Case StudyShadow AIGetting Started
From “we think people use AI” to a clear picture: an accountancy firm (illustrative)
An illustrative example of how a 120-person accountancy firm could move from uncertainty about AI use to a clear register, policy and plan.
Daniel Rogers
Co-Founder & COO
- Published
- Last reviewed
Illustrative example: 120-person accountancy firm
Accountancy and tax
- single AI register replacing guesswork
- 1
- single AI register replacing guesswork
- AI policy staff can read in five minutes
- 2 pages
- AI policy staff can read in five minutes
- from first Snapshot to a working governance rhythm
- 90 days
- from first Snapshot to a working governance rhythm
The challenge
Partners knew staff were using AI tools to draft emails and summarise documents, but nobody could say which tools, how often, or whether client data was involved. A large client had just added AI questions to its annual supplier review.
The solution
The firm used TURMA Snapshot to build a scored picture of AI use across teams, then wrote a two-page AI policy, approved a small set of tools and trained staff on what never goes into an AI tool.
This is an illustrative scenario, based on situations we commonly see in mid-sized professional services firms. It does not describe a specific TURMA client, and the figures show what the process produces rather than measured results.
The situation
A 120-person accountancy firm had grown its use of AI quietly. Some staff used free chatbots to draft client emails. The tax team had started summarising guidance with an AI tool. A new AI feature had appeared in the firm’s practice management software.
Nobody was doing anything wrong on purpose. But when a large client’s supplier review asked "Do you use AI, and how is client data protected?", the partners realised they couldn’t answer with confidence.
What they did
1. Got a clear picture
The firm started with TURMA Snapshot. In place of guesswork, partners saw which AI tools were in use, which teams used them, and where client data might be involved, along with a score and a recommended next step.
2. Set simple rules
Using what Snapshot revealed, the firm wrote a two-page AI acceptable use policy. It approved one business-grade AI assistant with appropriate data controls, and set a clear rule: no client data in any tool that isn’t approved for it.
3. Brought people with them
Rather than announcing a crackdown, partners explained the why. Staff had a short training session on using the approved tool well and checking AI output before it reached a client. Several people suggested new uses once they knew where the lines were.
4. Built a rhythm
The firm named an accountable partner and a day-to-day owner, set a quarterly review, and started keeping evidence in one place.
The outcome
When the next client questionnaire arrived, the firm answered it in an afternoon, with specifics instead of reassurances. Partners stopped worrying about what they couldn’t see and started asking where else AI could help.
Why it works
The firm didn’t start with a policy or a ban. It started with visibility. Once it could see what was happening, the right rules were obvious, and staff were more willing to follow them because they made sense.
You can’t govern what you can’t see.
