It’s a topic that’s been cropping up a lot in Data Culture Club discussions recently: not only is AI transforming how people work with data, it’s reshaping the skills, career paths and organisational responsibilities that underpin a strong Data Culture. So we dedicated our latest London event to a discussion of the challenges our members were facing in their own organisations.
The questions behind the conversation
There were a few different aspects we wanted to investigate:
- Everyone is having to learn new capabilities – although with AI replacing technical know-how, and natural language processing overtaking prompt engineering, should we now be focussing our training on the critical thinking and softer skills required to frame the question and challenge the answer?
- Talent paths are also shifting, with fewer opportunities for junior workers to learn their craft and experienced staff finding their hard-earned knowledge becoming obsolete or getting rusty. And this affects the whole organisation – so where do we draw the line between Data Culture and HR responsibilities?
- Meanwhile, data is becoming more important and less transparent. AI allows more people to analyse data, but access doesn’t create understanding. Tracing data lineage to comply with regulation is more complex when AI is synthesising information rather than simply retrieving it. Will the initial excitement for AI result in frustration with the underlying data management or give it a boost?
The event kicked off with a lively fireside chat featuring Data Culture Club member Dan Whiter (Compare the Market) and guest Charlotta Rames (Morgan Stanley), both of whom are on the ground supporting their respective firms’ AI initiatives. The conversation continued with a roundtable discussion where our members compared experiences, challenged assumptions and shared some wonderfully practical examples.
What emerged was a fascinating mix of excitement and concern. AI is opening up opportunities to work faster, tackle more interesting problems and make knowledge more accessible. But it also raises some uncomfortable questions about the skills we might lose, the expertise we need to preserve, and who’s responsible for helping people navigate the change.
The seven key themes that stood out
1. If AI does all the work, how do we learn to do it ourselves?
Charlotta shared how AI is already changing the nature of work in investment banking. Previously, junior employees might have spent hours processing data, preparing presentations and identifying trends. Now AI can do much of that for them, freeing them up to focus on more interesting analysis.
But there’s a catch: how do you develop the knowledge to create sensible AI prompts – and the judgement to challenge the outputs – if you’ve never done the work yourself the ‘old-fashioned’ way?
We concluded that there is value in completing a task manually a few times before handing it over to AI, whether that’s creating meeting minutes, writing Python code, or building dashboards. Dan suggested taking this a step further: do the work yourself, ask AI to challenge it, and then ask AI to improve on your version. That way, the technology becomes a learning partner rather than simply a shortcut.
In general it does seem helpful to think of AI as something to collaborate with, rather than delegate to. For example, one participant remembered how intimidating it felt to send their first professional email, agonising over the wording before clicking ‘Send’. Today’s graduates can ask Copilot or ChatGPT to produce something polished in seconds – but while AI might know how to write a business email, does it know how to write an email for your business, to your colleagues, in a way that sounds like you and builds the right relationship? Maybe one day AI will be able to do it all; for now the group recommended using AI to get 80% of the way towards a finished product, leaving the final 20% for human editing, tailoring and personalisation.
The takeaway: The technical skills that Data Culture programmes teach are still needed - but exactly what level of expertise is required has become more nuanced because we're now using them in conjunction with imperfect AI assistants and agents. We can use AI to help with both the learning and the 'doing' without losing the human checking and finessing.
2. Are we inadvertantly removing the first rungs of the career ladder?
The conversation then moved to a broader concern: what happens to career development when AI automates the routine tasks traditionally given to junior employees?
At a time when the news is full of reports of organisations reducing their numbers of interns, our speakers argued that graduate programmes remain vital pipelines for future talent. Not only can junior employees now contribute at a higher level much earlier in their careers, they’re also bringing to the organisation a significant amount of AI nous. One member described asking a young colleague, just a year out of university, to give her an informal introduction to AI tools – and he enthusiastically demonstrated capabilities she’d never encountered through her own company’s training.
Where entry-level staff still need guidance from experienced professionals is in understanding the business context, and the underlying assumptions and risks. And they may find progression up the ranks becoming more difficult. Dan described how he expects the traditional career pyramid to change into something more like an hourglass shape, with a broad base of junior talent and strong capability at the top, but potentially fewer roles in the middle. That raises questions not only about middle management, but also about how people develop the experience to become senior decision-makers.
The discussion also touched on the impact of remote working on early-career development. Junior employees have missed out on the informal learning that happens through overhearing conversations, observing senior colleagues and interacting with clients. AI could compound that challenge if it removes even more opportunities to learn through observation and experience.
