Executive Review - From Formula One to Generative AI: Lessons for the Modern Data Leader
From AI Hype to AI Reality: What Changes When the Constraints Change?
AI may be dominating boardroom conversations, technology roadmaps and investment decisions, but a recent Data Leaders Lounge discussion raised a more useful question:
What happens when AI changes the constraints our organisations have been built around?
Across two opening perspectives and a wide-ranging peer discussion, one theme kept resurfacing. Technology is moving quickly, but successful transformation is still determined by data, judgement, culture and an organisation’s ability to learn.
Huge thanks to our guest host and speakers who inspired some really illuminating discussion! Peter Henderson, Martin Blow and Geoff McGrath - thank you.
In brief:
AI is changing the economics and operating model of the data office, but it isn't removing the need for strong data foundations or human expertise. As AI becomes more accessible, competitive advantage may increasingly come from an organisation's proprietary data, context, domain knowledge, culture and ability to turn technology into better decisions.
Here are some of the ideas that stood out.
Constraint can be the catalyst for innovation
Many of the innovations we now take for granted were born from constraint.
The discussion explored examples ranging from flat-pack furniture and internet video to space technology, before bringing the principle back into the modern data office.
For years, data organisations have operated within constraints: storage, compute, bandwidth, budgets and, perhaps most importantly, people and capacity.
Those constraints have influenced how data teams operate. Organisations prioritise critical data, concentrate effort on the highest-value use cases and focus limited resources where they can have the greatest impact.
But AI and automation have the potential to shift some of those constraints.
If activities that previously depended on human capacity can increasingly be automated, organisations may be able to manage and extract value from a much broader proportion of their data estate. The result could be a change not simply in how much data teams can do, but in what a data office is there to do in the first place.
AI is changing the economics of the data office
A particularly interesting argument was that AI is effectively
repricing the data office.
It potentially changes both the cost of delivering data services and the value organisations expect to receive from them.
Traditional value cases have often centred on revenue, cost optimisation and operational efficiency. The discussion suggested that
speed, agility and time to value are becoming increasingly important measures too.
At the same time, expectations of data teams are expanding. Metadata, semantics, context, model evaluation and other capabilities are becoming increasingly relevant as organisations attempt to put AI into production.
The challenge is therefore not simply "how do we implement AI?"
It is:
How does the operating model of the data function need to change when AI becomes part of its everyday infrastructure?
AI-ready data still matters
One of the liveliest parts of the discussion challenged the idea that advances in AI somehow remove the need for good data foundations.
The consensus was very much the opposite.
AI may make it easier to access and work with structured and unstructured information, but context, semantics and business logic remain hugely important.
Participants shared examples of AI proofs of concept bypassing traditional data preparation and relying instead on poorly structured information, producing results that subsequently needed validation by experienced data teams.
As one contribution effectively framed the challenge: if an AI system says something unexpected, organisations still need the underlying foundations to understand why.
The conversation also raised the possibility of AI creating a new generation of "shadow IT": solutions created quickly because the barrier to entry has fallen, but which may leave organisations with governance, maintenance and quality problems later.
The takeaway was not that AI reduces the importance of data management.
Arguably, it increases it.
Don't automate judgement just because you can automate activity
Another strong thread concerned where agents should, and shouldn't, operate autonomously.
The experience shared in the room suggested that tightly bounded agents performing specific tasks can be easier to evaluate than agents given broad responsibilities. Auditability, lineage, change control and the ability to understand what has changed were all highlighted as increasingly important considerations.
But there was also a challenge to the assumption that "human in the loop" automatically equals good governance.
A human approval stage only adds value if meaningful judgement is genuinely being exercised.
This becomes important as businesses decide which activities can safely become automated, which require oversight and which fundamentally depend on human expertise.
What decades of high-performance engineering can teach us
One of the most compelling parallels came from high-velocity engineering.
Formula One provided the starting point.
A racing car was described not as a finished product, but as a prototype in continuous development. Performance is instrumented, feedback is rapid, and learning is fed directly back into the next iteration.
The same philosophy was described across engineering, energy and major infrastructure environments:
build, measure, learn, adapt.
In high-performance environments, data is useful because it improves the quality and speed of the next decision.
That shifts the question from:
"What does the dashboard tell me?"
towards:
"What should we do next?"
The discussion also highlighted something technology programmes can easily overlook. These high-performance environments don't simply work because they have better technology. They combine technology with delegated decision-making, trust and expertise close to the point where the information is richest.
The next competitive advantage may be context, not the model
There was also considerable discussion around large language models versus smaller, more specialised models.
