How MSPs can add AI services: what to sell first and when to bring in a partner

Start with a fixed-fee workflow assessment, follow it with a pilot on one repetitive workflow that's easy to undo, and then sell monthly AI management on top of the managed contract. Bring in a partner for the architecture, integration and security design you can't staff yet, and agree up front who owns the client and the support.

By Sagar Pandya, Co-Founder and CEO, Spej. Exited MSP founder and author of The AI Culture Blueprint.

I came up in managed IT. My first real IT job was at 17, working overnight inside a data center, and I later founded an MSP and exited it. So when MSP owners tell me their clients keep asking about AI and they don't know what to sell, I recognize the spot they're in. The client trusts you with their technology, the ask is vague ("can you set up AI for us?"), and you don't have an AI team sitting on the bench.

Why clients are asking, and what they usually mean

The demand is real, and most MSPs haven't turned it into revenue yet. In Kaseya's 2026 State of the MSP Report, 48% of MSPs ranked AI and automation as the top client need for 2026, and 13% said they were generating meaningful revenue from those services. Nerdio's October 2026 State of the MSP Business Report, which surveyed 153 MSP decision-makers managing Microsoft 365 and Azure environments, found that only 14% operated a revenue-generating AI service line, and only 14% consistently applied a documented AI governance framework across every customer using Copilot or other AI tools (Nerdio, via Telecom Reseller).

TD SYNNEX's fifth annual Direction of Technology report puts the gap in a wider channel context. Drawing on insights from more than 1,500 channel partners across 45 countries, it found that 66% of IT partners are increasing investment in AI solutions and services, but only 21% are generating meaningful revenue from those investments. Only 13% said most end-customer AI pilots successfully convert into scaled production deployments (Channel Insider, November 4, 2026; TD SYNNEX report announcement). These are channel partner findings, separate from the MSP surveys above.

The reported end-user barriers were security, legal and governance concerns (42%), an unclear business case or ROI (39%), and a lack of internal skills to manage AI after deployment (35%). My read: the assessment should settle the business case and surface governance gaps, and monthly management should give the client someone to own the work after deployment.

There is also a reason to start with a conversation you can explain plainly. In competitive AI deals, partners cited existing relationships and trust (45%), translating AI complexity into plain business language (37%), and knowledge of the customer's industry or business problem (34%), ahead of the strongest technical capabilities or certifications (32%) (TD SYNNEX, via Channel Insider). Your client relationship gives you a place to start. You still need the delivery skills behind the promise.

When a client says "AI," they usually mean one of three things: their people are already using ChatGPT or other tools and leadership is nervous about it, they bought Copilot seats and usage is flat, or there's one painful workflow (quotes, claims files, reports assembled by hand) they want fixed. Each of those maps to one of the offers below.

You don't need every answer before you start. On our October 30 live stream I said, "you don't have to be fully prepared with all the answers. You just gotta have the conversation. And that conversation itself is the thing that gets the AI started." You just have to be trusted enough to have it with your customers, and as an MSP you already are.

Offer 1: the workflow assessment

The assessment turns a vague AI ask into a ranked list. List the client's recurring workflows, roughly how much time each takes, the systems involved, and where a person has to approve something. Then rank them by value, effort and readiness. Check what AI tools people already use, including ones outside company approval, and whether the data the best candidates depend on is accurate and accessible.

The client should leave with a short ranked list, a recommended first pilot with a baseline to measure against, and the gaps to fix first. Sometimes the honest answer is "not yet," and the client is better off hearing it before they spend on a build.

Price it as a fixed fee. For reference, Spej's AI Roadmap, which does this work in more depth with a business case, is starting at $10,000. The final price depends on scope. When we worked as a regional P&C insurer's retained AI Office, the first two weeks were stakeholder interviews across seven departments, an audit of existing tool usage and a data readiness check (case study). For a national mechanical contractor, a readiness assessment across data, process, organizational readiness and technical capability found six high-ROI opportunities (case study).

Offer 2: the one-workflow pilot

Pick a workflow that happens often, can be reversed if something goes wrong, and is easy for a person to check. Record the starting number before you build: how long it takes today, how often it goes wrong. Then build it inside the systems the client already runs and measure the same thing afterward.

Our Design & Build work, starting at $20,000, follows that shape. Scope, price, target and measurement period are agreed upfront, there's a weekly 30-minute check-in, one use case goes into production in four weeks, training is included, and the client owns the code and documentation. The contractor's job intelligence prototype is a good example of a pilot scoped to data the company already had, using systems it already owned, so leadership could decide on a bigger investment with something real in front of them.

Offer 3: monthly AI management

This is the recurring line, and it fits naturally next to a managed IT contract. What goes in it:

• Governance: an acceptable use policy, who can approve new tools, and a regular review of guardrails.
• Access and permissions: Microsoft says Copilot only surfaces organizational data that a user has at least view permission to (Microsoft Learn). If last summer's interns still have access to the finance folder, AI will find it. Reviewing that before and after rollout is work an MSP is already set up to do.
• Usage and cost reporting: who is using what, and whether it's working.
• Training for the people using what's deployed.
• The next improvement, chosen with the client.

Spej's AI Office, starting at $10,000 a month, is a model of this. It keeps one backlog tied to value, risk and readiness, puts one initiative into production a month, reviews guardrails, sends a one-page monthly report, and sets the next three priorities at each quarterly review. At the insurer, governance meant an AI acceptable use policy, a model risk management process aligned to their existing enterprise risk standards, and a monthly AI steering committee with the C-suite.

What to keep in-house and what to hand to a partner

Keep the client relationship, intake, identity and access, and day-to-day support. Those are already yours and they're what the client pays you for.

Hand off, or co-deliver, when you need architecture and integration across several systems, security design for regulated clients, testing how a system behaves under real conditions, or adoption training at scale. There's a real shortage of people who can do this work correctly, and one freelancer can't cover all of it.

Three ways to work with a partner

Disclosure: Spej offers all three models below through our partner program. This is how we work with MSPs, and we have a commercial interest in being one of your partner options.

That's how Spej's partner program is set up. White label is especially useful for larger MSPs expanding their services. For referral economics, see AI referral fees for consultants and MSPs, and for the platform question, see managed AI platform or white-label delivery partner.

Where AI falls short for MSP clients

• Vague asks produce vague projects. If the client can't name the workflow and the number it should move, sell the assessment first.
• AI shows you whatever is already overshared. Messy permissions become an AI problem the first week.
• Messy data costs money every time AI reads it, because the model spends effort working out what each piece of data is before it does the task.
• A pilot nobody uses doesn't pay back, however well it's built. Budget training and a measured before and after into the pilot.
• Don't resell what you can't support after the partner leaves. Agree support ownership before you sign.

A first 90 days for an MSP

1. Pick three clients who have already asked about AI.
2. Run the workflow assessment with each and agree a baseline for the top workflow.
3. Run one pilot with the client whose workflow scored best, and measure it.
4. Write the result up for that client's next QBR and offer monthly AI management from there.

If you want a team behind you while you do it, talk with our partner team. We bring in the AI horsepower to be a friendly ally in the conversation, and you stay the advisor your clients turn to.

FAQ