top of page
Search

Leaders Who Don’t Value Their Data Will Struggle to Adopt AI

Writer: aiomniconversation
aiomniconversation
6 days ago
5 min read

By Professor Ashley Braganza


Many boards can explain how they make money, but not their data quality. Ask a leadership team to map its revenue sources and it will have a good attempt. Ask it to sketch the organisation’s structure and the people who matter, and it will manage that too. Ask it to describe the organisation’s data architecture, and the reply is a question of its own: why do I need to know that? That reply says more about an organisation’s readiness for AI than any count of licences or pilots. It describes a culture in which data is someone else’s problem.

My guest on The AI Adoption Podcast is Claire Thompson, Chief Data Analytics Officer for the Corporate Bank at Barclays. Claire has worked in data and analytics across several financial services organisations, each at a different stage of data maturity. 

The data shrug is a cultural verdict

Claire’s starting position is that AI is primarily a people and a culture challenge. The technology creates the opportunity; the culture, she says, determines how much of it becomes part of everyday work and how much value is realised. Few would quarrel with that, any more than with the hackneyed line that data is the oil of AI. The two ideas are usually held apart. My argument is that they belong together.

Senior leaders set the tone, norms and mores of an organisation. Culture, in practice, follows the things they pay attention to. When a leadership team can describe revenue and structure but treats data architecture as a technical matter beneath its notice, staff are likely to draw the obvious conclusion. Data becomes something the systems produce rather than an asset the business depends on. People stop connecting the revenues and profits they generate with the data that makes them possible, and data quality drifts accordingly.

Claire is clear about the cost. Without data of good quality, clear ownership, an understanding of where it comes from, and security that limits access to what each role needs, AI “can make for really unreliable outputs and also make scaling really hard”. She adds that AI will amplify organisational gaps and challenges. A culture that tolerated poor data when it fed monthly management reports cannot tolerate it once the same data feeds tools in employees’ hands across the business. The shrug in the boardroom becomes an error rate on the front line.

Investment in data does not cure a culture that ignores it

The strongest objection to this argument comes from Claire herself. When I put it to her that a culture which does not value data would slow AI adoption, her answer was measured: “Yeah, potentially, potentially.” Organisations, she pointed out, sit in very different places. A fintech would manage data differently to an organisation with a long legacy to deal with. In her experience, most organisations that she’s worked in have invested in data.

That is a fair corrective, and it sharpens the argument rather than weakening it. Investment and culture are different things. An organisation can fund data platforms and data teams for years while its leadership remains unable to describe the data structures it paid for. The questions Claire would have any organisation ask are diagnostic: is the obstacle the amount of investment, investment in the wrong places, missing skills, untrained teams, or senior leaders not ready for the change? Those questions have different answers in different organisations, and she is right to resist a single diagnosis.

Breadth changes the calculation. Claire notes that AI is no longer confined to specialist teams; it is being given to everyone in the organisation through different tools. When data work sat in a specialist function, leaders could delegate data to that function and get away with it. Once AI reaches desks across the organisation, the data culture of the whole organisation shows up in the outputs. Past investment in data is necessary but is unlikely to be sufficient.

A portfolio of pilots is not an AI strategy

The habit that leaves data unowned produces a second failure: adoption by accumulation. Some organisations have bought licences for everyone, and the result is a scatter of local use. Claire does not dismiss this. Personal productivity, she argues, is a good way for people to learn, and enthusiasts who push the boundaries are an asset. Her objection is to mistaking the sum of these efforts for AI strategy. Disconnected pilots are useful in driving some value, but fall short once you consider strategic value to the organisation”.

Claire’s point on governance is about timing. It is much easier to put the right governance in place at the beginning than it is ‘once the horse is bolted’, with use cases and agents sprawling across the organisation and little or no governance in place. The risk is that a tool built locally becomes critical to a process without anyone noticing, and then it fails.

AI arrives by three routes: tools placed in everyone’s hands, capability vendors build into their applications, and bespoke models built in house. Each needs its own controls, from sharper procurement questions for vendors, to a triage step before a personal tool is shared with 100 people, to monitoring and a backup plan for in house models in production. Claire can see a point where people manage teams of agents, though organisations are still working out how. Moreover, whether AI agents, which by Claire’s own definition offer a best answer rather than a certain one, can run a critical process with the reliability a team member needs is yet to be resolved.

AI readiness starts with who sits at the leadership table

Claire’s practical advice begins with the leadership table. For a team that has taken off the rose-tinted spectacles and judged its data maturity low, her first question is direct: who around this table is our expert? The more that an organisation has individuals around the senior leadership table that really understand the AI data technology, the more likely it is that you’re going to succeed. Without someone in the room who can hear the priorities and propose a different solution, it’s very difficult.

The second prescription is personal. Claire asks leaders to take a real interest, get curious, and to go and speak to their data and technology teams. She also asks them to model AI use visibly, telling the story of how they use it in their own jobs so the message becomes “it’s not a theory, this is how I’ve done it”. That, she argues, moves people from curiosity to confidence. The same logic applies to data. A leader who talks about data with the fluency they bring to revenue signals that data belongs to the business.

The third is foundations: data of good quality, clear ownership, data products people can find intuitively, and security so people reach only the data their role requires. This is a point I put to boards and senior managers I work with. A leadership team that cannot describe its data cannot judge whether that data is ready for AI.

Claire’s framing ends where it began, with people and culture. My challenge to leadership teams is narrower and harder. Before the next pilot is approved or the next licence bought, ask each member of the executive committee to sketch the organisation’s data architecture with the confidence they bring to its revenue model. If they cannot, the organisation’s AI ambitions are likely to outrun its data, and the gap will show first in outputs nobody trusts. The fix starts in the boardroom, not in the data team.


Listen to the full conversation

If you found this useful, you will also find these related conversations of interest

 

 
 
 

Comments

bottom of page