For organisations with deep proprietary knowledge, competitive advantage may increasingly come from domain knowledge, organisational context, business logic and proprietary data, rather than simply having access to the newest general-purpose model.
Participants discussed the potential for smaller or locally focused models in engineering and highly constrained environments where compute, power, latency or connectivity matter.
The wider implication is significant.
As models become increasingly interchangeable, the differentiator may become what an organisation can wrap around them: its context, data, processes and expertise.
And then we came back to people
Perhaps appropriately, the discussion ultimately moved beyond technology.
A question from the room captured it well:
Isn't this ultimately about culture?
Technology might enable change, but organisations still need people who know when to trust it, when to challenge it and how to respond when something happens that nobody anticipated.
Experience, judgement, trust, curiosity, tenacity and the ability to work through ambiguity all featured heavily in the conversation.
Rather than viewing AI purely as a replacement for people, the discussion increasingly positioned it as augmentation and support. Routine activity can move towards technology, while the value of genuinely human judgement may actually increase.
One particularly useful distinction emerged: when somebody wants a routine service, automation may be exactly what they need. When they need a person, however, they want a genuinely human experience, not simply another process interface.
Five questions leaders should be asking
For us, the conversation left five questions worth taking back into organisations:
- Which constraints have shaped the way our data function operates today, and do they still exist?
- Are we genuinely AI-ready, or are we putting increasingly powerful technology on weak data foundations?
- Where can automation safely replace activity, and where do we still need human judgement?
- Are we measuring AI by adoption, or by the speed and quality of decisions it enables?
- If models become increasingly commoditised, what proprietary data, context, domain knowledge and human expertise will differentiate us?
Perhaps the biggest takeaway is that the AI conversation is moving on.
"How should our organisations work differently now that it can?"
And answering that question may have considerably more to do with people, culture, data and operating models than with the technology itself.
AI, Data and the Future of the Data Office: Key Questions
How is AI changing the role of the data office?
AI has the potential to automate more routine elements of data management and delivery, changing both the cost of providing data services and the speed at which they can operate. The discussion suggested this could move data teams towards more dynamic services, greater automation and a stronger focus on context, evaluation and decision support.
Does AI reduce the need for structured, high-quality data?
No. The discussion strongly challenged the idea that AI removes the need for good data foundations. AI can work across structured and unstructured information, but context, semantics, business logic and data quality remain essential for reliable outputs and for understanding what has gone wrong when an AI system produces an unexpected result.
What does “AI-ready data” mean in practice?
The discussion positioned AI-ready data as more than simply making information available to a model. Organisations need reliable data alongside the metadata, semantics, context and business knowledge that help AI systems interpret that information appropriately.
Will AI agents replace data teams?
The discussion did not point towards the disappearance of data professionals, but towards changing roles. Routine services may increasingly be handled by agents, while people remain important for design, critical thinking, domain expertise, trust and judgement.
Where should organisations use AI agents?
A recurring view was that agents work particularly well when their remit is tightly defined. Narrower tasks can make outputs easier to evaluate and control, while judgement-heavy activities may still require people. Participants also highlighted the importance of audit trails, evaluation and change control as agent use increases.
Are smaller AI models becoming more important?
Potentially. The discussion explored environments such as robotics, engineering and edge computing where bandwidth, power and processing constraints can make smaller, specialised models particularly relevant. It also raised the possibility that proprietary models built around an organisation's own data and domain expertise could become a source of competitive advantage.
What will differentiate organisations as AI models become more accessible?
One of the strongest themes was that competitive advantage may increasingly come from context rather than the model itself: proprietary data, semantics, business logic, workflows, domain expertise and the organisation's ability to apply them effectively.
What is the biggest barrier to successful AI adoption?
Technology is only part of the equation. The discussion repeatedly returned to culture, trust, operating models, human judgement and an organisation's ability to learn and adapt. As one exchange put it, the challenge is increasingly about how organisations learn to use and deploy AI across their businesses, rather than simply gaining access to the technology.
What can business leaders learn from high-velocity engineering?
The engineering examples centred on continuous learning: instrument the product, understand how it performs, feed that information back into decision-making and shorten the cycle between learning and improvement. The principle is to use data and technology to improve the speed and quality of the next decision, rather than simply producing more information.
Be part of the next conversation
How is your organisation turning AI ambition into practical change? Continue the discussion with the Eden Smith Data Leaders Executive Lounge, connecting with fellow senior data and AI leaders to share experiences, challenge thinking and learn from one another.
Register your interest in joining our community to receive invitations and updates, or explore our upcoming events to find your next opportunity to connect.
