K7 Insights
How to Sell AI Services as an MSP: The Three Offers
Part of Grow Your MSP
Key takeaways
- Enablement helps employees get useful work done with AI and recognize where they need more help. That has value even when the client never buys a custom build.
- Sell a foundation engagement that ends, ongoing managed enablement, and separately priced implementation. Reuse learning content and session materials, then apply them to each client's work.
- Start with a client who has already asked about AI. Prove the first engagement yourself, document it with your delivery lead, and give that person the authority to quote additional work.
To sell AI services as an MSP, help your client’s employees use AI in their daily work. Package that enablement as a foundation engagement, offer ongoing learning and leadership guidance, and price implementation separately. Your client gets a practical place to start. You get an offer your team can deliver again without promising your help desk will build whatever someone dreams up. Here’s why I’d start there, how I’d deliver it, and what I’d count before calling it profitable as part of growing your MSP.
Why start with AI enablement?
Enablement helps employees do useful work with tools they already have while you learn where bigger improvements are possible. An owner asking about AI may be thinking, “Can my current team handle more work?” Ask them. I raised this in the webinar with John Harden published September 3, 2026: our concerns as MSPs can crowd out what the customer came to buy.
John made the case for teaching employees before waiting on a long custom build. Useful changes can reach several departments while that project is still being scoped. Test whether they help, including the time employees spend checking and correcting the result.
Employees also need permission to try. Which tool can I use? What information can I put in it? Who do I ask when I’m unsure? Answer those questions, let people practice, and give them somewhere to bring the result. That builds confidence they can use at work after the demonstration ends.
It also changes discovery. After trying AI on a routine task, an employee can tell you where it helped and where they got stuck. As I explained in the same webinar, education helps employees recognize opportunities in their own work. Some will need a project. Some will only need a better way to use the tool. Both are useful outcomes for the client.
What AI services should you sell first?
Sell three distinct offers: a foundation engagement, ongoing managed enablement, and implementation. I call this the ladder in my August 5, 2026 talk on selling AI services. It’s a model to test against your own clients and costs.
| Offer | What the client buys | Where it ends |
|---|---|---|
| Foundation | Six months of guided employee learning, an AI use policy, and a list of useful opportunities | The agreed sessions and deliverables are complete |
| Ongoing managed enablement | Continued employee learning, leadership guidance, and policy upkeep | The agreed subscription scope and support allowance; additional work gets a proposal |
| Implementation | A specific improvement to a business workflow, delivered as a project | The agreed result and handoff; continuing support has its own terms |
You also don’t need to march every client through all three offers. If they’ve already done the foundational work, meet them where they are. And if six months of help is all they need, finish the engagement well.
My broader AI for MSPs guide covers what AI changes inside your own shop. This is the offer you put in front of a customer.
What belongs in the foundation engagement?
The foundation is a project with an end date. I’d start with six months and one live session a month, supported by short learning assignments. The client keeps an AI use policy and a list of opportunities found through the work.
Use the same delivery loop each month. Here’s an illustrative session on customer follow-ups:
- Before the session, assign a short lesson on giving AI useful context. Ask employees where writing follow-ups slows them down.
- In a 30- to 45-minute live session, demonstrate with fictional meeting notes in an approved tool. Then have an employee try it. Check whether the draft reflects the notes and whether anything was invented. Count the review time when comparing it with their usual process.
- Leave a short job aid and an assignment to try an approved task before the next session. Ask what they used, what needed correction, and whether they’d use it again. Course completion tells you they finished a lesson; their work tells you whether it helped.
- Bring the findings to the client’s decision-maker. If writing improved but copying information from the CRM still takes too long, record that problem, its owner, and a proposed next step. The leader decides whether it’s worth investigating.
Set the initial rules before the first exercise: approved tools, permitted information, and who reviews output. Refine the policy as questions come up, with the client’s decision-maker approving the changes. Employees should know how to ask for help without feeling they’ve done something wrong.
