Why are AI and data leaders struggling to retain talent?

If your AI or data function is losing strong people, pay is the obvious place to look.

It may be part of the problem. It is unlikely to be the whole of it.

Lucent Search's AI leadership research found that 86% of leaders experience difficulty retaining top talent. That is a striking number for organisations that have often invested heavily in hiring scarce technical capability.

The external market clearly plays a part. Strong AI specialists have options and are approached regularly.

But people are also making a judgement about the organisation they have joined. Can they get useful work into production? Can decisions be made? Does their leader have enough influence to remove obstacles? Is there a credible career ahead of them?

If the answer to those questions is repeatedly no, another employer does not need to offer much more than the prospect of being able to get things done.

If you sit on the board, lead the business or own the people agenda, persistent attrition in an AI or data function should prompt a wider question: what are people experiencing once they arrive?

If good people are leaving, look at whether they can actually deliver

You can recruit an excellent AI team into an environment that makes success unnecessarily difficult.

The team may have been hired to improve productivity, build new products, automate processes or create new sources of value. Once they start, they discover that the data is not ready, systems are difficult to integrate, investment decisions take months or the business functions they depend on have different priorities.

Lucent's research found that 91% of AI leaders struggle to move pilots into production and 89% face challenges integrating AI with existing systems.

Those are not abstract technology problems when you are trying to retain people.

Consider the experience of a strong machine-learning engineer who has spent 18 months building models that never reach customers or operations. Or a senior data scientist whose team repeatedly proves that something works, only to find that nobody has agreed who will fund or own the next stage.

At some point, the issue becomes professional as well as organisational.

People want to see the work they are good at become useful.

If several strong people are leaving, look at how much of their time is spent producing valuable work and how much is spent trying to get permission, access, funding or attention.

Look at the organisation around the AI team, not just the team itself

AI rarely succeeds within one function.

Your AI leader may need technology to change a platform, a business function to redesign a process, finance to approve investment and risk or legal teams to agree how something can be deployed.

That creates dependencies on people outside the AI function.

Lucent's research found that 51% of leaders considered cross-departmental collaboration very challenging and another 36% moderately challenging.

You will usually see the consequences before they appear in an engagement survey.

Projects slow down. The same decisions return to the executive committee several times. Senior technical people spend increasing amounts of time managing stakeholders. Teams start choosing work according to what they can get approved rather than what would create the most value.

A good AI leader can improve those relationships. They cannot indefinitely compensate for an organisation in which nobody has resolved how cross-functional decisions will be made.

If you are seeing repeated frustration or turnover, ask where the team gets stuck and who has the authority to resolve it.

If the answer is effectively “nobody”, changing the retention programme will not solve the underlying problem.

Check the job you have given the leader

Problems further down an AI or data team can start with the senior role above it.

You may have appointed a capable leader but given them a job that is difficult to succeed in.

Perhaps they are accountable for AI outcomes but do not control the data or technology required to deliver them. Perhaps they need several other executives to agree every material investment. Perhaps the CEO is supportive in principle but rarely intervenes when another function blocks progress.

The team notices.

They see priorities change. They see projects lose funding. They see their leader repeatedly asking for decisions that never quite arrive.

Lucent's research found that nearly 63% of AI leaders planned to change roles within the following 12 months.

That figure does not prove that poor organisational design is causing people to leave. Senior AI leaders are in demand and careers move for many reasons.

But it gives you a reason to examine the quality of the role more closely.

If you want a senior AI or data leader to stay, they need a credible route to achieving something. The same applies to the people working for them.

A title, compensation package and seat at the executive table will not compensate indefinitely for a job with responsibility but little practical ability to act.

Pay can be part of the problem without being the underlying problem

You still need to know whether you are paying people fairly.

Lucent's research identified signs of internal pay compression, with some newly appointed leaders being paid as much as, or more than, longer-tenured incumbents. Some of the small group who had remained in post for five to ten years also appeared to have fallen behind the market.

That deserves attention, particularly if you are responsible for reward, succession or executive retention.

But increasing salaries every time somebody receives an external approach can hide a more fundamental issue.

A well-paid executive can still conclude that the job is no longer worth doing.

If their team cannot deploy what it builds, if resources remain uncertain or if every significant decision becomes a negotiation with another function, an additional increase in salary may simply postpone the resignation.

When you lose somebody good, establish whether they are leaving for a substantially better package or whether the new role also gives them something your organisation did not: greater authority, stronger sponsorship, better infrastructure or a more credible opportunity to make an impact.

Those are different problems and require different responses.

Strong AI people tend to care whether their work goes somewhere

AI and data professionals often have considerable choice over where they work.

The strongest people are not only comparing compensation.

They are also assessing the problems they will get to solve, the quality of the people around them and whether the organisation can turn technical work into something that is actually used.

That becomes more important once the novelty of joining has worn off.

A sophisticated project that never moves beyond a pilot stops being particularly sophisticated from the perspective of the person working on it.

A succession of proofs of concept can begin to look like evidence that the organisation is interested in experimenting with AI but less prepared to make the changes required to use it.

Your team does not need every project to generate an immediate financial return.

They do need enough evidence that good work can progress.

If you want to understand the health of the function, ask what happened to the best work the team produced over the past 12 months.

Did it reach the business? Did it change anything? If it stalled, why?

The answers will tell you considerably more than a retention target.

Your career structure may be forcing people to leave

There is another problem that becomes more visible as AI teams mature.

The people you hired two or three years ago may now be doing very different jobs.

An individual contributor may be leading a team. A technical specialist may have developed strong commercial judgement. A senior data scientist may want broader leadership responsibility rather than another variation of the same technical title.

