI taking away all the jobs or eliminating all the thinking. Focus on AI as interface and as orchestration. Dig deep into what it means for you and for your organization.

Are you still building yesterday’s information systems?

About me: I help organisations turn complex data into clear decisions and commercial outcomes. My focus is on enabling better decision-making and unlocking new value through data-driven innovation – especially where the stakes are high and the problems are difficult and poorly defined.

Follow me here and on LinkedIn for more insights.

But here’s the contradiction: while AI adoption is accelerating, many firms are still hiring, still training, and still struggling to fill roles. If AI is so disruptive, why hasn’t it delivered the promised efficiency gains or the mass layoffs?

The Problem: Misdiagnosing the decline 🤔

Everyone says AI is the culprit. Isabel Berwick’s FT Working It episode (no paywall on this one) captures this sentiment well: entry-level jobs are vanishing, and AI is doing 80% of what junior lawyers used to do. Jérémy Clédat, CEO of Welcome to the Jungle, observes:

“You see that this decline has been three times faster and higher than other positions for more senior people.”

Chris Eldridge, CEO of recruitment firm Robert Walters, adds:

“Companies just don’t want to risk their capital right now. And therefore, they’re just pausing the graduate intake. But that pause really has now lasted for two, two-and-a-half years.”

So here’s the twist: the hiring freeze began before generative AI went mainstream. Economic uncertainty, not AI, is the consistent factor. The result? A talent vacuum: not because AI replaced juniors, but because firms stopped hiring them.

So what’s really happening?

The Insight: AI is a convenient scapegoat 🔎

The real story is more nuanced. AI is not yet capable of replacing entry-level professionals wholesale. Most deployments are augmentative, not substitutive. Simmons & Simmons’ in-house tool “Percy” helps lawyers process documents, but as one partner explains:

“You’re never going to rely 100 per cent on what Percy tells you. I’ve still done that. But then what it’s allowed me to do is just ensure that the output that I’m creating is just that bit more accurate.”

Moreover, the skills AI supposedly replaces—repetition, analysis, admin—are precisely where juniors learn. Remove the “drudge work” and you risk hollowing out the learning curve. Julian Taylor, senior partner at Simmons & Simmons, warns:

“If we get too narrow at the entry point, I think that will create problems further down the line. It’s much easier to take your talent from your own trainees.”

The myth of AI-driven job destruction obscures a deeper issue: short-termism in workforce planning. Firms are experimenting, not transforming. And in doing so, they risk losing sight of the long-term value of training and development.

The Action: Reframe the narrative ✅

Instead of blaming AI, interrogate hiring strategy.

Try this:

• Before cutting junior roles, ask: what learning do they enable?
• Next time you hear “AI replaces X”, ask: what does it actually do—and what does it not do?
• Rather than fearing AI, use it to enhance rather than eliminate early career development.

If you thought AI was the villain, you may be surprised how often it’s just the alibi.

The Impact: A talent pipeline at risk 🚀

The real impact isn’t technological—it’s structural. Firms that pause hiring now will face a skills void in three to five years. The diamond-shaped org chart is a fantasy without a base. Apprenticeship models may return, but with troubling implications: fees, indentures, or long-term contracts that commodify training.

Isabel Berwick notes:

“Some people think that maybe that’s going to be a model for the future… a return of a kind of indentured labour.”

Meanwhile, graduates are questioning the value of university. If entry-level jobs disappear, what’s the point of a degree?

The kicker?

AI wasn’t the disruptor: it was the distraction. The real disruption is a failure to invest in people.

The Bottom Line: AI isn’t the problem. Blame poor strategy 📌

The conflict resolves simply: AI didn’t kill entry-level jobs. Poor strategic planning did.

☛ So let’s stop blaming the machines and start rethinking the model. What do you think—are we misreading the signals, or is the future of work already off course? ☚

A final thought: AI can (probably) not replace you in your job. But a salesman from an AI company with a billion dollar marketing budget can absolutely convince your boss that AI can replace you in your job….

About me: I help organisations turn complex data into clear decisions and commercial outcomes. My focus is on enabling better decision-making and unlocking new value through data-driven innovation – especially where the stakes are high and the problems are difficult and poorly defined.

Follow me on LinkedIn for more insights.

Because we treat data science like a production line, not a craft.

