AI Consulting and Automation for MSPs, Run by a Former MSP COO
I spent 18 years in the MSP industry and ran operations as COO while one MSP grew from $2MM+ toward $10MM. Now I help MSP owners put AI to work inside their own shop first, then sell it to clients without creating a churn problem.
In 2026, 7 MSP partners I work with launched profitable AI practices. Together they added about $120K in new AI MRR and about $75K in monthly recurring profit.
Two Reasons MSP Owners Hire AI Help
Most owners who call me want one of two things.
The first group wants AI and automation inside the MSP. Fewer hours on triage, documentation, reporting, and follow-up. More tickets per tech without another hire. Most of this page covers that work.
The second group wants to sell AI to clients as recurring revenue. That takes a service catalog, pricing, delivery, and enablement, and the AI Practice Build page covers it.
Do the first before the second. Your team needs to use AI every day before you sell it to anyone.
What 7 MSP Partners Got From AI in 2026
Every MSP partner I worked with in 2026 launched a profitable AI practice. Seven partners. Seven profitable practices.
7
MSP partners
~$120K
new AI MRR
~$75K
monthly recurring profit
$30K
new AI MRR from one partner
One partner alone added $30K in new AI MRR, and $20K of it is profit every month. I don't name partners. The numbers are current as of October 2026.
Where MSPs Win With AI, and Where They Waste a Year
Ticket triage and categorization
This is the highest yield place to start in almost every shop I've worked inside. It's also where the data problem shows up first. Before any model touches your queue, your board needs a categorization scheme a human defends. Most PSAs I open have 60+ ticket types. Half go unused, and three mean the same thing. Collapse the list, retag the last 90 days, then automate. Skip that step and you'll spend a quarter arguing with a bot about what "network issue" means.
Documentation and SOP cleanup
AI turns a resolved ticket into a draft runbook in seconds. It has no idea whether the fix was right. The workflow that holds up: the tech closes the ticket, the model drafts the doc, and one named engineer approves it before it enters the knowledge base. One approver. No committee. Shops that skip the approval step end up with a documentation library nobody trusts, and that's worse than having none.
QBR and reporting prep
A vCIO burning hours per QBR on data assembly is the most common hidden labor cost I find. Automate the gathering. Keep the narrative and the recommendation human, because clients spot a machine-written deck fast. The full method is in How to Automate MSP Client Reporting.
The year MSPs waste: client-facing agents first
Client-facing autonomous agents get the most attention right now. Build them last. They need clean documentation, stable process, tight permissions, and a signed governance policy. Owners who chase deflection first spend twelve months and real money to end up with an escalation path their techs route around. Do the boring layers first. They pay better.
An AI Offering Without Enablement Is a Churn Driver
If your MSP sells AI and doesn't train the people using it, your AI offering is a churn driver.
The client buys seats. Nobody shows their staff how to use them. Usage drops after the first month. At renewal, the client sees a line item with nothing behind it, and you're explaining why they should keep paying.
The fix is cheap. Put two things in the base price of every AI offering:
- ▸Onboarding sessions when each client goes live
- ▸How-to training materials. Pre-recorded videos and walkthrough documents work fine.
When a client wants more direct help, like a better prompt for their sales team or a workflow built around their intake process, that's billable work. Scope it and charge for it.
Who Should Own AI Inside Your MSP
Look for the person already leading the internal AI charge. Most MSPs have one. They built their own prompts, automated part of their own job, and keep showing everyone else what they found. If they're good and they know the tools, give them ownership.
Nobody fits? Start training someone now. If you need results faster than training gets you there, bring in a partner while your person catches up.
Build Automations in House or Outsource Them
Build in house when you have someone who works like a developer and you'll give them at least 25% of their time for it. Working like a developer means following a development lifecycle: requirements, testing, version control, documentation, and a rollback plan. Automations built by someone squeezing the work in between tickets break quietly, and nobody knows how to fix them.
Without that person and that time, outsource the build.
How the First 90 Days Run
- 1. Weeks 1 to 3: Where the labor goes. Ticket volume by type, time to resolve, rework rate, and the manual steps nobody wrote down. We find the workflows with real hours attached.
