Key takeaways
- An AI rollout has two halves: infrastructure (identity, permissions, licences) and adoption (provider choice, policy, training, use cases).
- Most MSPs are strong on the first half and lighter on the second.
- Reseller relationships can tilt an MSP towards the suite assistant. That may be right, but check the reasoning.
- The best structure is often a split: AI specialist leads, MSP handles identity.
If you have a managed service provider, they have probably already mentioned AI. For many small and mid-sized businesses the MSP is the most trusted technical partner they have, so it is natural to hand them the AI project too. Sometimes that is exactly right. Often it gets you a well-configured tool that nobody uses.
The two halves of an AI rollout
| Infrastructure half | Adoption half |
|---|---|
| Single sign-on and identity | Choosing the provider on the work, not the licence |
| Conditional access and device policy | Matching models to jobs |
| Licence purchase and assignment | Usage policy staff actually follow |
| File permission clean-up | Training built around real tasks |
| Data loss prevention settings | Spend caps and usage review |
| Offboarding | Measuring adoption and fixing what is not working |
MSPs do the left column every day for other systems. The right column is a different skill, closer to change management than to IT, and it is where most rollouts succeed or fail.
When your MSP is the right choice
- You have already decided on Copilot or Gemini and need the tenant prepared.
- The main risk is permissions: overshared files, old sharing links, too many global admins.
- Your MSP can show you two or three clients where adoption is high three months after go-live.
- They are willing to set hard spend caps and explain them, rather than just assigning licences.
Where MSP-led rollouts usually fall short
The tool is chosen by the licence, not the work
Many MSPs are Microsoft or Google partners. Recommending the suite assistant is the path of least resistance and sometimes the most profitable. For the majority of staff that recommendation is often correct. For a bids team or an analyst team, a standalone assistant may be much better. See Copilot vs ChatGPT Business.
Training is a generic webinar
"Here are the features" training produces a spike of curiosity and a flat line of usage. What works is showing each team three tasks from their own week, done properly with the tool. That needs someone who understands the work, not the console.
Nobody writes the policy
Infrastructure teams set technical controls. Somebody still has to write down, in plain language, what staff may put into the tool, what they must check before sending output to a client, and who to ask. Without that, people either avoid the tool or use it carelessly. Our AI usage policy guide covers what to include.
Success is measured as "licences assigned"
A rollout is done when people are using the tool weekly for real work, not when the last licence is assigned. Ask your MSP what they measure at 30 and 90 days.
The split that usually works
- AI specialist: chooses provider and tier, writes the implementation plan, configures the AI tool's admin settings and spend caps, writes the usage policy, runs training.
- MSP: fixes permissions, configures single sign-on and conditional access, assigns licences, handles offboarding.
- You: name an internal owner who reviews usage monthly and decides where seats go.
This is how our Implementation Package works where a client has an IT team or provider: we do the AI half and work alongside them on the identity half. If you want to compare all five ways of getting the work done, read who should implement AI in your business.
Frequently asked questions
Is my MSP qualified to roll out AI?
Most MSPs are well qualified for the identity, security and licensing side. Fewer have experience choosing between AI providers on the work, mapping models to tasks, writing usage policy and training staff on real jobs. Ask for two examples of each before deciding.
Will my IT provider recommend the best AI tool for us?
Not always. MSPs often hold reseller or partner arrangements with Microsoft or Google, which makes the suite assistant the natural recommendation. That may well be right, but you want the reasoning to be about your work rather than their margin.
Can an AI consultant and our MSP work together?
Yes, and this is often the best structure. The AI specialist leads provider choice, configuration of the AI tool, policy and training. The MSP handles single sign-on, conditional access, permissions clean-up and licence assignment.
What should we ask our MSP before giving them the AI project?
Ask which AI tools they have rolled out and for how many clients, how they decide between providers, what spend caps they set, who writes the usage policy, and how they measure adoption after 30 days.
Get a straight answer for your business
The Clarity Package gives you a written business case and a detailed implementation plan for 500 USD. The Implementation Package does the whole build, live in 30 days, for a fixed 5,000 USD.