If your organisation has no credible next move for them, the external market provides one.

You need to understand what progression looks like before a resignation forces the conversation.

That does not mean creating management roles for everyone.

Some people will want deeper technical responsibility. Others will want to lead teams, own products, take on broader business problems or eventually move into executive leadership.

The risk arises when leaving is the clearest way to grow.

If several of your strongest people move to bigger roles elsewhere, examine whether your own structure was giving them anywhere to go.

This is where the people agenda and the business agenda meet. Career architecture, succession and development may sit with the CPO, but the opportunities available to these people depend on how seriously the organisation intends to build AI and data capability.

Do not make the AI leader accountable for problems they cannot control

You should expect senior leaders to build good teams.

Your AI or data leader should create clarity, develop people, deal with weak performance and give strong people worthwhile work.

But be careful about treating retention as a simple measure of whether they are leading well.

There are limits to what an individual executive can fix.

They cannot personally correct a company-wide pay problem. They cannot fund projects if the capital is controlled elsewhere. They cannot resolve every cross-functional dependency if other executives are free to ignore agreed priorities. They cannot create meaningful career progression if the organisation will not support it.

Separate the things the leader genuinely controls from the conditions you are asking them to operate within.

That distinction is important for the board when assessing the leader, for the CEO when looking at execution, and for the CPO when deciding whether the problem is leadership, reward, development or organisational design.

Otherwise, you can end up holding the executive responsible for the consequences of decisions made elsewhere.

Sponsorship becomes visible when something difficult needs to happen

Most organisations will say their AI agenda has senior sponsorship.

You find out how strong that sponsorship is when the team needs something.

Perhaps a pilot has worked and now requires serious investment. A business unit needs to change a process. The AI team needs access to data controlled elsewhere. Risk and technology disagree about deployment.

Someone then has to make a decision.

Lucent's research found that roughly three-quarters of leaders experienced at least some difficulty securing organisational buy-in for AI projects.

If your AI leader has to rebuild the argument every time a difficult decision reaches the executive team, progress will be slow.

Their team sees that too.

Sponsorship is visible in what happens when priorities conflict.

Does somebody make the call? Does the agreed priority hold? Does the organisation release the money, data or people required to move forward?

Repeatedly failing those tests will eventually affect who wants to work in the function.

Treat resignations as data

When somebody strong leaves, you will normally be given a reason.

More money. A bigger title. An interesting opportunity. A well-known technology company.

The explanation may be perfectly accurate.

You still need to understand why they were prepared to listen.

One resignation tells you relatively little. Patterns tell you much more.

Look at who is leaving, when they leave and where they go.

If several of your strongest people leave after two or three years, pay attention. If they repeatedly move into jobs with greater ownership, examine why. If exit conversations keep returning to the same dependency, executive relationship or inability to get work into production, take that seriously.

Do not rely only on exit interviews.

By the time somebody resigns, the decision has often been made.

Speak to the people you most want to keep while they are still committed enough to tell you what is making the job unnecessarily difficult.

That conversation may sit with the CPO or another senior people leader, but the response may require decisions from the CEO or board.

You are not trying to eliminate turnover. Strong organisations lose good people.

You are trying to establish whether good people are leaving for reasons you could reasonably have addressed.

Fix the conditions before you hire the replacement

This is where retention problems become expensive.

A strong person leaves. You appoint another strong person. They enter the same structure, inherit the same dependencies and encounter the same difficulties.

Eighteen months later, you are having the conversation again.

Before you go back to market, establish whether anything needs to change around the role.

That may involve clarifying which decisions the leader can make, improving their access to the CEO, resolving a difficult relationship between functions, increasing resources or creating a clearer route for strong people to progress.

Sometimes you will still need a different leader.

But replacing the person without examining the conditions around the job risks treating the visible vacancy while leaving the reason it exists untouched.

So why is AI talent difficult to retain?

Scarcity matters. Good AI and data people have options.

But what they experience after they join also shapes your ability to retain them.

Can they get useful work into production? Can they make progress without continually fighting the organisation? Does their leader have enough authority and support? Can strong people see somewhere to go next?

If you are losing people you expected to keep, those are the questions worth examining before you conclude that the market has simply become too competitive.

For the CEO, this is partly an execution issue. For the board, it can be a signal about leadership, capability and organisational risk. For the CPO, it raises questions about reward, career structure, succession and the design of senior roles.

You are all looking at different parts of the same problem.

You cannot control how often your strongest people are approached. You have considerably more control over whether they believe they can do their best work by staying.

Related analysis and executive search

What do AI and data leaders need from the organisation to deliver?

How authority, sponsorship, resources and organisational conditions affect the ability of senior AI and data leaders to create results.

Can your CIO lead AI, or do you need a different leadership model?

How to assess whether the existing CIO has the capability, remit and authority to lead enterprise AI.

We’ve invested heavily in AI but aren’t seeing enough business value. What kind of leadership do we actually need?

How to identify whether the constraint sits in leadership, ownership, data, technology or execution.

Chief AI Officer Executive Search

For organisations appointing senior enterprise AI leadership.

CDO / CDAO Executive Search

For organisations appointing senior data and analytics leaders.

Losing strong people from your AI or data team?

If you are losing senior AI or data people, it is worth understanding whether the problem sits with the individual, the role or the organisation around them before you appoint a replacement.

Lucent Search works with CEOs, boards and Chief People Officers to understand the leadership capability and organisational conditions required for AI and data functions to deliver effectively.

Where there is a genuine leadership gap, Lucent conducts retained executive search for senior AI, data and technology leaders.

Speak with Rebecca