BCG's 2026 AI at Work research found that only 36% of employees feel adequately trained for how AI has changed their role. The result shows up in the numbers finance cares about. Untrained teams run up licence and token spend with nothing on the P&L to show for it.
We hear it on first calls: "We bought AI licences and nobody uses them." The tool is rarely the problem. The training was.
Generic courses do not change how people work
Most AI training teaches features: how to write a prompt, what the tool can do, a few examples from someone else's industry. People leave interested and go back to work exactly as before, because nothing connected the tool to a task already in their week.
Training that changes behaviour is built from the participant's own work. A compliance officer practises on a regulatory question they face this month. A finance team works on its own management pack. The exercise is the job.
Tailor on three axes
We tailor every programme on three axes: your industry, each business unit and each role, from the board to the operators. Examples of what teams train on:
- Commercial: pricing and performance analysis, competitor monitoring, campaign content and client meeting preparation.
- Operations: incident write-ups, manuals searched in plain language, scheduling scenarios and quality reporting.
- Corporate: contract and supplier review, board and management packs, policy and regulatory change summaries.
- Financial services: KYC and source of funds summaries, credit papers, investment memos and committee packs.
Around 75% of the effort happens before the first session: discovery, a readiness diagnostic through our AI360 toolkit, interviews with unit leads and building every exercise from your own material.
What it looked like at one regulated firm
A regulated financial services firm with more than 100 staff had AI licences on more than a hundred desktops with little use. We ran a five-day programme, department by department, from the chairman and CEO to operations. It was built around real tasks: board minutes, fund reporting, KYC checklists and client correspondence.
Within three months, 84.8% of licensed staff (89 of 105) were actively using AI. In the first 28 days, staff saved 652 hours and submitted 8,423 prompts. One participant from the compliance team was using what they had learned on a real fund regulatory question the next morning.
Draw the governance line at the same time
Adoption without guardrails creates a different problem. Good training makes clear what data can and cannot go into a tool, against the frameworks that apply to you. Risk committees sign off faster when that line is drawn in the training, not after an incident.
Make the capability stay
Training should end with something the firm keeps: an AI hub on your own intranet, prompt libraries, a log of use cases and a route in for new joiners. Hours saved then convert into cost a board can book, and the next wave of use cases comes from your own people.