- 2. Weeks 4 to 6: Governance and data hygiene. An AI usage policy your team follows, permission boundaries, human review points, and enough PSA cleanup to give the automation honest data.
- 3. Weeks 7 to 12: One workflow shipped and measured. A named owner, a baseline, a post-change number, and a decision about what comes next based on that number.
If the result isn't there, I say so and we stop.
Book a 30-Minute CallPlatforms Worth Reviewing
I don't push one platform, and I take no referral fees. Do your own due diligence and pick what fits your team and your clients. These are the ones I tell partners to review:
How to Judge Any AI Consultant for Your MSP
Use this list on anyone you talk to, me included.
- ▸They've held an operating seat inside an MSP. Ask which seat and for how long.
- ▸They name the PSA and RMM platforms they've worked in.
- ▸They want to see your ticket data before they recommend a tool.
- ▸They ship one measured workflow before they sell you a roadmap.
- ▸They take no referral fees from the platforms they recommend, or they tell you when they do.
- ▸Enablement is in their plan.
- ▸They tell you when AI is the wrong fix.
What This Looked Like at a $3M MSP
A $3M MSP came to me with automation tools already paid for and nothing running. Service issues and incident response ate every hour the team had. We put automation behind process stability, installed an operating rhythm, and started turning the tools they owned into margin instead of shelfware.
"It has been so great to work with Mike and have a sounding board to help reduce decision fatigue and get the right product for the right solution."
Brian G.
Founder, Los Angeles Based MSP
"Mike has been able to guide us towards real organization and stability in our processes, immediately reducing chaos and the constant need for band-aid fixes."
Evan D.
Service Manager, Michigan Based MSP
Who This Fits
A fit:
- ▸MSP owners at $2MM to $10MM who want AI tied to margin and service delivery
- ▸Shops with AI or automation tools paid for and nothing running
- ▸Owners ready to give one person real time to own AI
- ▸MSPs that want to sell AI and keep the clients who buy it
Not a fit:
- ▸General AI curiosity calls
- ▸Tool-first projects with no business goal
- ▸Owners who want a client-facing chatbot before their own documentation is clean
FAQ
What does AI consulting for an MSP cover?
Two sides. Inside the MSP, I find the workflows with real hours attached, set an AI usage policy, clean up enough PSA data to automate on, and ship one workflow with a measured result. On the client side, I help you build the offering, the pricing, the delivery model, and the enablement so clients keep using what they buy.
How is AI consulting for MSPs priced?
I work on monthly retainers, month to month, scoped by how much of a COO or CTO seat you need each week. I don't bill hourly. Project work is on the table when the scope is clear. We set scope on the first call.
How long before an MSP sees a return on AI automation?
Plan on 90 days for the first measured workflow. Weeks 1 to 6 go to finding the hours and cleaning the data. The first workflow ships and gets measured in weeks 7 to 12. All 7 partners I worked with in 2026 launched profitable AI practices that year.
Should we fix our process before we automate?
Yes. Automation runs your process at speed. If ticket categories are a mess or nobody owns the workflow, automation makes the mess faster. Clean up the one workflow you plan to automate. You don't need to fix the whole shop first.
Which AI automation platform should an MSP use?
The one that fits your stack, your team's skills, and your clients. Review Rewst, n8n, Microsoft Copilot, Synthreo, and Hatz AI, then test against one real workflow before you sign anything. I take no referral fees, so I have no stake in your pick.
Do we need to hire a data scientist or a developer?
No data scientist. You need one person who works like a developer and gets at least 25% of their time for automation. Without that person, outsource the build.
What's the difference between AI consulting and an AI practice build?
AI consulting puts AI to work inside your MSP. An AI practice build turns AI into a recurring service you sell to clients. Most MSPs need the first before the second. See AI Practice Build.
Want AI Running Inside Your MSP Within 90 Days
Book a 30-minute call. Bring your PSA and the one workflow eating the most hours. You'll leave with a straight answer on whether AI fixes it.
Book a 30-Minute CallSelling AI to clients? See AI Go-to-Market for MSPs.
Reading first? Start with AI Use Cases for MSPs, the MSP Automation ROI Framework, or RPA vs AI Automation.