Building a Finance Function AI Roadmap

Most finance functions adopt AI in fragments — a tool here, an enthusiastic analyst there — without a plan that a finance leader owns. A roadmap changes that. It sequences adoption so that value and governance move together, and it gives the CFO something concrete to take to the board. McKinsey’s work on finance functions is blunt on this point: pilots launched without direction rarely scale. This guide sets out how a finance leader builds a roadmap that does.

Why a roadmap, not a rollout

A rollout pushes tools into the function and hopes for the best. A roadmap decides, deliberately, what to adopt, in what order, with what controls, and who owns each stage. The difference matters because ungoverned AI in finance is where the risk lives — and because a board is far more comfortable with a phased plan than with an open-ended experiment. It also matters because the evidence shows unled adoption tends to fail: the functions that capture value tie AI to specific business needs and transform domain by domain rather than all at once.

The pitfalls a roadmap is built to avoid

It helps to design the roadmap explicitly against the failure modes McKinsey identifies in finance functions:

• Waiting for perfect data. Teams that delay until every data set is clean never start. A good roadmap delivers value on today’s data while strengthening the data foundation in parallel.
• Transforming all at once. Holding back until the whole function is ‘AI ready’ stalls progress. The better path is domain by domain, building momentum and capability.
• No direction. Pilots with no roadmap rarely scale. Every use case should trace to a defined business need.

The phases of a finance function AI roadmap

Phase one — foundation and governance

Before any tool touches live data, set the ground rules: what AI is permitted for, where human review is mandatory, how data security is handled, and who is accountable. This phase is about governance, not technology — and skipping it is the most common serious mistake.

Phase two — low-risk, high-value tasks

Start where the value is clear and the risk is contained — drafting and first-pass work that a person then reviews. This builds the team’s literacy and confidence without exposing the business, and it demonstrates value to the board early. It is also where the 20 to 30 percent time savings McKinsey observed typically first show up.

Phase three — embedding review discipline

As adoption widens, the constraint shifts from production capacity to review capacity. This phase is about making human checking a designed part of the process rather than an afterthought — so that a faster function does not become a less controlled one. It is also where a function begins to consider agentic workflows, which act rather than merely draft and therefore demand tighter governance.

Phase four — capability handover

The endpoint is a governed capability owned by the permanent team, with clear ownership and a policy that outlives any individual. If a fractional leader built the roadmap, this is where they hand it over.

Change management runs through every phase

The technical steps are the easy part of a roadmap. The difficult part is bringing the finance team with you — addressing the reasonable fear that AI is there to replace people, and reframing it as a change in what the team spends its time on. The evidence is clear that tech-first adoption, which neglects this, underperforms: organisations putting people at the centre are markedly more likely to realise their AI investment. A roadmap that plans the technology but not the people is only half a roadmap. This connects to the wider work of building AI literacy across the team.

How long a roadmap takes — and why that varies

Finance leaders often want a timeline, and the honest answer is that it depends on where the function starts. A function with clean data, an open team and a clear business need can move through the early phases in a matter of months; one wrestling with fragmented data and a resistant team will take longer — and should, because rushing the foundation is how adoption fails. The roadmap’s value is partly that it makes this realistic: it replaces ‘when will AI be done’ with a phased plan the board can track.

Crucially, a roadmap is not a one-off project with an end date so much as a sequence that hands off into business-as-usual. The endpoint is not ‘AI adopted’ but ‘AI adoption owned by the team as a normal part of how the function works’.

Measuring whether the roadmap is working

A roadmap the board can track needs measures. The most useful are not about how many tools are in use but about outcomes:

• Time redirected from routine processing to analysis and business-partnering — the 20 to 30 percent shift the best functions achieve.
• Whether AI-assisted outputs are consistently reviewed before use, evidenced rather than assumed.
• Whether new use cases pass through governance before deployment.
• Team confidence and engagement with the tools — because adoption that the team resists is adoption that will not last.
• Value tied to specific business needs, not activity for its own sake.

Measuring outcomes rather than tool counts keeps the roadmap honest and keeps the board’s attention on whether AI is actually improving the function — which is the only test that matters.

A worked domain: what a roadmap does to FP&A

Forecasting is a good illustration of the domain-by-domain approach, because it shows how a roadmap changes not just the tooling but the fundamentals of what a finance domain does. Traditional FP&A locked finance into annual cycles; AI-enabled forecasting allows rolling forecasts that update monthly, weekly, or even daily depending on business cadence. The leading finance teams are treating this as a capability rebuild rather than a software deployment — retraining planners to set assumptions differently and using AI to test scenarios rather than to replace judgement.

A roadmap that reaches FP&A therefore does not simply add a tool; it re-poses the question of what the function is forecasting, why, and for whom. That is exactly why sequencing matters — a domain like this rewards deliberate rebuild and punishes a tool bolted onto unchanged processes. The teams that answer the fundamental questions first redefine what FP&A does; the rest, in one memorable phrasing, are simply updating spreadsheets faster.

Taking the roadmap to the board

One of the roadmap’s quieter benefits is that it gives the CFO a coherent story for the board. Rather than fielding an open-ended ‘what are we doing about AI’ with a list of tools, the finance leader can present a phased plan: here is where we are, here is what comes next, here is how it is governed, here is how we will know it is working. That reframes AI from a source of board anxiety into a managed programme with milestones — which is both more reassuring and more honest than either hype or vagueness. A roadmap, in the end, is as much a governance and communication instrument as a technical one.

What owning the roadmap says about a finance leader

A CFO or FD who can build and lead a roadmap like this is demonstrating exactly the capability boards now look for — deliberate, governed, commercially proportionate AI adoption that avoids the known pitfalls. It is one of the clearest ways a finance leader shows they are equipped for the way the function is changing, and a capability FD Capital assesses when placing senior finance leaders.

Call 020 3287 9501 or email recruitment@fdcapital.co.uk to discuss a finance leadership appointment where owning the AI roadmap is part of the brief.

FD Capital — CFO and Finance Director Recruitment

Fellow of the ICAEW | Placing finance leaders who can plan and deliver a modern finance function since 2018. We recruit permanent, interim and fractional finance leaders across the UK. 4,600+ network. 160+ placements. Shortlists in 3–7 working days.

This guide is general information for finance leaders and does not constitute professional advice. Businesses should take their own advice on their specific circumstances.