Fourteen people asked, before a line was written.
The question was never whether. It was where.
An accountancy practice runs on deadlines stacked on deadlines. OFM’s partners could see that AI ought to be taking routine work off the team, and they were clear-eyed about the harder question: precisely which work, in a firm handling client financial data where being approximately right is not good enough.
They could have bought a tool on a recommendation and hoped. A great many firms do. Instead they backed a proper diagnostic first — which is the decision that made everything after it work.
Survey, then scope, then build.
The audit
Every active member of the team answered a structured survey in their own words — trainees, accountants, payroll, financial planning — and both partners were sat with on site. The firm was mapped against four pillars, and the findings were written up rather than presented as slides.
The readout, then a scoped build
The audit produced a written analysis, an AI policy and a recommendations roadmap. The build was scoped from that document rather than from a conversation, and commenced on 1 July with the scope and terms agreed in writing first.
A partnership, not a handover
The build was delivered through July; the partnership runs to the end of the year on Iris, with on-site sessions continuing into the autumn. Onboarding completed in August, and the work has carried on from there.
This engagement is ongoing. The audit and the build are delivered and the platform is live; the on-site sessions run through the autumn and the partnership term ends in December. We publish outcome figures when we have measured them, so there are none on this page yet — there will be.
The answers are held further down.
Interviewing fourteen people individually takes longer than sitting down with two partners, and it is the entire reason the build was the right one. Leadership can describe a process as it is designed. The detail that decides whether automation works — which step takes the time, which report gets rebuilt by hand, where the judgement actually sits — is held by the people doing it daily. OFM gave us access to all of them.
Returning for a second round in August mattered just as much. Nine further responses, three months after the tools landed, answer a question no pre-rollout survey can: what people are genuinely using, rather than what they expected to use.
If you run a practice and want the same picture of your own firm, that is what an Operational Audit is: a written readout you can act on, and the scope of whatever gets built next.