Most MSP owners I talk to place themselves one level higher on the AI maturity ladder than they actually are. They bought a couple of AI-enabled tools, someone on the service desk uses ChatGPT to draft ticket responses, and they’ve decided that makes them an AI-forward shop. It doesn’t. It makes them a Level 1.
I’ve been watching MSPs adopt (and avoid) AI for the past three years, across coaching clients, peer groups, and my own operation. The pattern is consistent enough that it’s worth naming. Here’s the ladder. Find yourself on it honestly, because the gap between levels is widening fast, and it compounds.
First, drop the idea that AI is optional
AI is part of the MSP industry now. Period. It’s not a specialty, not a side interest for the one tech who likes to tinker, not a product category you evaluate when a vendor calls. It’s a core competency, the same way Windows and M365 are core competencies. You wouldn’t keep a tech who couldn’t hold a conversation about Entra or Intune. Within a year or two, the same standard applies to AI, and your clients will enforce it whether you do or not.
Because your clients are already asking. They’re getting pitched AI by their software vendors, reading about it in their industry publications, and hearing their competitors talk about it at trade association meetings. When they ask their MSP, and they will ask, “should we be using Copilot?” or “is it safe to put our data in ChatGPT?”, every person on your team who touches that client needs to give an intelligent answer. Not a deep answer from everyone, but a competent one. “I don’t really follow the AI stuff” from a senior tech is the 2026 equivalent of “I don’t really do cloud.” It tells the client to start shopping.
The best MSPs go one step further. They have someone, often the owner, sometimes a vCIO or account manager, who understands business well enough to have a real business conversation about AI with a client’s leadership: where AI fits their operations, what it costs, what the risks are, what to do first. That conversation is where AI stops being a topic and becomes a project. And then the MSP implements it. The shops that can run that full loop, business conversation to implementation, are converting AI curiosity into some of the highest-margin project and advisory work available in this industry right now. The shops that can’t are watching that revenue go to consultants who will never touch a firewall.
That’s the backdrop for the ladder. The levels below aren’t about how enthusiastic you are about AI. They’re about how much of that loop your shop can actually run.
Level 0: The holdouts
These are the MSPs that either actively resist AI or have parked themselves in “wait and see” mode. Some have legitimate-sounding reasons: security concerns, client contractual restrictions, bad early experiences with hallucinating chatbots. Most just haven’t made it a priority.
Wait and see felt reasonable in 2023. In 2026 it’s a strategic decision to fall behind. Your competitors are compressing ticket resolution times, cutting documentation hours, and quoting projects faster than you can. And a Level 0 shop can’t run any part of the client conversation described above, which means every AI question from a client is a moment of visible weakness.
The security objection deserves a straight answer, not dismissal. Yes, you need a data handling policy. Yes, you need to know where client data goes. Those are solvable problems that Level 3 and Level 4 shops solved a while ago. They’re not a reason to stand still.
Level 1: AI as the new Google
This is where the majority of MSPs sit today. Someone on the team uses ChatGPT or Claude the way they used to use Google: paste in an error message, ask how to write a PowerShell script, get an answer, move on. Maybe the shop bought an AI-enabled product or two, something that summarizes tickets or suggests responses in the PSA.
This is real value. It’s also table stakes. Individual technicians get faster, but nothing about how the business operates has changed. The tools are doing what the vendor designed them to do, and the shop is a passenger. If your AI strategy is a list of products you purchased, you’re at Level 1.
Level 2: Deliberate operators
Level 2 is where usage stops being ad hoc and starts being intentional. The shop has rolled out Copilot seats or Claude accounts with actual guidance on how to use them. Individuals build structured prompts, projects, and reusable workflows for the tasks they do repeatedly: QBR prep, proposal drafts, runbook documentation, client communications. Someone owns the AI question internally, even if it’s not their whole job.
The tell for Level 2: individuals are measurably faster, but the processes themselves haven’t changed. Same workflows, same handoffs, same org chart. The tickets flow the same way they did in 2022; people just move through their pieces quicker. That’s a fine place to pass through and a dangerous place to camp, because the shops above you aren’t making people faster at the old process. They’re replacing the process.
Level 3: Process transformation
Level 3 shops have redesigned how work gets done, internally and sometimes for clients. They’re using platforms like Claude, Claude Cowork, and Microsoft Copilot not as a smarter search box but as a working layer inside the business. Skills and standing instructions encode how the shop does things. Monthly billing reconciliation that took an accounting person half a day runs as a guided workflow. Meeting transcripts flow into the CRM without anyone retyping notes. Client-facing deliverables get produced from templates the AI knows how to fill.
The difference from Level 2 is structural. At Level 2, if the person with the good prompts leaves, the capability leaves with them. At Level 3, the capability lives in the business: documented workflows, connected systems, repeatable outputs. Level 3 is also where the client-facing loop closes. Because the shop has transformed its own operations, it can walk into a client’s business, have the business conversation with credibility, and then actually implement what it recommends. You can’t sell what you haven’t done. This is also where AI starts showing up on the P&L in ways you can point to, not just “the team feels faster” but actual hours reclaimed and margin captured. A shop running 62% LLGM on managed services got there partly because service delivery costs came down while headcount held.
Level 4: Agentic operations
The top of the ladder, and it’s sparsely populated. Level 4 shops have agents doing actual work: multi-agent platforms like OpenClaw, Hermes-style frameworks, and GrokBot handling defined jobs end to end. Not suggesting a ticket response for a human to approve. Monitoring the sales pipeline overnight and flagging deals that went quiet. Running the monthly close checklist. Watching alert queues and executing remediation runbooks within defined guardrails.
The distinction that matters: at Level 3, AI transforms processes that humans still drive. At Level 4, agents own outcomes. A human sets the objective and the boundaries, then reviews results instead of performing steps. This requires real engineering investment, real security architecture, and real thought about what an agent should and shouldn’t be allowed to do. Nobody accidentally arrives at Level 4. But the shops operating here are doing things with 10 people that used to take 25, and some are packaging that capability as a service line for clients.
Where the money is
Level 0 and Level 1 shops compete on the same cost structure they had three years ago. Level 3 and Level 4 shops are pulling cost out of service delivery while holding price, which drops straight to true net profit. Best-in-class MSPs run 18% true net or better; the median sits around 8%. AI maturity is becoming one of the clearest dividers between those two numbers.
You don’t jump from Level 1 to Level 4. You climb one level at a time, and each level requires different decisions: Level 2 is a rollout and training problem, Level 3 is a process redesign problem, Level 4 is an engineering and governance problem. Knowing which problem you’re actually solving keeps you from buying tools that don’t match your level.
Figure out your next rung
Two ways we can help. First, join an MSP Advisor peer group. The fastest way to calibrate your own AI maturity is to sit in a room with owners who are a level or two ahead of you and see exactly what they’re doing, what it cost, and what broke along the way. Nothing punctures self-assessment like a peer showing you their working system.
Second, if you want direct help climbing, work with our MSP Advisor AI and Automation Sherpas. We’ll assess where you actually are, map the path to the next level, and build alongside your team instead of handing you a slide deck.
Either way, start with an honest answer to one question: what level are you at today? Not the level you’d tell a prospect. The real one.