The Problem: Testing does not trump insights

Marketing was perhaps the first and the worst offender. Rigour in analysis and understanding were downplayed in favour of A/B testing, often with vanity metrics. I get it. Except the vanity metrics.

I may not have invented A/B testing but I did create¹ an approach to marketing called Customer Value Management (CVM) that was heavy on the automation while closely monitoring and optimising the short and long term financials. (The key innovations were the change process, the links to financials, and the processes for sustaining innovation.)

https://www.cybaea.net/CVM.html

It was great, for its time. We went from not using data to using data; from decisions based on hunches and whoever was paid the most to decisions based on data; from marketing vanity metrics to financial metrics that actually matter for the organization. We reliably increased profits by 5%, every year.

Too few companies today do even this.

Aside: Stop the data envy already

We were lucky to do this in mobile telecommunications. This was the industry that had loads of data on their customers and their behaviours. My definition of Big Data is that you have it when you are forced to throw away more data than you are able to keep. Mobile telcos with their detailed network information checked that box.

Other sectors were envious of our rich data, but that was the wrong obsession: They did nothing (very smart) with the data they had.

Having lots of data meant that we could try lots of things, but also that there is a strong temptation to ‘just do data’ and spend less energy on understanding.

Fast forward twenty years since that first CVM implementation at Vodafone Netherlands, and today every business has a lot more data than it used to possess.

And too many of them are still looking with envy at other businesses with more data, using that as an excuse for not doing enough with what they have.

Enough with the excuses already.

The insight: AI can do the testing work but not replace humans

Every business now has more data. And the “easy” data work—the simple A/B tests and vanity metrics that were once common in marketing—is being automated by #AI. Those jobs are mostly gone, and AI will replace the remaining ones.

So, how do you remain relevant?

And how do businesses gain true Return on Data? How do we deliver the impact that transforms our performance?

The Action: Rediscover craftsmanship

We remain relevant by embracing craftsmanship. By going beyond the algorithms to deeply understand the data, the problem, and the business.

If you are doing data science and you are not reading “Applied Machine Learning for Tabular Data”, then you are doing it wrong.

https://aml4td.org/, now available up to chapter 15.

It is a fantastic resource because it teaches you to go beyond and behind the statistics and algorithms to help you understand your data and thereby ultimately whatever the data is about. Say, your customers.

They treat machine learning as a holistic and practical craft, rather than a purely theoretical or algorithm-focused discipline. The emphasis is on the entire modelling pipeline, with a strong belief that the steps before and after model fitting are just as important, if not more so, than the choice of the algorithm itself.

This is not a ‘cookbook’ of oversimplified rules. It is not a ‘black box’ approach promoting the authors’ pet algorithm (or, worse, trying to sell you a proprietary one).

Instead this is for the pragmatic practitioner. It is a book about the craftsmanship of data science.

Aside: Learn from the best

You are doing it wrong, first because Max Kuhn is super smart² and someone from whom you want and need to learn.

Second, because the level of understanding of the data, the problem, and the algorithms you get from following the methodologies you learn here are exactly what you need to remain relevant in a world where #AI promises or threatens to automate the easy work.

Max comes from pharma and clinical trials which always required deep understanding. Unlike, say, marketing.

The Impact: Craft + curiosity finds opportunities and delivers change

Whether you read the book or not, please take this away:

• Getting the right insights from data is a craft that requires curious humans.

• Getting the right business outcomes requires humans to build connections, understand opportunities, and deploy solutions across the organization.

The Bottom Line: Curiosity didn’t kill the cat, it won the race

We include AI in all our products these days. Not to replace humans but to make us more curious and adventurous.

☛ How are you building your data craftsmanship to stay ahead of the AI curve? Share your thoughts below. ☚

About me: I help organisations turn complex data into clear decisions and commercial outcomes. My focus is on enabling better decision-making and unlocking new value through data-driven innovation – especially where the stakes are high and the problems are difficult and poorly defined.

Follow me on LinkedIn for more insights.

Here’s what I’m seeing go wrong – and more importantly, what actually works.

The Problem: Falling in Love with the Algorithm

Recently, I spoke with a claims director who’d spent 18 months building a sophisticated machine learning model to predict fraud. The model was technically brilliant with high accuracy in testing. But it’s been sitting unused for six months because nobody could figure out what adjusters should actually do with a “73% probability” score.