The takeaway: Career paths have become less linear, and it's a false economy to use AI to replace human resources. Data Culture and HR leaders need to rethink how they develop future experts, deliberately creating opportunities for learning and human interaction rather than assuming these will happen naturally.
3. Human skills are becoming even more valuable
With AI now routinely used to handle administrative and technical tasks, there’s a shift in the types of training that employees are requesting.
A popular topic at the moment is critical thinking (learning to evaluate evidence, question assumptions and identify biasses, to improve problem-solving and strategising). As is data storytelling (to effectively communicate ideas and insights), along with creativity, curiosity and good judgement.
These ‘soft skills’ – rebranded ‘power skills’ by some organisations to reflect their significance – are either helpful for working effectively with AI, or purely human abilities that AI cannot (yet) replace.
Another aspect that is becoming more important is an understanding of business processes and priorities. One participant recalled the concept of the ‘purple person’: someone who can bridge the gap between technical data specialists (‘red person’) and those with business acumen (’blue person’). It’s a role that has always been valuable in the data industry, but it remains surprisingly rare. Rather than teaching our business people about data, perhaps we should be teaching our data people about the business!
The takeaway: As AI makes technical capabilities more accessible, the ability to understand business context, exercise judgement and communicate clearly becomes an even greater differentiator. Helping people connect data with business understanding has always been a central part of Data Culture, the focus now is on using that knowledge when working with AI.
4. You can't just tell people to experiment, you have to give them time
One of the most practical ideas of the evening came from Dan, who has been encouraging his team to experiment with AI by assigning two hours of protected time every Tuesday afternoon – no prescribed training programme, no requirement to produce something useful, just an invitation to bring a question, share a tip or try something new.
As he pointed out, “nobody just gets on a surfboard and looks cool.” Learning something new involves falling off, swallowing half the sea and looking rather foolish. The important thing is to create an environment where people feel safe doing exactly that.
Charlotta described a training initiative that taught colleagues how to build a simple birthday card generator using AI. Some initially questioned why they needed to learn to build an application when their jobs involved booking travel, scheduling meetings or supporting executives. But the point wasn’t the birthday card. It was learning how to approach a task differently and recognise where AI could help.
Importantly, though, this is not just about automating tedious tasks – it’s about being innovative and finding ways to accomplish things which would be impossible without AI. Once processes are made more efficient, the time freed up can be reinvested in learning, experimentation and gradually developing the skills and confidence to build more advanced agents and tools.
The takeaway: AI adoption needs dedicated time for experimentation, not just training materials and an expectation that people will somehow fit learning around their existing workload. Data Culture professionals have long worked to build the confidence, curiosity and psychological safety that encourage people to try new things - AI gives us another reason to make these behaviours part of everyday working life, rather than treating learning as a separate activity.
5. Who is responsible for the people side of AI?
One of the roundtables had a particularly lively debate about who should be responsible for ensuring employees adopt AI in their everyday work. It may have started out with the Data Office, but organisation-wide capabilities are usually the responsibility of a central Learning & Development team. Or is it like any other office desktop tool which falls under the Technology function?
One participant described how their organisation had successfully used established HR leadership programmes and regional learning priorities to introduce AI training: HR provided the structure and mechanisms, while data and AI specialists contributed the expertise. Another explained that their small L&D team was enthusiastic but already stretched across essential priorities such as health and safety. Finding space for data and AI training was difficult, even when everyone agreed it mattered.
The lucky ones had been able to bring HR, data, technology, AI and internal communications together to coordinate their approach. Data and technology teams provide use cases and real success stories. HR and L&D teams coordinate learning and development. Internal communications spread the message and generate interest.
This combined approach can make it difficult to work out who is ultimately accountable. We can perhaps take inspiration from the way organisations have approached data management, making maturity a measurable leadership priority. Divisional leaders given numerical improvement objectives as part of their performance scorecards mean the responsibility is tangible, visible and difficult to ignore.
If AI had a similar maturity framework, with leadership accountable for progress, what would it measure? Training completions, tool usage and engagement scores might demonstrate adoption and usage, but they don’t necessarily show that people are working differently. We also need to look at outcomes: What can people now achieve that they couldn’t before? How much time is being saved? What business value is being created?
The discussion even ventured into whether HR might one day need to manage AI agents as part of a mixed human and digital workforce. If agents are increasingly completing whole tasks, who manages their performance, responsibilities and development? It sounds rather futuristic, but some organisations are already planning for this kind of scenario.
The takeaway: AI adoption works best if there's a coordinated approach across HR, data, technology and internal comms. This is a familiar conversation for anyone working in Data Culture, as we know changing behaviours takes collaboration across the organisation, with leadership support and a shared sense of responsibility.
