guavaHow should an MSP handle AI policy and shadow AI for its clients?
Treat it as a recurring service. Give each client an AI acceptable-use policy, keep an inventory of the AI tools, connectors and agents its people use (shadow AI), review access and connectors, and hold a quarterly governance review. Guava’s Ongoing Optimization & Governance ($4,000 a month) sets it up in your own MSP and coaches your team to offer it.
What AI governance covers
- Acceptable-use policy Which AI tools are approved, what data can go into them, who approves a new tool, connector or agent, and how people report a mistake.
- Shadow-AI inventory Every AI tool, connector and agent in use, including the ones nobody signed off: personal ChatGPT or Claude accounts, browser extensions, and apps connected to company email or files.
- Access and connector reviews Who holds admin rights, which Claude and ChatGPT connectors and Copilot agents can reach which data, and which should be switched off.
- Quarterly governance review What was added, what was removed, and what the client signs off on for the next quarter.
- Incident guidance What to do when someone pastes the wrong data into an AI tool, or an agent does something it shouldn’t.
Start with your own shop
Run it on your own MSP first. Your techs probably already use AI tools on their own, which is normal and where governance starts: inventory them, write your own policy, and review your own connectors. Then you sell governance you already practice. New tools, connectors and agents keep arriving, so governance sells like patching: every month, not once.
How Guava does it
- Ongoing Optimization & Governance $4,000 a month after any engagement, cancel anytime. The policy, inventory, reviews, quarterly review and incident guidance above, with your team coached to run governance itself and to offer it to your clients.
- AI Rollout $25,000, 3–6 weeks. On Microsoft 365 its Ready phase plans the oversharing cleanup and sets sensitivity labels before Copilot goes live.
- Implementation & Offer Launch $35,000, 3 months: MSP to MIP in 90 days. Governance, readiness and rollouts packaged and priced as AI services for your clients.
Fixed fees, agreed in writing before work starts, and you own everything built. See the engagements
What happened at one firm
As Chief AI Officer at an SF-based tech staffing company, I set up its AI services as an Assessment, then an implementation, then ongoing support. 80% of its implementations stayed on for ongoing support. That’s one firm, not a forecast for yours. See the track record
Questions MSP owners ask
What is shadow AI?
AI tools people use at work without sign-off: personal ChatGPT or Claude accounts, browser extensions, and apps or agents connected to company email and files.
Should clients ban personal AI tools?
Usually not. A ban tends to push the use out of sight. Give people an approved tool and a clear policy, then inventory what’s left.
What goes in an AI acceptable-use policy?
The approved tools, what data can go into them, who approves new tools, connectors and agents, and how to report a mistake.
How often should AI governance be reviewed?
Quarterly, with an access and connector review whenever a new tool, connector or agent is added.
Will Guava deliver governance to our clients for us?
No. I coach your team to run governance itself and to offer it to your clients, so your people deliver it without me.
More questions MSP owners ask
Talk it through.
Twenty minutes on what AI your team and clients already use. Prefer email? romeo@workwithguava.com