Decades ago, I spoke with a marketing director who had a brilliant model to predict that someone was “85% likely to churn” but had no tools or insights to retain them. Predictably, discounts were the easy answer, leading the industry to a race to the bottom. (See also: Commercial Churn Modelling.)

This is the classic mistake: starting with the technology instead of the decision.

The Insight: Start with the Action, Not the Algorithm

The most successful AI implementations I’ve seen follow what I call the “Decision-First Principle”:

1. Map the actual decision – What choice does someone need to make? By when? With what constraints?
2. Identify the decision bottleneck – Is it lack of information, too much information, or unclear trade-offs?
3. Design the intervention – What would change the quality or speed of that decision?
4. Then build the analytics to support that intervention

The Action: Three Questions Before Your Next AI Project

Before investing in any AI initiative, ask:

• Who makes the decision this AI is supposed to improve?
• What would they do differently if they had perfect information?
• How will we measure whether decisions actually improved?

If you can’t answer all three clearly, you’re not ready for AI – you’re ready for a conversation about decision-making. (Give me a call – I can help.)

The Impact: From Proof-of-Concept to Profit Center

Companies that start with decisions rather than data science see dramatically different outcomes. One client increased their underwriting speed by 40% not because they built a better model, but because they redesigned the decision process around what underwriters actually needed to know.

The AI was still sophisticated - but it was sophisticated for a purpose.

The Bottom Line: Great AI Projects Look Simple

The irony? The best AI projects often look deceptively simple. That’s because they solve real problems rather than interesting technical challenges.

What’s your experience with AI projects that worked vs. those that didn’t? I’m curious about the patterns you’re seeing.

About me: I help organisations turn complex data into clear decisions and commercial outcomes. My focus is on enabling better decision-making and unlocking new value through data-driven innovation – especially where the stakes are high and the problems are difficult and poorly defined.

Follow me on LinkedIn for more insights.

“Our data isn’t good enough for AI yet.”

I heard this phrase 23 times during our recent survey of Lloyd’s market leaders. It’s become the most sophisticated way to avoid making a decision about artificial intelligence - and it’s costing companies competitive advantage.

Here’s why the “perfect data” myth is more dangerous than messy data.

The Problem: Waiting for Data Perfection

Our survey found that insurance professionals consistently name “data quality” as their number one concern about AI implementation. On the surface, this sounds prudent. Responsible. Professional.

But here’s what’s really happening: companies are using data quality as a sophisticated excuse to delay decisions they’re uncomfortable making.

Meanwhile, their competitors are moving forward with imperfect data and learning faster.

The Insight: Humans Work with Messy Data Too

Here’s the insight that changed everything for one Lloyd’s managing agent we surveyed: “If human resource can be trained to cope with and adapt to those data vagaries and shortcomings, then so can AI – in time, probably better and faster.”

Think about how your best underwriters make decisions. Do they wait for perfect information? Of course not. They work with incomplete, inconsistent, sometimes contradictory data and still make profitable decisions.

The question isn’t whether your data is perfect. The question is whether AI can work with imperfect data better than your current processes.

The Action: The 80% Rule

Instead of asking “Is our data good enough for AI?” ask these three questions:

1. What decisions are we making with current data quality? If humans can make profitable decisions with your existing data, AI probably can too.

2. Where would 80% accuracy beat current performance? You don’t need perfection – you need improvement.

3. What would we learn by trying? The fastest way to improve data quality is to start using it systematically.

The Impact: Learning While Competitors Wait

One survey respondent described using AI for classifying claims information and identifying trends in pricing and reserving work. Not because their data was perfect, but because even imperfect pattern recognition was better than manual processes.

Result? They’re not just getting operational benefits – they’re discovering data quality issues faster and fixing them systematically. Their data is actually getting better because they’re using AI, not despite using imperfect data.

Meanwhile, their competitors are still in “data preparation mode.”

The Bottom Line: Perfection is failure

Data will never be perfect. But waiting for perfection is a perfect way to fall behind.

What decision could you improve with 80% of the data you wish you had?

Why? Because they’re asking the right questions.

The Problem: The AI Adoption Paradox

Everyone’s talking about the competitive advantage of AI, but here’s what our survey revealed: half the market is represented by firms with limited or no AI implementation.

It is interesting because it’s not that they don’t see the potential. Actuarial professionals are actually more optimistic about AI because “the quantitative nature of actuarial work means that many in the field see the opportunities and benefits of AI and ML tools as more tangible.”

