“We’re not really using it yet.”
That’s what most professional services leaders respond with when we ask them about AI. But a lot of them don’t notice their associate using ChatGPT to summarize a deposition transcript, their bookkeeper running financial models through Copilot, or their marketing coordinator drafting client newsletters in Gemini. By that point, the actual answer is “We’re using it everywhere; we’re just not aware of it.”
What leadership believes is happening compared to what is actually happening on the ground is the gap that matters most about AI in your business right now. Tool selection and vendor choice come later. What comes first is an honest assessment of your base.
Why the picture is fragmented
Managing partners, COOs, and practice managers are being asked to make decisions about AI with information stitched together from trade press, LinkedIn posts, the occasional vendor pitch, and whatever their IT person has mentioned in passing. What’s usually missing is a clear-eyed view of where their own firm stands today.
Part of the reason is that “AI in business” is really two different things.
What AI in business looks like right now
The first category is generative AI tools that employees adopt themselves. ChatGPT, Claude, Gemini, Perplexity, and a growing list of specialist platforms for legal research, tax preparation, and document review. These tools live in a browser tab, cost twenty dollars a month or nothing at all, and don’t require anyone’s approval to start using.
The second category is AI built into software your firm already pays for. Microsoft 365 Copilot now runs inside Word, Excel, PowerPoint, Outlook, and Teams. It drafts responses, summarizes threads, analyzes data, and takes actions on documents. Microsoft’s own guidance for its Dynamics products shows the direction of travel. In many business tenants, Copilot features arrive activated by default, and administrators have to opt out rather than opt in. The same pattern is repeating across Google Workspace, Zoom, Slack, Salesforce, and most SaaS tools a modern professional services firm runs on.
To summarize, employees are using AI whether they know it or not. The meeting summary at the top of a Teams recording. The smart compose in Gmail. The “analyze” button in Excel. Each of these is an AI model processing your work content.
Where it’s probably already happening
In our experience supporting DMV professional services firms, AI is in use across at least four parts of the team, regardless of whether leadership has formally approved it.
A paralegal pastes a stack of deposition transcripts into ChatGPT and asks for a three-page summary. The lawyer reviewing the brief gets a cleaner document, faster. Nobody logs the fact that those transcripts, containing privileged material, left the firm’s systems.
An accountant copies a client’s trial balance into Copilot inside Excel and runs a variance analysis against prior quarters. The output goes straight into a working paper. Nobody notes which model processed the numbers, or whether the engagement letter permits it.
A business development manager uses Gemini to draft a pitch for a prospect, pulling in phrasing from the firm’s previous work. The draft goes out after light editing. Nobody realizes the model also had access to the source folder, which happened to include a competitor’s commentary a team member had saved for reference.
An associate records a client call using a free AI notetaker they installed on their own laptop. The transcript is more accurate than their handwritten notes. The recording consent requirements in their jurisdiction were not checked, and neither was the tool’s data retention policy.
None of these employees are trying to cause a problem. They’re trying to do their jobs faster.
The part that gets labeled ‘shadow AI’
When this use happens outside sanctioned tools and policies, it gets called shadow AI. It’s a real risk and worth naming. A 2025 MIT study found that employees at over 90% of companies regularly use personal AI tools for work, while only 40% of companies have paid for any official AI subscription. Most AI use inside most firms is happening off the official record.
But shadow AI is only one symptom of something larger. Framing it as the central issue misses the bigger picture. AI is already embedded across multiple layers of your firm’s operations, in tools employees have chosen for themselves and been given, and in tools nobody consciously switched on. Shadow AI is only the part that’s easy to see.
What leadership typically doesn’t see
Three things, in our experience:
The full extent of employee usage
Self-reported numbers almost always understate the reality, partly because staff don’t want to admit to unapproved tools and because they don’t count the AI features built into software they already use.
Which data is moving where
Most firms can’t trace what exactly has been processed by an external AI model in the last thirty days or which tenants the resulting outputs were stored in.
Whether the insurance, regulatory, and client contractual positions line up
Professional liability policies, state bar guidance, and client engagement terms are all starting to address AI use. Many firms haven’t yet checked whether their current practice is compatible with any of them.
You can’t climb on ground you haven’t checked
Every firm we’ve worked with that has handled AI well started with a proper accounting of where they stand. What’s in use, by whom, with which data, under what controls, and against what obligations. That’s your base camp. Before a firm decides where to go with AI, it needs to know the ground it’s standing on.
You can’t build a sensible AI policy on an inventory you don’t have. You can’t train your team on tools you haven’t mapped. And you can’t tell a client their data is safe if you can’t name every model it’s touched. Every climb starts at base camp – book a call, and we’ll check the ground with you.
