Most clients already know their staff use AI. What they don't know is which tools, on which accounts, or with what information. That gap is where the conversation should start. When an MSP opens with risk, warnings and worst cases, the client hears a sales pitch. When it opens with a question, the client hears a partner.
Our view is simple: Governance starts with visibility. You can't govern what you can't see. Every step below follows from that.
Start With What They Already Use
Don't lead with what could go wrong. Lead with what is already happening.
A good first line is a question the owner can answer: "Which AI tools does your team use for work today?" Many owners can name one or two, and suspect there are others they couldn't name. That honest "I'm not sure" does more for the conversation than any warning you could give, because it's theirs.
Keep the tone neutral. Staff who use AI are usually trying to do their jobs faster, not causing trouble. If the conversation treats them as a threat, the owner will defend them, and the conversation stalls. If it treats AI as a useful tool that needs a few rules, the owner leans in.
Ask Before You Show
Before you show any tool or any data, ask five questions. They take ten minutes and tell you what the client actually cares about.
- Which AI tools have you approved, and with which accounts? Work accounts and personal accounts carry very different terms.
- What information would hurt most if it left the business? Client records, financials, contracts, source code, credentials. The answer sets your priorities.
- Who outside the business expects you to control this? Clients, insurers, auditors or regulators. This tells you which evidence will matter later.
- Do your staff know what is allowed? Often the honest answer is "probably not."
- If something went wrong, how would you find out? This is usually the question that turns the conversation from interesting to important.
Write the answers down. They become the first draft of the client's rules.
Show Their Own Picture, Not Headlines
Stories about other companies' AI mistakes are easy to find and easy to dismiss. The client's own picture is neither.
Offer a short, observe-only look: deploy our agent to the client's devices for two weeks, with every rule set to Monitor, and see which AI tools show up and how they are used. Nothing is blocked or redacted, and nobody's work is interrupted. Before you start, agree with the owner on how staff will be told what is being looked at, and have the client confirm with its own counsel what notice or consent the law or any employment agreements require. The visibility you gain should be visibility they know about.
At the end of the two weeks, walk the owner through what you found: the tools, the accounts, and the kinds of information that went into them. The conversation that follows will be about their business, not about a headline. This is the point where visibility turns into governance, because now there is something real to decide about.
Talk About Decisions, Not Bans
Once the owner can see what is in use, the question becomes what to allow. Frame it as a set of decisions, not a list of things to shut off.
- Approve the tools that earn it. Pick one or two AI tools the business can stand behind, and move staff onto work accounts.
- Set rules for sensitive information. Decide what should be warned about, redacted or blocked before it reaches any AI tool, whichever tool it is.
- Block only what's risky. Reserve blocks for the apps or uses the client has decided against.
Blanket bans risk pushing AI use onto personal phones and home laptops, where nobody can see it at all, and that works against the whole point. A client that approves a tool and sets clear data rules is, in our view, in a better position than one that tries to block everything.
Put It in Writing
Decisions that live only in a meeting fade quickly. Write them down in an AI Use Policy: which tools are approved, what information must never go into AI, what staff should do when they are unsure, what the business monitors, and how much of a prompt is kept.
Keep it short enough that people will read it. Ask each employee to accept it, and keep a record of who accepted which version. If you need a starting point, our AI policy template covers the common sections. The client owns the final wording, and for anything touching employment law, they should have their own counsel review it.
Make It a Standing Agenda Item
AI tools change fast. New ones appear, existing ones add features, and staff try new things. A policy written once and never revisited will drift from what really happens.
Make AI use a standing item in your regular business reviews. Show what changed since last time: new tools that appeared, decisions that need making, and anything the data rules caught. Ten minutes a quarter keeps the policy and the practice in step, and keeps you in the conversation as the partner who saw it first.
The Short Version
- Open with a question, not a warning.
- Ask what matters to the business before you show anything.
- Show the client their own picture with a short, observe-only look.
- Turn what you find into decisions: approve, set data rules, block only what's risky.
- Write it down and have staff accept it.
- Revisit it every quarter.
Governance starts with visibility. You can't govern what you can't see. Once a client can see how AI is actually used in their business, the rest of the conversation gets much easier, for them and for you.
This article is general information, not legal, compliance or security advice, and does not create a professional relationship. Product details are accurate as of the publish date; see ProctoredAI.com/terms for what the service does.