So what’s holding them back?

The Insight: Fear of Bad Decisions, Not New Technology

The survey uncovered three critical concerns that smart executives are wrestling with:

1. Validation Anxiety: How do you validate an AI output when you don’t fully understand how it reached that conclusion?
2. Regulatory Compliance: Using third-party AI tools raises compliance concerns that “has held back the pace of adoption.”
3. Data Reality: Models are only as reliable as the data they’re trained on – and most companies know their data isn’t perfect.

These aren’t irrational fears. They’re legitimate business risks that require thoughtful solutions.

The Action: The Smart Adoption Framework

Based on our research, here’s how leading firms are moving forward:

Start with augmentation, not automation: Use AI to supplement human decision-making rather than replace it entirely. Classification, trend identification, and pattern recognition are proving to be sweet spots.

Invest in interpretability: The need for “transparency and interpretability” isn’t just regulatory compliance – it’s business intelligence. If you can’t explain the recommendation, you can’t improve the process.

Accept imperfect data: As one respondent noted, if human resource can be trained to cope with and adapt to those data vagaries and shortcomings, then so can AI – in time, probably better and faster.

The Impact: Competitive Advantage Through Smart Risk-Taking

The firms that will win aren’t necessarily the fastest adopters – they’re the smartest adopters. They’re building AI capabilities while maintaining the risk management discipline that makes them successful insurers.

As Sanjiv Sharma from the LMA put it: Firms that successfully integrate AI and ML into their actuarial and risk functions and beyond, can gain a competitive edge and make more informed strategic decisions.

The Bottom Line

The best insurance executives aren’t avoiding AI – they’re approaching it like underwriters: assessing risk, managing exposure, and making calculated bets.

☛ What’s your biggest concern about AI implementation in your organization? Validation, compliance, or data quality? ☚

Most return to work in the office (RTO) policies are simply the visible outcomes of repeated failures of leadership.

There is nothing inherently wrong with working in an office and companies have every right (within legal limits) to arrange their activities in the way management thinks is best. Equally, employees have every right to reject working for companies that doesn’t work for them.

My issue with RTO is that most of it is just lazy leadership.

“Come to the office to keep our culture strong,” they say.
“Come to the office and you will learn more,” they say.
“Come to the office and you will be more productive,” they say.
“Come to the office and you will build strong networks,” they say.

This puts so much down to chance rather than intention.

Culture emerging by the invisible, random force of office osmosis was, is, and will remain, a terrible idea and a failure of leadership.

Good leaders are deliberate about developing the company’s culture. They see it as a core part of their job. Relying on chance is a failure of leadership. The office doesn’t make culture, people do.

Some skills are best acquired through direct, hands-on experience, much like apprenticeships. If your organisation has structured on-the-job learning led by experienced professionals guided by a deliberate development programme: good for you. And for your staff. Some jobs need to be done in specific locations.

But for the average knowledge worker hotdesking in a noisy open plan office, there is no apprenticeship and minimal structure to their learning. They just have to hope that they bump into the right person at the right time; hope that they can learn and connect; hope that someone, anyone, will coach and mentor them.

Learning by serendipity is no way to develop staff. Your most important job as a leader is to help your people grow. Learning and development is not something an underfunded, unloved department in HR does. It’s your job. Leaving it to chance means you are failing.

As for productivity – even if we limit it to productivity for the company, ignoring all of our other commitments and how some workers find it easier to juggle their responsibilities as carers, find time for their families, or manage their neurodiversity when working from home – for knowledge workers the busy open plan office is a terrible place for any sort of deep work.

These workers may be more busy in the office, but they are unlikely to be more productive.

Running your business on busyness instead of true productivity is an abject failure of leadership. It is not the office versus remote working that’s the problem. It’s you.

By all means get everyone back to the office. Or go hybrid. Or fully remote. Whatever. Work on Mars if you want. Just don’t be lazy and think that any one of these choices will somehow magically fix your culture, your learning, your productivity, or any other issue.

Effective leadership is proactive and intentional, not lazily reliant on outdated norms.