Put the limits on the same page as the offer. How many people or departments are included? How long is each session? How much help do they get between sessions? Custom integrations and workflow builds need their own agreement.
If your client asks for six monthly payments, fine. The project still ends in month six. Billing frequency doesn’t change what you sold.
How can Empath help you deliver this repeatedly?
Empath can supply the learning platform and content you use around the live sessions. I’m Co-Founder & COO of Empath, MSP education software. Empath Grow lets you manage separate client organizations from a shared pool of seats, assign courses, and add optional catalogs such as Bigger Brains. Choose material that fits the topic you’re teaching.
That lets you reuse foundational learning across clients and assign it to the next hire. I described the delivery approach in the Ladder talk: show someone how to do the task, have them do it, and assign supporting courses.
The full AI facilitator and curriculum kit I discussed is still unfinished. You can prepare your own session materials: a short agenda with speaking notes, a tested exercise and worked answer, and a job aid. Keep a question log that says who follows up and whether the work is included. Rehearse with the person who’ll deliver it.
You still lead the session and get decisions on what comes next. Reusing the content and preparation keeps each new client from becoming a new course-development project.
What makes ongoing enablement worth a monthly fee?
Ongoing managed enablement earns its fee by keeping employees able to use AI as their work and tools change. It includes employee learning and leadership guidance. For example, you might sell a monthly department session with assigned learning, onboarding for new hires, and a quarterly leadership review of results and next steps. Define the support allowance and policy-maintenance work too.
I learned the maintenance lesson with much less exciting software. At the MSP I ran, we spent four hours building a lunch-and-learn about Microsoft licensing for our engineers and customers. It helped for about three months. Then we hired new people who’d never seen it, and we were back where we started. I told that story in the same talk because the mistake carries straight into AI training: we’d built a session and left the next hire out of the plan.
At the leadership review, bring examples employees have actually used and decide what needs attention next. Your vCIO offering may already have the right person to guide that conversation. When a department spots a useful automation, they can say:
“We’ve found something worth building. I’ll put a separate proposal together so you can decide whether the result is worth the cost.”
That’s the boundary between enablement and implementation. State how any included working hours are scheduled, whether they expire, and what happens when the client needs more. If the advisor promises a build during the meeting, your service desk inherits the promise.
I call undefined AI work stuffed into an all-you-can-eat agreement “MRR cosplay.” You get to report more recurring revenue while your team figures out how much labor you accidentally sold.
Where do implementation projects fit?
Implementation is the separately priced work that turns an agreed opportunity into something the client can use. It can come out of either enablement offer. The client should be able to explain what will improve before you start selling the build.
Say their front desk wants help scheduling appointments. Teaching the team what’s possible belongs in the session. Connecting the scheduling system and making the workflow function is a different engagement, with its own delivery cost and person responsible.
Establish that you or a delivery partner can do the work before you commit to it. An advisor who can teach prompting hasn’t automatically become an integration engineer.
Licenses and usage fees follow the work that needs them. If there’s a resale margin, count it. Then price continuing support according to what you’re agreeing to maintain. A recurring license charge doesn’t cover an unlimited obligation to change the workflow.
How much should you charge for the first AI offer?
Build your price from the full delivery cost and the margin you need. In the talk, I used 16 hours at a $50 loaded hourly cost as a starting example for the six-month foundation. In this model, those hours include setup, session preparation and follow-up, policy work, and administration. It’s a planning estimate to test with reusable material already prepared.
Here’s a hypothetical 25-person engagement. The software allowance is a budgeting assumption, not an Empath quote.
| Cost over the six-month engagement | Assumption | Cost |
|---|---|---|
| Total client delivery labor | 16 hours at $50 loaded cost | $800 |
| Allocated software cost | Hypothetical six-month allowance | $300 |
| Total modeled direct cost | Labor plus software | $1,100 |
At $500 a month for six months, you’d collect $3,000. Subtract $1,100 and you have $1,900 in modeled gross profit, a 63.3% gross margin. The 65% target I proposed in the talk would require about $3,143 at that cost. Doubling your cost gets you to 50%.
