Published on 17 November 2026
The first level, and what AI takes on
The first level of IT support is listening, sorting and simple steps: understanding what isn't working, gauging the urgency, having the person try what can be done without special rights, and passing on the rest with what was learned. It is the most repetitive part of the job, and the part AI handles well, at any hour, in writing.
Second and third level stay with your technician, in-house or external: administrator rights, work on the computer, the network, security. The AI gives them better-described tickets, and time for those.
Ticketing, helpdesk, service desk: the terms
Ticketing software turns each request into a ticket: a number, a category, a priority, an owner, a history, a status until resolution. The helpdesk is the desk that receives these requests and answers them; the service desk is its broader version, which also handles service requests (an account, an access, a new computer) and tracks response-time commitments.
Level 1 means taking the request and the simple steps, level 2 the technical intervention, and level 3 the expert or the vendor. GLPI and Zammad, two widely used open-source tools, hold the ticket; the AI works upstream, in the conversation.
The IT agent at work
A team member writes to the assistant: "Outlook hasn't received anything since this morning." The assistant hands the question to the IT agent, the support AI agent. It opens the incident from the first message and sorts it: the category (email and collaboration, out of nine), the urgency she states, the impact it estimates, and the resulting priority, from P1 to P4.
It then follows the diagnostic tree for that problem, shipped with the product, and suggests one step per message: look at the status at the bottom of the window, check offline mode, restart Outlook, check the account status. It notes each step and its result. When it is solved, it confirms with her.
The trees provided, and the steps that belong to the technician
As of 17 November 2026, six diagnostic trees ship with the product, written from the vendors' documentation and dated: the printer, the slow computer, the password and two-factor authentication, OneDrive that no longer syncs, Outlook that no longer receives mail, the VPN. Each tree says what to isolate first, which steps to suggest, what to note for the file, and when to hand over straight away. More will follow.
The split of roles is written in the code, not just in the instructions:
| The IT agent guides | It hands over straight away | These steps belong to the technician |
|---|---|---|
| One to three simple steps, up to five when the solution is close: check a status, test on the web, restart an application, switch an option. | A security risk (phishing, virus, compromised account). | Recreating an Outlook profile, deleting a mail data file. |
| One step per message, noted with its result. | A possible loss of data. | Touching the registry, resetting the computer. |
| Three hypotheses with no result, and it hands over. | A need for administrator rights. | Disabling the antivirus, the firewall or two-factor authentication. |
| The person carries out the steps, it guides them. | Several people affected, or the person asking for someone. | Asking for a password or a code, installing downloaded software. |
The ticket goes into your tool
When it hands over, the IT agent passes the conversation to a person on the team with a structured file: the symptom, the category, the priority, each step tried and its result. The technician picks up where the conversation stopped, without asking the questions again.
If you have GLPI or Zammad, the incident connects to it: it goes there with the file as an internal note and its priority mapped to your tool's scale. The connection is set up in the agent's page, with your tool's key. Without a ticketing tool, the file arrives in the team's escalation queue.
What gets solved becomes an article
When the person confirms it is solved, the IT agent writes a draft article: the symptoms, the steps that solved the problem, a summary. The team reviews and approves it, and it joins the knowledge base. The second time, the answer is there, with its source.
For business software, the IT agent also answers from its documentation, and writes to the vendor's support when that is not enough: that is the vendor relay.
The questions we get asked
- Do you need to change your ticketing software?
- No. Your ticketing software keeps tracking the tickets; the AI works upstream, in the conversation, and the incident connects to GLPI or Zammad with its file. Without a ticketing tool, the files arrive in the team's escalation queue.
- Can the AI take control of a computer?
- The IT agent works in writing: it guides, and the person carries out the steps themselves. Actions that need rights or carry a risk (the registry, the antivirus, resetting the computer) belong to the technician, and the code removes them from every answer.
- What happens when a whole team is affected?
- Several people affected at the same time means immediate escalation: the IT agent suggests no step and hands over straight away to a person on the team.
- Who approves the articles written after a resolution?
- A person on your team. The IT agent writes a draft when the person confirms the problem is solved; it only enters the knowledge base once approved.
- Is it useful to an external IT provider?
- Yes: the provider receives tickets already sorted, with the steps tried and their results, and spends its time on interventions. It has a place in the platform, through a dedicated profile.
To see it answer
The demo of a manufacturing company opens with your personal link, received by email. Ask the assistant an IT question, a printer that isn't responding or a forgotten password, and watch it go to the IT agent, with its sources. Go-live then sets up its trees and the connection to your ticketing tool. Try the demo, then, once you have made up your mind, the purchase page states the price with nothing hidden.


