• Good leaders intentionally develop company culture and skills through structured programs, not by relying on random office interactions. If you are in the office, how are you intentionally going to use the office to develop culture; if working flexible, how will you intentionally use that flexibility to grow your culture?
• For knowledge workers, learning and productivity are not automatically improved in noisy, open-plan offices with frequent interruptions. How can you be intentional about using different environments at different times and stages in people’s development or projects?
• Leadership should focus on deliberate development and true productivity, developing the business and not just busyness.

Most offices are terrible; they don’t earn the commute.

I happen to like the office. In moderation. But most offices are terrible. Aditya Sanghvi, a partner at McKinsey, says it well: “To beat the working-from-home alternative, [office] spaces should ‘earn the commute,’ and you do that by creating space where people actually want to work, not just the space where they have to work.”²

None of this is easy, but it’s not exactly new either. Let’s share our stories in the office one of these days. Or virtually. Or in a coffee shop. It’s all good.

Follow me for more insights on delivering commercial value and growth in the data economy.

Our calculations show that NI on pensions contributions would add over 1% to total expenditure for the average school. Keep in mind that the average school operates a 4% surplus to allow for future maintenance and development, so this one measure alone would take a quarter of the surplus. The calcullation is based on the schools on Data Navigator for Schools database and assumes the current NI rate of 13.8%. Staff costs typically account for 60-70% of the total expenditure for an independent school and schools typically pay generous pensions contributions compared with other private sector small-medium enterprises.

At a time like this, being able to make data-backed decisions and benchmark yourself to your peers is a significant advantage and can provide reassurance for strategic choices.

Data Navigator is a free tool for independent schools. It is developed by Barnett Waddingham in collaboration with Independent Schools’ Bursars Association to enable bursars to be able to view and compare their financial performance to other schools and explore different forecasting scenarios.

For example, bursars can use the forecasting element of the Data Navigator to view how an increase in pensions costs might affect their overall finances – and more importantly, explore the subsequent impact of different mitigating strategies. The example below shows how the effect of a 13% increase in pension costs requires an offsetting 2% decrease in teacher salaries to remain budget neutral. The result on the school’s overall surplus gives an indication of how these changes might affect its financial performance, both in isolation and compared with the chosen benchmark group.

One of the great strengths of the Data Navigator is that it allows you to choose which schools to benchmark against – would you prefer to compare yourself to: schools near you, schools of a similar size, schools in an urban location, or some combination of all three? The Data Navigator allows you to select the characteristics of your benchmark group and view all your data alongside this comparator cohort.

Below you can see the distribution of employer pension contributions for teachers, for schools in the Data Navigator. The average rate (indicated by the red triangle) is 16%, which has been shown to be representative of the sector. A bursar seeing this could choose to compare themselves to a subset of these schools and use that to drive strategic decision-making. In the example below, the Demo School has an above average contribution rate and can conclude they might be relatively more impacted than their peers if NI were introduced on these contributions.

Any increase in staff costs may also require bursars to review their overall employee benefits and look for alternative ways to attract and retain talented staff. Barnett Waddingham supports schools and other businesses with the management of employee benefits, benefit design and implementation, and communications and offer an engagement platform with a range of flexible benefits which fit the needs of your teachers and support staff. Read more here.

If you are not yet part of the Data Navigator, you can click here for more info or book a call with one of our friendly colleagues to talk through our pre-population service and the benefits the Data Navigator offers.

Thanks to Chris Brownlie for substantial help on this article.

Follow me for more insights on delivering commercial value and growth in the data economy and for information for the independent schools sector.

• You have to watch LEADERSHIP LAB: The Craft of Writing Effectively.

• One insurance firm employs “explainability engineers” to help understand the outputs of ai systems. Source: How businesses are actually using generative AI from The Economist.

• I tell my new consultants, in my usual over-dramatic style, that, Your job, your only job, is to get your client promoted. Jim Stevenson has a more nuanced version of this as number four in a recent interview with Authority Magazine: Jim Stevenson Of Bletchley Group: Five Things I Wish Someone Told Me Before I Started My Consulting Business

• Why did I not know about https://www.openculture.com/ before?

• Mastering Real-Time Decision Making by Annie Duke is a good watch. She says some of the things I say, such as her version of When a decision is hard, it’s easy.

Reading, watching, listening

Interesting business stuff this week:

• Listening to The Professional Services Pursuit Ep. 61 - Navigating the Evolving Landscape of Consulting: From Hourly Billing to Products and Partnerships w/ Tom Rodenhauser

Interesting tech stuff this week:

• My colleague Chris Brownlie published This week in R.

