Yorkshire Tech Bills £18,000 for Work Mostly Done by AI

John Benson has spent around 30 years in IT and runs Yorkshire Tech, a managed service provider based in Harrogate with a few hundred endpoints under management. He came to the call with two numbers rather than a feature wishlist: an £18,000 client invoice for work that was mostly done by AI, and a data centre move that went from two weeks of planning to about an hour of work.

What makes his story useful to other MSP owners is how ordinary the setup is. Mingo, OpenFrame's technician copilot, sits alongside ChatGPT and Claude in his working day. Mingo handles the endpoints, the other two handle the writing and the coordination, and John runs the whole thing from an iPhone and the OpenFrame mobile app. He no longer carries a laptop. Fifty-six of his devices are on OpenFrame today, with the rest of the fleet still on the tools he already pays for.

He is also clear about what OpenFrame needs before he moves everything over: chat folders and sorting, a one-glance patch and device dashboard, admin and user permission levels, and a Linux agent. He calls those polish. The rest, in his words, is a total game changer.

Yorkshire Tech runs a conventional MSP stack: NinjaOne for Windows endpoints, Addigy for Macs and Halo for the service desk. Each tool does its job. None of them investigates a problem, writes anything down or hands a finished piece of work to a client. Every incident still meant a senior engineer assembling the picture by hand, and John has used almost every RMM system on the market looking for something that does more than show a dashboard.

Two jobs in 2026 showed what that costs. A client offboarding turned into an incident when the client's Microsoft 365 tenant was compromised and its terminal servers came under attack. Investigating every endpoint, collecting evidence and writing it up as a report the client could act on is days of skilled time. The second job was moving servers out of Yorkshire Tech's York data centre into a client office. The usual approach is two weeks of planning: pulling backup configurations, stripping out security data, preparing rollback plans and briefing the engineer who will do the physical move.

Underneath both jobs sits a business decision. John wants to spend more of his time on consulting and less on routine delivery, and he wants to run the company without being tied to a desk. That only works if the tooling can take on real work, not just alerting, and if it can do so on a fleet that is still split across several vendors.

Yorkshire Tech joined OpenFrame early access in June 2026 and became a paying customer in September 2026. The tenant has 56 registered devices. John did not rip out NinjaOne, Addigy or Halo. He added OpenFrame alongside them and let Mingo take the jobs the other tools could not do.

On the Microsoft 365 incident, Mingo did the endpoint investigation: checking machines, pulling what it found and keeping the evidence in one place. Mingo cannot produce Word documents yet, so John used desktop ChatGPT to coordinate the output, and he is finishing an 18-page incident report for the client from that material. On the data centre move he used Mingo, ChatGPT and Claude together to pull the backup configurations, strip the security data out and hand a clean brief to a returning engineer. He is precise that this was a combination of tools, with Mingo on the endpoints and the other two on the writing, and that is how the case study presents it.

Day to day, he trusts Mingo to patch. He checks in from the iPhone through the OpenFrame mobile app, which is why the laptop stays at home. His plan is to keep OpenFrame as an addition to the current stack until PSA, quoting and billing exist in the platform, and to roll it out to the rest of the team once admin and regular user permission levels land, since today every user is a super user and his confidential chats are visible to staff.

The headline number is the invoice. For the Microsoft 365 incident and offboarding, Yorkshire Tech billed the client £18,000. In John's words: "I've got an invoice here for £18,000 for work we've done. Most of that was done by AI, not by us." The client got the endpoint investigation and an 18-page incident report, and most of the hours behind that invoice were AI hours rather than engineer hours.

The data centre move is the second result. The planning that used to take two weeks took about an hour: "In the past it would have been two weeks worth of planning and rollback." The returning engineer walked in with a clean brief instead of a pile of configs to decode. Between those two jobs and the daily patching he now leaves to Mingo, the way John works has changed enough that he describes his job as running AI agents and has dropped the laptop: "This has just changed how I work now. I don't carry a laptop."

What he still wants is a short list, and he frames it as polish: folders, tags and recency sorting for chats; a simple AI-populated dashboard showing online and offline devices, patch status and what still needs doing; admin and user permission levels; a Linux agent for his AWS virtual machines; and file uploads into Mingo. None of it has slowed him down. "None of that bothers me because what we use it for, it's fantastic." Asked how it compares with everything else he has run in 30 years: "There's nothing else like OpenFrame, nothing at all."