Most companies that buy AI agents pay for access: a license per seat, per device or per month, whether the agent gets the work done or not. If the agent handles half the requests and sends the rest back to your team, you still pay the full price, and someone on your side still has to watch it.
A launch this month shows a different way to buy an agent, and it says a lot about what a good agent offer should include.
What launched
On November 4, Fixify made its help desk agents generally available to managed service providers, the companies that run IT for small and mid-sized businesses. The launch came during IT Nation Connect, a major conference for managed service providers, the same week Lemhi opened its managed AI platform to all providers.
Here is what Fixify is selling, according to its announcement:
- Agents that finish the ticket. They handle routine Level 1 requests from first contact to resolution: password and multi-factor resets, account lockouts, application access, group membership changes, and onboarding and offboarding.
- Inside the tools the provider already uses. The agents work in the ticketing systems providers already run, including ConnectWise, Autotask, HaloPSA and NinjaOne.
- Named people supervising. Fixify’s own IT analysts watch the agents around the clock, handle exceptions, and step in whenever a decision or an approval needs a person.
- Built and maintained for each client. Fixify writes and updates the playbooks the agents follow, so the provider doesn’t need senior technicians to build or manage them.
- Paid per resolved ticket. Providers pay for tickets the agents successfully resolve, instead of a license fee per device, technician or client.
Fixify hasn’t published resolution rates, volumes or customer results, so this is a description of the offer and nothing more. It’s the shape of the offer that matters here.
Customer service got there first
In 2024, Zendesk moved its AI agents to outcome-based pricing, where customers pay only for issues the AI resolves on its own and can keep human agents for everything else. What’s new this month is the full package arriving in IT services: the agent, the people supervising it, the playbooks and the per-resolution price, sold together as one contract.
Why the model is worth paying attention to
Per-resolution pricing changes who carries the risk. If the agent fails, the vendor doesn’t get paid for that ticket. That gives the vendor a reason to keep the agent working well, and it gives the buyer a simple way to see whether it is.
The supervision piece matters just as much. An agent that resets passwords and changes access is making decisions about security. Having named people watching it, with a clear rule for what gets handed to a person, is what makes it safe to let the agent finish the work.
The same three things apply to any agent you’re considering, whether it handles IT tickets, invoices, purchase orders or customer emails:
- Who supervises it? A named person or team, with time set aside to do it.
- What gets handed to a person? A written list of exceptions and approvals the agent never decides on its own.
- What are you paying for? Access to the tool, or work the agent actually finished.
The fine print to check
Paying per resolution only works if everyone agrees on what a resolution is. Before signing any outcome-based agent contract, get clear answers to these:
- What counts as resolved? The agent closing the ticket, or the requester confirming the problem is gone?
- What happens when a ticket reopens? Is it billed twice, refunded or free?
- Who decides when the agent hands off? And does a ticket a person had to finish still get billed?
- Can you see every action the agent took? You’ll want a record for every ticket, especially the ones involving access and security.
- What does the agent do when it’s unsure? The safe answer is that it stops and asks a person.
How we answer the same three questions
Buyers should ask us these questions too, so here are our answers.
Who supervises it? A named Metacto lead runs every agent that’s live, and someone on the client’s team is accountable for its results.
What gets handed to a person? Every consequential step, until the agent earns more autonomy on a written rule: 20 decisions, 90% agreement with the reviewer, and no rejections in 14 days.
What are you paying for? Work you can measure. The first agent is measured against a baseline set in the first month and proven within eight weeks. AI usage is billed at cost on the client’s own model accounts, with no markup, and the agents run in the client’s cloud and keep running if the partnership ends.
The help desk launch is a good sign for buyers. It shows that agent vendors can be asked to stand behind the work, with people watching it and a price tied to results. That’s a fair standard to hold every agent offer to, ours included.
If you have a high-volume process where requests pile up waiting for a person, show us the work. Describe it in about three minutes and we’ll send back a plan for your first agent: what it would do, what we’d measure, and how it goes live. If you’d rather talk it through, talk to an expert.
Sources
- Fixify, Fixify Launches Agentic AI Automation for MSP Help Desks, November 4, 2026.
- Channel Insider, Lemhi Makes Managed AI Platform Available to all MSPs, November 3, 2026.
- Zendesk, Outcome-based pricing for AI agents, August 28, 2024.