What I am reading now:

• Bulletproof Problem Solving - The one skill that changes everything by Charles Conn and Robert McLean (Amazon paid link UK | US) is the new book we are studying in our team

• Scaling People - Tactics for management and company building by Claire Hughes Johnson. Stripe Press.

Until next week: Travel well, whichever road you are on 👋.

• I attended the TBD Conference and learnt that I have no appetite for virtual conferences any more. Not even one as good as TBD. Or, perhaps: especially not one as good as TBD.

• I think I learnt that there is a lot of angst in the world, perhaps leftover from Covid and built up before. Several of the TBD talks were of the “let’s all stick together” variety, which is different from previous years.

• I learnt more about AI in education though the short version seems to be, “Not yet, but watch this space.”

Reading, watching, listening

Interesting business stuff this week:

Conference, health issues, and a busy week at work meant that this was not a good week for learning.

Interesting tech stuff this week:

What I am reading now:

• Value Proposition Design by Strategyzer. Re-reading this with the team. The book is ten year’s old and still be most practical, visual, and downright enjoyable guide to the subject I know. We are going to finish it on Monday.

• Scaling People - Tactics for management and company building by Claire Hughes Johnson. Stripe Press.

Until next week: Travel well, whichever road you are on 👋.

• This was a good week for learning.

• I learnt about languishing from Adam Grant: How to stop languishing and start finding flow | TED

• I learnt that You can’t build a billion dollar despicable me startup from Alexander Jarvis’ article Sole founders are bad which is linked from Microsoft’s official training on how to build a startup.

• I enjoyed the The Wisdom Tree which I found out about from a post by Eddie Cheng but I have referenced the original A Short Guide to Reading and Learning for Investors by Vishal Khandelwal over on Safal Niveshak.

• I learnt what is a Zibaldone and as far as I am concerned anything that connects H.P. Lovecraft with 14th century Venice and Thomas Jefferson is good. Very good. How to Keep a Zibaldone, the 14th Century’s Answer to Tumblr

• I learnt that the EU Court of Human Rights Rejects Encryption Backdoors because backdoors violate human rights (!). The “confidentiality of communications is an essential element of the right to respect for private life and correspondence,” the ECHR’s ruling said. Thus, requiring messages to be decrypted by law enforcement “cannot be regarded as necessary in a democratic society.”

Reading, watching, listening

Interesting business stuff this week:

The Coaching Habit by M.B. Stanier (Amazon paid link: UK | US).

This is a terrible book: the binding is awful, pages are falling out, the paper quality is rubbish. The contents is good – a solid introduction to thinking like a coach in the workplace – so that’s something. Thank you Christine Kerr for the recommendation.

Interesting tech stuff this week:

• Why pandas feels clunky when coming from R by Rasmus Bååth who five years ago switched from R to python at work.

What I am reading now:

• Value Proposition Design by Strategyzer. Re-reading this with the team. The book is ten year’s old and still be most practical, visual, and downright enjoyable guide to the subject I know.

• Scaling People - Tactics for management and company building by Claire Hughes Johnson. Stripe Press.

Until next week: Travel well, whichever road you are on 👋.

• I think I will take a break from Weeknotes until I find an alternative to Quarto for the website, as mentioned last week.

• This means that when leaders become busy, the indifference becomes mutual. A reminder to be present, always. Source

• Epic. Demo here

• Music on heavy rotation Spotify playlist from Tobias van Schneider.

Reading, watching, listening

Interesting business stuff this week:

Interesting tech stuff this week:

What I am reading now:

• Value Proposition Design by Strategyzer. Re-reading this with the team. The book is ten year’s old and still be most practical, visual, and downright enjoyable guide to the subject I know.

• Scaling People - Tactics for management and company building by Claire Hughes Johnson. Stripe Press.

Until next week: Travel well, whichever road you are on 👋.

• Never sell what you do. Sell the outcome of what you do. Source

  • Allan’s corollary: Never sell your time. It doesn’t scale and it is the one thing you never get any more of.
  • Time is not a commodity for which anyone can charge, as the angelic doctor apparently argued (though I can not find the primary source).

• I am not a big fan of standing at vendor exhibition stands at conferences. However:

  • Bursars of independent schools are a small community and it is great to see the people.
  • Having a stand is worthwhile, possibly essential. It is just if I need to be there with a grin on my face.

