AI adoption in a finance team — 90% people, 10% tech

AI adoption in a finance team is 90% behavioural change and 10% technology — rolling out a Copilot or Claude license takes an afternoon, but getting the team to actually use it takes months. For finance teams the specific resistance traces back to three patterns (audit anxiety, control instinct, professional identity), and only rituals addressing these directly work — not generic change programmes.

The technology isn't the problem. Copilot, Claude, and ChatGPT are ready in ten minutes; rolling out licenses to a finance team of twelve takes an afternoon. And yet AI adoption stalls in most SME and scale-up finance functions within three months — not because the tool doesn't work, but because the team doesn't change. AI transformation in finance is 10% technology and 90% change management. Underestimate that and you'll pay licenses without anything changing in the close cycle.

Why AI adoption in finance is different

Three traits make finance an odd department for AI adoption.

Precision culture. Finance staff are trained on accuracy, double checks, and four-eyes control. A tool that gives "approximately right" answers is a learning point in marketing; in finance it's a red flag. Adoption often stalls on understandable mistrust: "that AI gave something last week that wasn't right, I won't believe the next thing either."

Audit awareness. Any work landing in the books or going into a formal report has to be reconstructable. A controller who doesn't know how an AI commentary came about doesn't dare sign it. That's not stubbornness; that's professional discipline.

Specific job anxiety. Outside finance, the message is "AI doesn't replace people, it helps people." In finance, a stronger worry persists: "automation hits my role first because it's repetitive, rule-bound work." That worry isn't unfounded for pure data-entry roles, but misses the broader reality — the scarcity of finance talent means experience shifts to review and judgment, not that jobs disappear.

The fear you have to address

Practice shows four specific concerns in finance teams:

• My job — "is my work going to disappear, and if not, will I become a review machine?"
• My experience matters less — "I've built 15 years on this ledger, and now an agent gets to do the work?"
• Audit liability — "if the AI gets it wrong, will I get blamed anyway?"
• My own standing — "apparently everyone already understands this, except me."

The last one is often the strongest brake in finance. An experienced controller seeing LinkedIn posts about agents feels publicly behind — and goes quiet. Adoption stalls not through resistance, but through a polite kind of invisibility.

The shift — from data entry to review

The frame that does work in finance: AI doesn't replace finance staff, it shifts their role from entering to reviewing. The person who used to type a journal entry now reviews an AI draft. The person who wrote a variance commentary checks and contextualizes the AI version. The person who typed an AR mail approves the personalized draft.

That's a real competency shift, not cosmetic. It requires finance staff to learn to read critically, supply context, and stay accountable for numerical quality — even when they haven't typed a letter themselves. Experience counts more, not less: only those who know the work well can judge AI output well.

Karim Lakhani (Harvard Business Review) sums it up sharply: "AI isn't going to replace humans, but humans with AI are going to replace humans without AI." That's the frame that gets a controller out of defense mode. Not "you have to cope with threat X," but "you get reinforcement Y, provided you learn to use it."

The seven pillars — finance edition

Successful finance AI adoption has seven building blocks in place:

1. Sponsorship — CFO or finance manager as active sponsor, with budget and visible commitment.
2. AI champion in finance — someone with mandate, time, and finance domain knowledge.
3. Bottom-up culture — finance staff dare to experiment, are allowed to make mistakes.
4. Tools, data access, and policy — right licenses, accounting connection, clear policy on Tier 1-4 data.
5. Training and hands-on practice — not only power users, the whole team.
6. Project cadence — ownership per pilot, rhythm of check-ins.
7. Use-case prioritization — process to bring in ideas and choose between them.

Of these seven, a champion can influence three on their own (2, 3, 7). The other four — sponsorship, tools/policy, training scale, project cadence — require explicit leadership decisions. Appoint a controller as champion without putting these four in order and you're setting an enthusiast on an island.

What works in a finance team

A few narratives and rituals that land in practice:

• Start with the why. The CFO explains why AI matters for this finance function, in two sentences anyone can repeat. "We want the close from 8 to 5 days, so we can steer on numbers earlier. AI is a tool, not a substitute." Concrete and pointed, not generic.
• Make it concrete with in-house examples. One pilot where variance commentary goes from 2 hours to 25 minutes, with real numbers and the work in hand, does more than ten webinars. An AP clerk showing how she now handles 80% of invoices in 5 seconds instead of 3 minutes convinces colleagues within the quarter-hour.
• "AI win of the week" in the finance standup. Someone from the team shares one concrete win each week. No grand stories, just a screenshot and one line of explanation. Normalizes experimenting, builds visibility.
• Explicit permission to experiment. Finance staff who think they "have to be productive" don't dare spend an hour trying things out. A CFO or finance manager who literally says "I expect you to spend 2 hours per week experimenting with AI on your own work focus" removes that block.

What doesn't work in finance: an internal memo using the word "transformation," a one-time half-day training, rolling out a tool without guidance and hoping it lands.

Top-down and bottom-up — combined for finance

A pitfall is the choice between top-down (CFO, strategy, roadmap) and bottom-up (controllers, AP/AR, daily experiments). Either alone doesn't work. Top-down without bottom-up produces presentations nobody executes; bottom-up without top-down produces scattered experiments that never reach the close cycle.

A workable split in finance:

• Top-down (CFO/finance manager) delivers: the why, the priority, the sponsor role, the budget, and governance-by-design (data classification, GDPR, audit requirements).
• Bottom-up (controllers, AP, AR) delivers: concrete workflows, the business case per use case, the action plan for quick wins, and the daily learning behavior.

The finance champion is the connector. Without one, the CFO talks to a void and the team flails without direction.

Where this lands in practice

• A family-owned logistics business (120 FTE) names a senior controller as finance AI champion, gives them a day a week, and starts with three use cases (variance commentary, bank reconciliation, AR mails). After three months one has scaled, one has stopped (data quality not good enough), one is iterating. Stopping was publicly celebrated as "we've learned what doesn't work."
• An accounting firm (45 FTE) introduces the "AI win of the week" routine in the Monday standup. After six weeks, more wins come from administrative staff than from consultants — the sign that finance adoption is landing broadly.
• An SME manufacturer (180 FTE) rolls out Copilot to finance without training or policy. After four months, 70% of licenses turn out to be unused; the 30% used sits with three enthusiasts. Classic tech-first, people-last — and a predictable outcome.
• A consultancy (30 FTE) uses the AI assessment to determine per finance role where the biggest blocker sits, and ties it to a 30/60/90-day plan. The change becomes diagnosis-driven instead of ideology-driven.