Departments · IT AI in IT: Support, Knowledge Assistant and Rules for AI Use

AI in IT: five tasks at a glance

These are typical workflows, not client projects. In each one the AI prepares, and a person from IT checks and decides.

1. Triage tickets and solve standard requests

A large part of the requests to IT repeat: password, access, printer, software.

• Input: the ticket or the email to IT.
• AI prepares: recognises type and urgency, assigns the ticket to the right group, searches the knowledge base and earlier tickets for the solution and writes a reply or instructions.
• IT checks: replies for anything that concerns rights, accounts or data. For selected standard cases the reply can go out directly after a trial phase.
• Output: a triaged ticket with a suggested solution or a sent standard reply with a log.

2. Create users and document permissions

New colleagues need accounts, groups and access, and every change has to be traceable.

• Input: the notice from HR or the department and your role templates.
• AI prepares: assembles the accounts and groups needed from the role template and writes the change request with a reason.
• IT checks and approves: every change of rights. The AI changes no rights itself.
• Output: a change request and, after approval, the documentation of the rights granted.

3. Knowledge assistant for manuals and systems

Knowledge about systems sits in manuals, wikis and heads, and new staff search for a long time.

• Input: manuals, wiki pages, operating instructions and earlier tickets.
• AI prepares: answers questions from these sources and names the passage. If it finds nothing, it says so.
• IT maintains: the sources. The assistant reports outdated pages but does not change them.
• Output: answers with a source, plus a list of questions without an answer from which the next documentation comes.

4. Summarise monitoring messages

Monitoring systems report a lot, and what matters is lost in the noise.

• Input: messages from monitoring, backup and security systems.
• AI prepares: groups messages into incidents, recognises repetitions and suggests which to look at first.
• IT checks: the summary and acts. The AI does not intervene in systems.
• Output: a daily summary with the messages behind it.

5. Interfaces between programs

Many small workflows between programs run by hand because a dedicated interface never paid off.

• Input: a recurring workflow between two or more programs, such as form, spreadsheet and ticket system.
• AI prepares: we build the workflow as an automation, for example with n8n or Make, and use AI where text is read or assigned.
• IT checks: the workflow before the start and the messages in operation.
• Output: a documented workflow your IT understands and can change. More in the article on hosting n8n yourself.

Shadow AI and AI governance: rules instead of bans

Shadow AI is the use of AI tools for company work that IT does not know about: private accounts, free plans, browser extensions. The data leaves the company in the process. A ban rarely changes that, because the tools are useful. An offer that people prefer works better.

• Ask first, then measure: Who uses which tools for what? An open survey shows more than network data. Evaluating network or log data concerns employee data, and data protection and the works council belong in beforehand.
• Approved tools: one or two tools with a contract, administration and processing in the EU that cover the most common need. Which ones fit is in the articles on Copilot or ChatGPT and on introducing ChatGPT in the company.
• AI policy: one page with clear rules on what is allowed and what is not. The contents are described on the page on the AI policy.
• Data classes: What never belongs in a chat window (personnel data, health data, customer lists, trade secrets), what needs care and what is free.
• Logging with a sense of proportion: What is logged and who sees it is agreed beforehand with data protection and the works council.
• Duties under the AI Act: Anyone who uses AI professionally has duties as a deployer. The pages on the AI Act and on the duties of deployers give an overview.

Programs and access: connected to what you already use

IT knows its systems best. The table lists the usual cases. Which route your system offers, we clarify before the quote.

The names are for orientation and do not mean a partnership. The principle for access: the AI never has more rights than the person it works for. Where a service account is needed, it gets read-only rights unless a person approves a change.

Connect, do not replace

You do not need a new ticket system or a new directory service, and we do not sell either. The AI sits in front of your systems and works with their interfaces.

• Your systems stay in charge. Tickets, accounts and rights live there, and changes go through your existing routes.
• Your IT providers work as before. We do not run IT for you and do not replace support, we complement it.
• One workflow at a time. We start with one step, for example triaging tickets, and run it alongside the existing work first.
• The way back stays open. Everything the AI does can be traced in your systems. Without it your IT carries on as before.

Approvals and rights

In IT, rights and accounts are sensitive. So before the start we agree what the AI may do.

• Rights: The AI acts with the rights of the person it works for, never with more.
• Changes to accounts and systems: only as a request that a person checks and approves, under the four-eyes principle.
• Standard replies: only for topics you release, for example instructions. Anything that concerns rights or data goes to IT.
• Credentials: Passwords and keys stay in your vault and are not copied into prompts or documents.
• Log: Every action is recorded with time, trigger and approval.

The table lists the main sources for orientation. Which rules apply to your company depends on sector, size and country.

This overview is general information, not legal advice. Your legal adviser or data protection officer can confirm what applies to you.

Data, knowledge and operation

A knowledge assistant is only as good as the sources it can access.

• Your own data with a source: Manuals, wiki pages and earlier tickets are searched and quoted with the passage. How this works technically is described in the article on AI with your own data.
• Cloud in the EU or in your own house: Depending on the data class, the AI runs with a provider processing in the EU or on your own hardware. We tell you when which is worth it.
• Currency: Outdated manuals lead to outdated answers. The assistant reports which sources contradict each other or have not been updated for a long time.
• Operation and monitoring: We set up cost and error monitoring and document what operation looks like so that your IT can take it over.
• Data protection: Tickets contain names and system data. We process them on servers in the EU, show the AI only what it needs, and agree retention and deletion with you. More under AI and data protection.

Your own tools for IT

Some things are missing not in the ticket system but in front of it or next to it. We build them as separate programs when standard products do not fit:

• Self-service portal for standard requests: request access, software and devices, with an approval route and a log.
• Internal interfaces: the data exchange between your programs that someone used to do by hand.
• Documentation store: one place where instructions and operating documentation are maintained and the knowledge assistant finds them.

These programs complement your systems, they do not replace them. More under software development.

Where to start

These three questions help with the assessment:

1. Which three requests reach your IT most often every week?
2. Which AI tools do your employees already use, and is there a rule for them?
3. Where does knowledge about your systems sit: in manuals, a wiki or in heads?

If you send us the answers, we tell you which workflow is worth doing first and what has to stay with your IT. We begin with a trial run alongside the existing work.