• I think From St Albans With Love was my first live event since Covid…. We had a great time in support of a good cause.

  • The event included a lovely (pun intended) reading of “This is what love is” from Captain Corelli’s Mandolin. Here it is on YouTube with John Hurt and Penélope Cruz. Love itself is what is left over when being in love has burned away.

• This made me very happy this week. #Zoozve

• This made me very worried that we have learnt nothing from Covid.

Reading, watching, listening

Interesting business stuff this week:

The conference consumed my time.

Interesting tech stuff this week:

• Be a data visualisation friend. Friends don’t let friends make bad graphs. Example:

What I am reading now:

• Value Proposition Design by Strategyzer. Re-reading this with the team. The book is ten year’s old and still be most practical, visual, and downright enjoyable guide to the subject I know.

• Scaling People - Tactics for management and company building by Claire Hughes Johnson. Stripe Press.

Until next week: Travel well, whichever road you are on 👋.

• Last week I moaned that Quarto was too slow for blogs. Can Hugo rescue me?

  • The consensus in open source community seems to be to use Hugo, “The world’s fastest framework for building websites”, for the elements of the site that do not require running code.
  • There are instructions on how to use Quarto with Hugo.
  • I have some reading and experimenting to do….

• Using Levels of Work (what Elliott Jaques called Modes) to re-evaluate the organisation.

• Re-reading Value Proposition Design with the team. Timing is right as we are pivoting our effort from building the solution to communicating its value.

• Ended the week with ten meetings, down from eleven last week. Friday is my ‘meeting day’, so it is fine.

Reading, watching, listening

Interesting business stuff this week:

• This video is great: Master the Business Model Canvas.

• Good introduction to Levels of Work: A hierarchy required by nature.

• Now is the time to create new corporate ventures. McKinsey thinks there are four ‘unlocks’ for success:

  1. Talent: get the right team, a diverse team, a team of talent that is going to get their hands dirty and really deliver and do things.
  2. Sponsorship: A new venture does need to be CEO- and board-sponsored. That helps attract attention and helps ensure they’ll not just fund it for a quarter and then give up to save money.
  3. Governance: getting the right governance in place and setting up your venture close enough to the corporate to exploit all those competitive advantages you described earlier. You need access to the brand, customers, and some of the partners.
  4. Distance: at the same time, you need to be far enough away so you can move at the speed and pace of a start-up, so that you can test and learn. Getting that right is easy to say, but harder to do.

Interesting tech stuff this week:

• Very satisfying: The source of the cake dataset. Unleash your inner data sleuth.

What I am reading now:

• Value Proposition Design by Strategyzer. Re-reading this with the team. The book is ten year’s old and still be most practical, visual, and downright enjoyable guide to the subject I know.

• Scaling People - Tactics for management and company building by Claire Hughes Johnson. Stripe Press.

Until next week: Travel well, whichever road you are on 👋.

• Rescheduled team away-day in London next week due to train strikes - bummed; it would have been great to get together.

• Upgraded to Quarto 1.4. This website still takes way too long to render. If anything, it has gotten worse with this version. Over 16 minutes to render 215 pages. I need to find another system for managing this site.

• This quote resonated: Perhaps part of breakthrough leadership is talking to others like giants. Expecting the most and holding others to it, trusting that they’ll rise to the challenge. Never backing down. Standing for the giant in us all. (From: Gap International)

• “When a measure becomes a target, it ceases to be a good measure.” Making a long list of personal development metrics to track, and am conscious of Goodhart’s Law and weary of McNamara’s Fallacy

• This is brilliant: The emotional arcs of stories are dominated by six basic shapes. The authors use 1,327 stories from Project Gutenberg to plot the six basic stories.

Reading, watching, listening

Interesting business stuff this week:

Britain, consultation nation (FT paywall - use search to bypass). Yo! It’s us. Management consultancy is Britain’s fourth-biggest export.

Interesting tech stuff this week:

ML + XAI → Strong GLM. This is so good. I always did this “by hand” and while this is not fully automated, it is a great toolbox. See also the published paper SHAP for Actuaries: Explain any Model. I need to do more with this approach.

What I am reading now:

Scaling People - Tactics for management and company building by Claire Hughes Johnson. Stripe Press

Until next week: Travel well, whichever road you are on 👋.