The State of AI in Finance Teams 2026.

01The headline numbers

Read together, the numbers describe a market in an awkward adolescence. Adoption headlines look healthy, but most teams are stalled between a pilot and a payoff: AI is in the building, not yet in the close. And the constraint has moved. It is no longer budget or belief; it is skills, workflow design and trust in outputs.

02What finance teams ask AI assistants (our data)

We can add something the industry surveys cannot see: what finance professionals ask AI assistants themselves. AI assistants have quietly become a research channel for finance decisions, and because our finance guidance is regularly cited in AI-generated answers across Microsoft Copilot and partner experiences, the citation logs give us a rare window into the questions being asked.

Across roughly 1,900 AI-assistant citations of our finance guidance in the second quarter of 2026, the questions cluster into three intents:

• Tool choice. Which AI tool for which finance job is the dominant question. Practitioners are past asking whether to use AI; they are deep into which one, for which task, on which data.
• Finding the opportunity. A large share of questions amount to "where would AI actually pay off in a business like ours". Teams believe in the value; they cannot yet see where it sits in their own workflows.
• Making the stack work. ERP-adjacent questions, AI alongside Dynamics 365 or Workday, automating KYC and onboarding, suggest the market's centre of gravity is shifting from chatbots to systems.

Two implications. First, if your team is asking an AI assistant which tool to buy, so are your competitors, and those answers are shaping shortlists before any vendor conversation happens. Second, the questions mirror the adoption gap in section 01: the market is not short of interest, it is short of a map. The pattern is global, too: the finance professionals reading this guidance span the UK, Southeast Asia, North America and the Gulf.

03The tool landscape: nobody should pick one winner

Our own testing across ten finance tasks, published and updated quarterly, keeps reaching the same conclusion: the "best AI tool for finance" question is malformed. The honest answer is a stack.

The full scored comparison, kept current with each major model release, is in our ChatGPT vs Copilot vs Claude for Finance test results, where you can also download the one-page Finance AI Tool Decision Matrix.

04What AI actually does in finance today

Across the industry research and our own engagements, the tasks that have genuinely moved from pilot to production share a shape: high volume, rule-adjacent, human-reviewed. Gartner's survey puts knowledge management (49%), accounts-payable automation (37%) and error and anomaly detection (34%) at the top of the deployed list. Our client work matches that pattern from the ground:

• Reconciliation matching: most matches are obvious; AI confirms them and routes only exceptions to people (how the setup works).
• First-draft variance commentary and reporting packs: the analyst reviews instead of writes. In one engagement, automating month-end close workflows removed roughly 60% of repetitive volume and unlocked £100k+ of annual capacity (case study).
• Document-heavy compliance: KYC and client onboarding, where AI-assisted review with human sign-off cuts processing time without cutting the control (case study).

What has not crossed into production at most firms: autonomous judgement. Forecast sign-off, close approval and anything a regulator will one day ask about still ends with a qualified human, and in our view should.

05The gap that defines 2026: budgets are moving faster than skills

The most important finding in this year's research is a mismatch. Deloitte's 2026 CFO Signals survey reports a sharp rise in CFO confidence and AI use, and 87% of CFOs say AI will be extremely or very important to their finance operations in 2026. Yet in AICPA & CIMA's survey of 1,446 senior finance leaders, only 8% feel very well prepared for the AI transition, and 56% name generative AI as their single most prominent skills gap. Organisations say they prefer to close that gap with internal training (62%) and on-the-job learning (61%) rather than hiring.

Money is arriving faster than capability. The teams that convert budget into results in 2026 will be the ones that treat AI as a skills programme with software attached, not the reverse.

This matches what we see in training rooms: the blocker is rarely the tool licence. It is that nobody has shown the team how to prompt against their own trial balance, their own board pack, their own close calendar. Generic AI training does not survive contact with a month-end.

06What a finance team should do with all this

1. Pick tools per task, not per brand. Use the stack logic in section 03; run any new tool through a structured evaluation (we publish ours as the F.A.I.R. framework: Fit, Accuracy, Integration, Risk).
2. Automate the drudgery first. Reconciliation, first-draft commentary and document review are the proven, low-regret starting points. Keep sign-off human.
3. Close the skills gap deliberately. Train on your own files and workflows, not demos, and give the team a shared prompting method (ours is RACEF).
4. Map before you buy. The 45% of teams stuck in pilots mostly skipped the step of identifying where AI actually pays in their workflows. A structured audit of your processes is cheaper than a stalled pilot.

07What the published research cannot tell you, so we're asking

Industry surveys skew to large enterprises and to CFOs. They say little about what individual practitioners and mid-market teams experience: which tools people actually open every morning, the honest hours saved per person, and what specifically blocked the teams where AI went nowhere. So we are running our own survey of finance professionals worldwide: 10 questions, 3 minutes, anonymous.