Now check the commitment behind the software allocation. Grow’s published terms, checked September 18, 2026, start at 100 seats for 12 months, with pricing at checkout. If your first client uses 25 seats for six months, you’re still paying for the unused seats and the remaining term. That cash obligation belongs in your launch budget; it doesn’t disappear into future clients you haven’t sold.
Count the work to prepare reusable material too. Another 20 hours at $50 costs $1,000. Charged entirely against that first $3,000 sale, it leaves $900 after the modeled costs, before any unallocated software commitment, travel, selling costs, or overhead. Owner time needs a realistic cost as well.
Replace these assumptions with your actual hours and bills before quoting the next client. Price ongoing enablement from its own promised work using the same cost and margin method you’d use for a vCIO retainer.
If you’ve got a client in mind, bring your first AI offer to a call. We’ll pressure-test the delivery cost and who’s going to own the work before you quote it.
How do you make the first sale and hand off delivery?
Start with one existing client who has already asked about AI. Run the first one or two engagements yourself with the person who’ll take over beside you. Your job is to make the offer repeatable.
Ask what their employees are already trying and where they get stuck. You may find they need basic education, or a specific project they already want. Let the answer decide the offer.
For a client who needs the foundation, the conversation can be this simple:
“Yes, we can help your employees use AI in their daily work. We’ll meet once a month for six months, with learning assignments between sessions. You’ll leave with rules your people understand and a list of work worth improving. We’ll agree on the support included. Anything you want us to build gets its own price.”
You can use a lunch-and-learn to help them decide. If you make it free, choose its limits and count the cost as sales activity. Leave them with something useful, then ask for the engagement.
Have your delivery lead teach a session while you watch. Can they get the employee to do the task? Can they handle a question they don’t know the answer to without making something up?
They also need authority to hold the boundary. If every request for additional work comes back to you, you’ve kept part of delivery on your desk.
Give the help desk a useful route for those requests: “We can help with that. I’ll get our advisor involved so we can agree on the work.” Someone then has to own that follow-up. A ticket closed as “out of scope” leaves the client to find their own answer.
Count owner time during the pilot. Once someone else takes over, check their actual hours against the price too.
What’s the first move this week?
Write one offer a client can buy and your team can finish. Start with the next client who asks about AI, put the deliverables and limits on one page, and cost the work with the person who’ll eventually deliver it.
That’s the part of this ladder you can test now. The broader work of building an MSP that grows beyond its owner happens when you can sell that offer again and keep the promise without being in every session.
- Do I need to hire an AI developer before I sell AI services?
- You need someone qualified to deliver the work you sell. Teaching a client to use an approved tool and helping them choose a first use case can fit an advisor's skills. Integrating AI into a business system requires different technical capability. Find that capability, internally or through a partner, before you commit to the build.
- Does every client need to buy all three AI offers?
- No. A client may finish the foundation engagement and stop. Another may already have trained staff and an AI use policy, so ongoing enablement is the better starting point. A client with a defined need may buy an implementation project directly. Use the offers to match their readiness and keep the work clearly priced.
- Can I bill the foundation engagement monthly?
- Yes. Six monthly payments can spread the cost of a six-month project. Keep its end date and deliverables clear. Monthly billing doesn't make an engagement that ends into an ongoing subscription.
- Should AI licenses be included in the service price?
- State exactly which licenses and usage costs the fee covers, including any limits. You can bill them separately or include a defined allowance. Either way, count them in your delivery cost and explain what happens when usage exceeds the allowance. Buying licenses on the client's behalf also leaves the training and implementation work to price.
Watch the conversation