6. Leaders need to be early adopters - but some need convincing
When it comes to adoption, we know from our Data Culture experience how important it is for the bosses to be involved, actively encouraging their teams and leading by example.
With AI there can be an additional tension here. Senior colleagues can be harder to engage with AI, particularly when they’ve spent decades working successfully in a particular way, or have team members that already prepare presentations and analysis for them.
Sometimes it’s about finding the particular example or benefit that will persuade a reluctant leader to take the first step. Others might need to be ‘pushed into a corner’ before they’ll adopt new ways of working. Generally the group agreed that adoption can be personal, with different people responding to different motivations.
Once they’re convinced, though, visible leadership support can make an enormous difference. Charlotta described how a senior executive publicly endorsed AI training after experiencing the benefits himself, and soon afterwards, other leaders began requesting training for their entire teams.
The takeaway: We've seen through Data Culture initiatives that when leaders ask questions and demonstrate that they're willing to learn, it gives others permission to do the same. AI adoption is no different. Leaders need to show curiosity, make learning possible and help their teams understand why it matters.
7. AI changes how work happens, but not who is responsible
Amongst all the excitement about AI’s capabilities, conversation often returned to risk, governance and accountability.
The group challenged the idea that AI governance should be treated purely as a data team responsibility. After all, AI is increasingly being used throughout the business. It might change how quickly a task happens or how much is automated, but it doesn’t transfer responsibility away from the team benefiting from the output.
The risk still belongs to you. It's just that AI changes the way it happens and the speed at which it happens.
There are particular challenges around introducing AI tools in a highly regulated environment like investment banking, where access to sensitive information needs careful management. As tools become more powerful and gain access to internal documents, organisations need to consider not only what AI can do, but also what information it should be allowed to use. And as AI moves from assisting people to acting on their behalf, the risks increase – with examples appearing in the news recently of autonomous AI agents communicating or taking actions beyond their intended instructions.
We noted an interesting paradox when considering the different levels of understanding and adoption of AI in an organisation: you might expect the people with greater knowledge of a topic to have more confidence in it – but they’re often more cautious because they understand its complexities and risks.
With vendors promoting ever more sophisticated capabilities and organisations keen to demonstrate that they’re ahead of the curve, there’s a temptation to chase every new development. But the winners may not be those who adopt everything first. The thoughtful ‘fast followers’, who understand their business problems and choose appropriate tools, could ultimately be more successful.
The takeaway: Building a strong Data Culture isn't just about encouraging people to use data. It's about helping them understand it, question it and take responsibility for the decisions they make. As AI takes on more tasks, we need to make sure people don't lose sight of their own accountability, particularly as technology becomes more autonomous and its outputs become harder to question.
What did we take away?
It was an evening full of interesting questions, practical examples and the occasional uncomfortable prediction about the future of work.
We didn’t arrive at a definitive answer to what everyone should learn, or how every organisation should approach AI adoption, but a few messages came through particularly strongly:
- Teach people to question, not just prompt. AI literacy needs to include critical thinking, business understanding and the confidence to challenge outputs.
- Protect opportunities to develop expertise. If AI removes the tasks through which people traditionally learned, organisations need to create new ways to develop judgement and experience.
- Value human skills. Curiosity, communication, adaptability and the ability to translate complexity into something useful will become increasingly important.
- Give people permission and time to experiment. Learning requires practice, mistakes and a supportive environment.
- Make AI development everyone’s responsibility. HR, L&D, data and technology teams all have a role, but leaders must take accountability for developing their people.
- Measure outcomes, not just activity. Tool usage and training completions are a starting point, but the real test is whether people can do something better.
- Remember that accountability doesn’t disappear with automation. AI may perform the task, but people and organisations remain responsible for how it’s used and what happens next.
Perhaps the most encouraging message was that AI doesn’t make the work of Data Culture professionals less relevant – quite the opposite! We’re already familiar with many of the challenges organisations now face with AI: encouraging curiosity, building confidence, developing skills, overcoming resistance, helping people trust information appropriately and creating an environment where new behaviours can stick.
The technology may be changing rapidly, but the people challenges are remarkably familiar.
Araminta Huitson
Data Culture Lead & Data Culture Club Founder
Join the next conversation
We’ve nurtured a community of Data Culture leaders who meet regularly, online and in-person, to share ideas, swap advice and get real about what works (and what doesn’t) when it comes to building healthy data cultures that last.
With more than 100 members representing multiple countries, and working for a range of industries, we’re able to bring together a great mix of perspectives helping shape the discipline.
We hold regular events so people are able to meet in person and virtually to talk about their challenges and successes with people who get it. We’d love for you to join us!
