AI pre-filters support tickets — how we bring context back into tickets

Support enquiries without context you probably know from your own organisation. A sentence lands in the inbox: "The page doesn't work." Or: "I need admin rights." No project, no URL, no time, no error pattern. What follows is email ping-pong: ask, wait, ask back, interpret. By the time someone is actually working on the solution, half a day has gone.

We asked ourselves how we could break this pattern without imposing yet another rigid form on our customers. The answer for us: AI-supported intake. It isn't spectacular, it's pragmatic — and that's exactly why it works.

What happens before a human sees the ticket

When an enquiry comes in to us, it runs through a structured pre-process. A specialised AI agent analyses the text, matches it against the project context, and adds missing information as far as it can be determined automatically:

• Which system this is about — from the sender domain, project-specific keywords, and previous tickets.
• Were there suspicious monitoring alerts, deployments, or error spikes in this system in recent hours?
• Which role is speaking here — editorial, IT, business — and which typical concerns have been associated with this role in the past?
• What urgency realistically follows from the content (not from the perceived tone)?

The result is an enriched ticket draft our team can grasp in seconds. Instead of "I need admin rights", the responsible colleague reads, for example: "Editorial staff member, project XY, currently author role, refers to campaign on Thursday, possible link to new media folder."

What the AI agent doesn't do

It's important to us to name clearly where the line sits. The AI agent doesn't decide. It doesn't answer tickets either. It doesn't assess permissions. All of that stays with humans. Its only job is to establish the context needed for a good decision — the context that would otherwise be added across multiple emails.

That's a deliberate design choice. We believe AI in support adds value above all when it reduces friction without taking on responsibility. Your data, your access, your operating processes stay in the hands of our colleagues, who know your project.

Why you'll feel it

You don't notice this pre-processing directly. You notice the consequences:

• Faster initial assessment. Because the context is there, the first qualified response comes earlier — often without coming back to you.
• Fewer "please add…" emails. The information we can find out without you, we no longer have to ask you for.
• Better prioritisation. What's urgent gets treated as urgent. What can wait isn't artificially pulled forward just because the wording was nervous.
• Consistent documentation. Every ticket contains a clean context block. That's also valuable for you when you want to understand in six months why a particular decision was made.

How this plays with the rest of our methodology

The fact that we can build something like this at all comes down to our foundation. Our 3-layer model ensures that projects are built comparably. Because of that, an AI agent can reliably attribute context — it finds the same log structures, the same configuration keys, the same role patterns again. On a wildly grown zoo of individual systems, the same pre-processing would be a permanent source of errors.

This connection is central. Standardisation is the precondition for sustainable automation. Without standard, even the best AI only helps in patches. With standard, it becomes a quiet lever across many systems.

One building block of our AI Agent as a Service

The support intake is just one example from our AI Agent as a Service offering. For customers we build comparable agents for other tasks: pre-processing incoming applications, automatic enrichment of content elements, suggestions for metadata and SEO snippets, or structured intake of PDF inputs into internal systems. Always to the same pattern: the agent provides structure and context, the human decides.