Your AI IT engineer for your MSP
Autonomous AI agents that resolve tickets end to end.
Claudia · AI IT engineer
Acme Corp support portal
Claudia: Hi Priya, I’m Claudia, Acme Corp’s AI IT engineer. What’s going wrong? Pick one, or describe it in your own words.
Interactive demo: you’re the user, then the technician. Names and data are illustrative; the fix is simulated.
The fleet, from the operating model
- 20
- AI agent roles
- 77
- Agent skills
- 157
- Workflows
- 15,000+
- Tools across 70 connectors
The problem
Where your desk loses time
The ticket from the demo, worked the way most desks work it today, and the same ticket with Claudia.
- 4
- people touch it today
- 1
- person touches it with Claudia
Counted from this example, not measured across desks.
Priya emails the desk
“Email’s not working.” One line, no detail.
- Waits in the queue
The dispatcher reads it and assigns it
Every ticket crosses your desk before anyone can work it.
Dispatcher or team lead →- Waits for a technician
A technician calls Priya back for the details
Half of every ticket is working out what it actually says.
Service desk technician →Escalated to the Microsoft 365 admin
Who finds yesterday’s policy change and fixes it.
- Notes written later, if at all
The client mentions it at the next review
You learn a client is unhappy at the renewal call.
Account manager →
What Claudia reads
Context first, questions second
Before it asks your user anything, Claudia reads the systems your desk already runs.
Pick a source.See how it connects
Gathering context · Autotask
Reads ticket history, configurations and alerts.
Autotask: Device LAPTOP-PS14 · 2 similar tickets this month, both fixed in Microsoft 365For “I can’t send or receive email”- Autotask: One lockout ticket for Priya last quarter, fixed by re-signing a phoneFor “My account keeps locking me out”
- Autotask: In warranty until March 2027 · no hardware ticketsFor “My computer is really slow”
- Autotask: No open printer tickets for AcmeFor “The printer isn’t printing”
Anthropic · Claude
OpenAI · GPT
Guardrails
Every fix waits for your technician
That is how the supervised launch works. Pick an action to see what Claudia checks first, and what it refuses.
Ask Claudia to…
Checking permissions
- Claudia can unlock accounts in Microsoft Entra
- Unlocking a user is inside Acme’s approved scope
- J. Okafor can approve account changes for Acme
Technician approval
Unlock Priya Shah’s account in Microsoft Entra
Requested of J. Okafor · Microsoft 365 admin
You’re J. Okafor now. Nothing changes until you decide.
Illustrative: the checks for Acme Corp in the demo above.
Your data
- Your data never trains a model
- Ticket content, documentation, and client data are never used to train or fine-tune any model — ours or a provider's.
- Credentials are never sent
- Passwords, API keys, and secrets are blocked or masked before anything leaves your environment. Agents never see them.
- Your permissions still apply
- Agents inherit the access rules already configured in your PSA. They cannot see what a technician could not.
- Graduated autonomy, with a human in the loop
- Claudia starts supervised on every MSP. Your team sets the level she may act at and raises it over time, on evidence, with the tools we provide.
Claudia is invite-only for a limited supervised launch.
How it works
Five agent roles, one ticket, one approval
They ask the right questions, get your technician’s approval, fix the problem, document it and close the ticket.
01
Ticket opened
Concierge
02
Gathering context
Triage
03
Questions
Concierge
04
Diagnosis
Resolution
05
Checking permissions
Dispatch
06
Technician approval
Dispatch
+ your technician
07
Running the fix
Resolution
08
Verify
Resolution
09
Documented
Resolution
10
Ticket closed
Watchdog
- Concierge
- Asks the user for what’s missing, before a technician has to.
- Triage
- Gathers context, classifies the ticket and sets priority against the client’s SLA.
- Dispatch
- Checks permissions and routes the approval to the right technician.
- Resolution
- Diagnoses, runs the approved fix, verifies it and documents it.
- Watchdog
- Watches the SLA and follows up with the client.
The roles come from The Agentic MSP Operating Model, a free, public model of how an MSP runs. Claudia is one AI engineer working in all five.
Read the model, version 4.361.0The Agentic MSP Operating Model — © 2026 Emil Isanov / AI Autopilot (https://aiautopilot.ai), licensed CC BY 4.0.
Built inside an MSP
Built inside ETech 7, the MSP Emil Isanov founded in 2012 — and run there first.
- 2012
- ETech 7 founded
- 600
- ticket types in the model, written by people who ran the desk
The list · 26 pages · PDF
The 10 ticket types that cost an MSP the most
The ten ticket types that cost your desk the most, with the lever for each.
Questions
What MSPs ask first
The launch, who stays in control, and where your data goes.
Getting started
Which stacks are the first launch focused on?
Starting with Autotask + Datto RMM and ConnectWise PSA + ConnectWise Automate. Claudia is invite-only for a limited supervised launch. Claudia’s AI agents resolve tickets end to end: they ask the right questions, get your technician’s approval, fix the problem, document it and close the ticket. Pilot scope and prerequisites are confirmed with your team.
Can the agents act without a human approving?
The current limited launch is supervised. The broader product uses graduated autonomy, but unattended execution is not part of this launch offer. Any later change in autonomy needs a separately agreed scope and evidence.
Do we have to migrate off our PSA?
The initial launch is designed around your existing Autotask + Datto RMM or ConnectWise PSA + ConnectWise Automate pairing. Confirm tenant access, permissions and the technician's workflow during scoping; catalogue coverage alone does not establish that every setup is ready.
Who does the setup work?
The published offer includes a two-hour onboarding session at no charge. Our team reviews the agreed workflow, integration prerequisites and technician walkthrough with you. The session is not a guarantee that every tenant or broader catalogue workflow will be ready in two hours.
Your data
Which AI providers do you use?
Anthropic Claude and OpenAI, hosted on Microsoft Azure within the United States. Both are configured so your data is not retained after processing and is never used for training.
What happens to our data if we leave?
Customer data is deleted within 30 days of contract termination, unless a longer period is required by law or you request earlier deletion.
Are you SOC 2 certified?
Not yet — our SOC 2 Type II audit is underway, with the report expected Q1 2027. In the meantime our 73 security controls are continuously monitored by Vanta and published live in our Trust Center, covering infrastructure, internal procedures, product, organisational, and data privacy controls. Ask us for the report once it is issued.
From the blog
Latest writing
Automation · August 17, 2026
Ticket triage automation for MSPs: the complete guide
10 min read
Automation · August 17, 2026
AI ticket dispatch: skill, load and shift instead of whoever is free
9 min read
Business · August 17, 2026
AI in your PSA: integration realities and the build-vs-buy question
11 min read
Start with a supervised workflow
Want to take Claudia for a spin in your IT environment? Claudia is invite-only for a limited supervised launch.