AI Closes the Tickets Techs Learned On
The work that used to teach a technician the job is leaving the shop, and so is the work at the top.
Start with IgnitHQ, which makes an AI-native platform for running an MSP. Founder and CEO Yaron Baitch told Channelholic’s Rich Freeman that about a quarter of its partners now let its AI engine, Duke, close tickets, when it can, with no technician involved. Another half require a person to approve. The last quarter still do the work themselves. Those partners have pushed about 10 million tickets through the platform. Baitch says trust usually comes after about a month of heavy use with someone watching, and that smaller MSPs get there first. His explanation: they can take on bigger projects without adding headcount.
ServiceNow, whose AI platform Thrive runs, put out a press release showing the same pattern at Thrive, a managed service provider. Thrive says 73 percent of support tickets are now resolved on first contact. The time it takes to connect a customer with an engineer fell from six minutes to under one. It counts an estimated 14,500 engineer hours returned to customer-facing work. That’s roughly seven technicians’ full working year.
The building is moving too. Rewst rebuilt its automation platform so users describe a workflow in plain language and the system builds it. Rewst says its early-access group, about 110 MSPs or roughly a tenth of its partners, created about 150 automations a day over one 30-day stretch. Every other Rewst partner combined made about 75. About 80 percent of those early-access automations were built from inside ChatGPT, Claude or another chatbot. CEO Aharon Chernin put it plainly: “The MSP isn’t using the product. Claude is.”
That’s the bottom of the ladder. At the top, the senior AI work is being offered for rent. On this show, Brian Kelly, who ran his own MSP before founding Automated Edge, described selling managed AI through MSPs under the MSP’s own brand. He told me every client assessment his company has run found staff already using tools like ChatGPT. He sells that service, so weigh his view accordingly.
In plain terms, the routine tickets are closing themselves, a chatbot is writing the automations, and the AI practice can arrive with someone else’s engineers behind your logo.
Which raises the question this whole shift turns on: where does a technician learn the job now?
Renting Skips the Apprenticeship
Expertise in an MSP has always been built the same way: by doing the work, inside a pyramid. Juniors close the routine tickets, learn the client’s environment, and move up into the harder work.
Forrester’s Ted Schadler, looking at the big public service providers, writes that the pyramid is exactly what AI disrupts. Large teams with pyramid delivery structures are the ones disrupted, because AI absorbs the coordination and compresses the execution work at the base. Forrester’s model puts technology services near the top of more than 200 markets for AI disruption. The big providers’ response is to put their own engineers next to the client. Schadler notes service providers are scaling forward-deployed engineering teams, and now the technology and model makers are bringing their own.
At MSP scale, the top of the pyramid is being offered as a product. In Channelholic, Rich Freeman framed the problem behind Pax8’s 8bot, now in early access. To make real money on AI, an MSP needs two skills: diagnosing what a client’s business needs, and building something that meets it. Both take time to develop. Pax8’s answer is to skip that. Chief product officer Elizabeth McIlhany told him the goal is for partners to “deliver AI value without building a custom AI solution.” There’s no service engagement. The client accepts an invitation and logs in. She says the agent keeps building context and memory until it learns the client’s business.
Notice where that learning lands. It accumulates in the agent, on the vendor’s platform. It doesn’t accumulate in your technician.
I put the same point to Brian Kelly. His company owns the software, the servers and the engineering; the MSP owns the relationship and a margin. So what does the MSP own at the end of a three-year contract? His hope, he said, is “a customer that wants to just continue on.” The system and the data belong to the client.
That’s an honest answer. It’s also the mechanism. The skill used to be a byproduct of the work. When the routine work closes itself and the hard work is rented, nothing passes through the middle for anyone to learn from.
In plain terms, AI didn’t just take tasks out of the shop. It took the apprenticeship.
And nobody has written the job that replaces it.
In a small shop, that gap has a very specific shape.
The Missing Middle Job
For the MSP, it shows up as a missing job.
Start with CompTIA, which sells IT certifications, so it has a stake in this. It announced plans for a credential called AgentOps+, covering how to deploy, secure, monitor and govern AI agents. Its own announcement says the skills to manage agents responsibly remain undefined. The certification isn’t due until the second quarter of 2027. Meanwhile, CompTIA’s survey of more than 3,600 IT professionals, including its own certification alumni, found fewer than half satisfied with their chances to advance. At small and mid-size businesses, a third said their employer offers no ongoing learning support, or they didn’t know what was available. That’s the person who would fill the middle role. They’re looking for a path, and in a small shop nobody has drawn one.
Big companies are drawing theirs. KPMG surveyed 400 CEOs of U.S. companies with at least $500 million in revenue. Fast Company reports that 86 percent are redesigning roles, career paths and team structures around people working with AI. Three-quarters have reassessed the skills needed for entry-level jobs. And 89 percent expect headcount to grow over the next three years.
Put those together. Large companies are rewriting the entry-level job on purpose. In an MSP, the entry-level job is being rewritten by default, by whatever the automation doesn’t close. Nobody decided it. And a twenty-person shop has nobody whose job it is to decide, unless the owner does. It’s not just your next hire, either. A lot of the people who own MSPs today started on a help desk. That queue is where this industry’s next owners come from.
So here’s the choice. Keep one path inside your shop that runs from tier-one work to AI operations. Deliberately leave some work your automation could close with your juniors, so they still learn on it, and accept the efficiency cost. When you rent an AI practice, write knowledge transfer into the contract, so your people work alongside the vendor’s engineers instead of just reselling them. Or rent both ends, let the tools take the rest, and own the relationship and nothing underneath it.
That choice gets made one ticket queue at a time.
Because the path from tier one to AI operations only exists if someone puts it on the schedule. Pick the ticket categories your juniors keep even when automation could close them, and write down which skill each one teaches. Then review that list every quarter, on the same day you decide which automations to turn on.
What to Consider
- Make approval mode the apprenticeship. Before you let automation close a ticket category on its own, decide whether juniors learn on that category. If they do, run it with human approval and show the automation’s suggested fix, so your technician learns the problem and checks the tool at the same time. That’s the setting half of IgnitHQ’s partners are already using. The difference is choosing it on purpose for the work that teaches.
- Put knowledge transfer in the rental contract. If you white-label an AI practice, ask for named sessions where your technician works alongside the vendor’s engineer on your own clients, plus access to the configurations and documentation they build. The system and data may belong to the client at exit, but the know-how shouldn’t be the one thing that walks out with the vendor.
- Write the middle job before you need to hire for it. One page is enough: which agents this person monitors, which permissions they set, and which vendor work they check before it reaches a client. If you can’t fill in that page, you’ve found the gap the rented practice is covering for you. Thrive’s chief operating officer, Scott Steele, told Contact Center Pipeline that when AI took over intake, the company moved 32 engineers out of Level 1 intake roles and into resolution work. That’s the middle job, written on purpose.
If this trend continues: By 2029, an MSP that needs someone to run its AI operations will choose between hiring that person from a vendor and promoting someone it kept a path for, and only the second will already know its clients.