Connecting TOPdesk to AI: a ticket that belongs to the right caller straight away
TOPdesk can be connected to AI. The assistant first looks the caller up in the person register, then logs an incident with a factual description and the right classification, finds existing incidents back by status or operator and updates an incident. Because the caller can follow their own incident, every change is also a message to that person. Updating therefore happens on request and never by itself.
What the assistant does in TOPdesk
Look up the caller
Before anything is recorded, the assistant finds the person in the TOPdesk person register. That prevents incidents hanging on the wrong name or on nobody at all, and makes sure the feedback lands in the right place.
Log an incident
The assistant records a new incident with a factual description and the classification you use: category, subcategory, call type, impact, urgency and priority. The description names what is not working, since when and who is affected.
Find incidents back
Existing incidents can be looked up by status, operator and period. That is the quick route to the question of what is still open for a department, without anyone having to walk a list.
Update an incident
The assistant can change the status, extend the description or add a progress line. That is exactly what becomes visible in self service, so it happens on request and with the text shown beforehand.
Account for what the caller sees
An incident is not an internal notepad. The caller follows their own incident and reads whatever is added. The assistant therefore writes in the tone you would use with the caller looking over your shoulder.
What this looks like at a service desk
Government
- Situation
- At the desk an employee reports that the printer on the second floor has been down for two days.
- What the agent does
- The assistant finds the employee in the person register, records the incident with the right classification and notes since when it has been going on and how many people it affects.
- Result
- A complete incident without anyone having to fill in fields between two conversations.
Healthcare
- Situation
- A ward team leader wants to know which of her team's incidents are still open.
- What the agent does
- The assistant looks up that department's incidents by status and operator and adds what the latest progress was.
- Result
- One overview instead of a round of phone calls, and it is clear what the service desk is waiting for.
Education
- Situation
- At the start of the academic year the same questions about access and accounts arrive in dozens of variants.
- What the agent does
- The assistant records them under the same category and subcategory and refers in the description to the earlier incident where one exists.
- Result
- Incidents that belong together are stored together, so the pattern becomes visible instead of disappearing.
Facility management
- Situation
- A failure affects an entire wing and separate incidents come in about the same problem.
- What the agent does
- The assistant looks up the existing incidents on that subject and proposes linking the new one to the same line, with the right impact and urgency.
- Result
- The scale of the failure is visible at once instead of scattered across separate incidents.
Setting it up inside your TOPdesk
-
1
Connecting through the platform
An administrator records access to your TOPdesk environment in IntraGPT once. On the TOPdesk side that access can be withdrawn again at any moment.
-
2
Setting permissions
You decide whether the assistant may only look up or also record and update, and which roles may use it. Many organisations start with looking up and logging, and add updating later.
-
3
Using the assistant
After that you simply say what is going on, or ask what is still open. The assistant finds the caller, drafts the incident and shows what it is about to record.
-
4
Visible in the log
What the assistant looked up, logged or changed is recorded in the log, with the moment and the question that preceded it. That helps during an audit and with a question afterwards.
Incidents are about people, not only about equipment
The service desk of a municipality, a care organisation or a school holds names, locations and sometimes sensitive situations. What you do with that has to be explainable.
Where the data lives: Nederland of de EU
The caller reads along
Whatever the assistant puts in an incident, the caller can see back in self service. That is why it shows the text before recording or updating anything, and never does it unprompted.
Roles and departments
You decide per role who may use the assistant. A service desk employee has a different need than someone who only wants to follow their own incidents.
Accountable afterwards
Every retrieval and every change lands in the log. For organisations that have to account for how they handle personal data, that is not a side issue.
Processing on your own server
Incident contents are processed on our own server in the Netherlands with open-weight models. That content does not go to an external model provider and is not used to train models.
Questions about TOPdesk and AI
Can I connect TOPdesk to ChatGPT?
For a municipality, a care organisation or a school that is not a workable route. You would give a public service access to a system holding names, locations and incidents about people, without roles and without a log. IntraGPT connects your own TOPdesk environment, with permissions you set, and processes the contents on our own server.
Does IntraGPT work with TOPdesk?
Yes. TOPdesk is in the connector catalogue and in use at customers. The assistant can look up persons, log incidents, find incidents back and update incidents. Which of those possibilities are switched on is up to you.
Why does the assistant look up the caller first?
Because an incident without the right caller is worth little. The feedback then reaches nobody, the incident does not appear in that person's self service and departmental reporting does not add up. The assistant therefore searches the person register first and uses the person it finds, instead of typing a name over.
Does the caller see what the assistant does?
Yes, as far as you have set self service up that way. An incident and its progress are visible to the caller. That is a reason to have the assistant show the text first, and to agree who may have an incident updated or closed.
Which classification does the assistant use for a new incident?
The same one you use yourselves: category, subcategory, call type, impact, urgency and priority. In many organisations that classification drives routing and the agreements on turnaround time, so we settle beforehand which values the assistant may choose and when it had better ask.
Does this work for changes and other TOPdesk modules too?
This connector covers incidents and persons. Other modules are not included, and we do not write them up as a possibility either. If you use more of TOPdesk and want something there, that is a separate conversation.
Does this meet our requirements around personal data?
The content is processed on our own server with open-weight models and does not go to an external model provider. Nothing is trained on it, and every retrieval is recorded in the log. IntraGPT is used by organisations that test against ISO 27001; what exactly has to be in your processing agreement we go through together.
What does a TOPdesk connector cost?
Connecting is part of the project we run together. What drives the work is whether the assistant may only look up or also record and update, how elaborate your classification is and which roles may use it.
How long does setup take?
The connector itself stands quickly and can be tested the same day. The lead time usually sits in your own classification: which categories and call types the assistant may choose, and which agreements apply to updating a running incident.