Ask Us Anything: Knowledge Management, with Stefan and Camilo

Hi ​@NickB,

Thank you for these questions, we are starting of strong 💪

The answer fro Q1

There's no mandatory format from TOPdesk right now, but based on how our own AI features actually use knowledge items, and on how AI-driven knowledge bases perform best across the wider industry, we can give some grounded advice.

Why this matters

The AI Answer assistant, which suggests a draft answer to operators based on a knowledge item, matches on semantic similarity between the ticket and the KI content, then quotes directly from that KI. If there's no good match, the system deliberately shows nothing rather than making something up. So the quality and structure of your knowledge items directly determines how well this and other AI features work for you.

The more cleanly your content is broken into manageable, self-contained pieces, the more reliable the AI output. A knowledge base that reads pleasantly for humans often contains exactly the kind of noise that confuses an AI.

What we'd currently recommend

1. One knowledge item per problem or topic, not catch-all articles. Smaller, unambiguous pieces are easier to retrieve than large documents mixing multiple topics.
2. Write and title in the end user's language and phrasing, not just internal terminology. AI matches best when the structure of your knowledge item resembles how someone would actually ask the question, for example Q&A format rather than flowing prose.
3. A complete, self-contained solution in the text itself, not just a reference to an attachment or external document. AI can only quote what's actually in the knowledge item.
4. Clearly distinct, separate solution steps instead of dense paragraphs. Short and concrete works better than long and narrative.
5. Regular maintenance. Keep knowledge items current so outdated information doesn't get shared with the end user.

So this is direction, not a mandate. As soon as TOPdesk releases more concrete guidelines or tooling for this, we'll share it here in the community.

Answer for Q2

From a best practice knowledge management (BPKM) perspective we advise the following:

• A clear problem/issue statement, written the way the requester would describe it, not internal jargon.
• Environment/context (what system, version, situation this applies to).
• The resolution, broken into discrete, numbered steps.
• Metadata: article state (work in progress, validated, approved), reuse count, and last-touched date.

Answer to Q3

Yes, AI-assisted keyword/metadata generation is both feasible and already possible with the KI generation feature in TOPdesk, provided there's still a review step before it gets published.

Answer to Q4

Good question, and honestly, one we hear a lot. We know there's strong demand for embedded video in knowledge items, it's a request we've seen repeatedly and take seriously.

At the same time, with limited resources we've had to make tough calls on where to focus first. Video support hasn't been the priority over the past period, other areas have needed our attention more urgently.

That said, it's still very much on our radar for the future, not something we've shelved. Part of why it's not ready yet is that we first need the right supporting architecture in place, for example in incident management and the self-service portal, so that video actually works well end to end rather than being a bolt-on feature. That foundational work is something we're actively building toward.

So the short answer: not yet, but not forgotten either. As soon as there's concrete progress or a timeline to share, we'll give an update.

Answer for Q5

The practical difference: an organization can be very good at "created" (lots of articles exist because people wrote them down after solving things) while being weak at "managed" (nobody revisits them, duplicates pile up, stale articles keep surfacing). We treat these as two different jobs requiring different habits, not one. Created knowledge answers "does an article exist for this," managed knowledge answers "is it still accurate, still used, and still the best version of the answer." The mature state, is when creation is automatic (a natural output of solving problems) and management is deliberate (a continuous quality process), rather than knowledge being managed only when someone remembers to audit it.

Hi ​@katwood.

Thanks for the questions.

About the visibility filters

Let's start with the good news: this is still very much part of our vision and roadmap for the future. We remain committed to building this.

Now the less good news. We started by building and testing a prototype based on a new concept we're introducing across the TOPdesk platform: Access groups, a combination of folders, linked operator groups and permissions. This felt like the right foundation for operator filters going forward.

While testing that prototype, we quickly realized this approach wasn't scalable to other modules. That means we're back at the drawing board on the underlying concept. We hope to restart our development efforts in the first half of 2027.

Created/updated date on Self Service

There are plans to add this yes. I can't give you the exact as this is something for the team owning the SSP. I will make sure your request is shared with them

Hi ​@Marianne Schulte,

That doesn't sound good. I looked up what went wrong here. Your restart has failed since you chose a date and time in the past. This is mostly our fault since the form let you choose a date and time in the past, we should have warned you.

Please request another restart.
 

Hi ​@Paul Verhoeff,

Good question. 
We have a nice knowledge article for this, which you can find here: How do keywords in knowledge items work?
 

Short answer: It sees it as two separate words but also work when searched on together.

Here is the longer answer: 

For keywords to work, it does not matter which separator (comma, space, new line) you use. Furthermore, the order of the keywords is irrelevant.

When searching for parts of words, words that are joined together will not be found, and vice versa. So when you search for mail import, a knowledge item with keyword mailimport will not be found, and when you search for mailimport a knowledge item with the keywords mail and import will not be found. Placing the keywords between quotes will not affect this.

When you search for two words (separated by a space) between quotes, words that are comma separated and words separated by a dash won't be found, but words with other separating characters will be found (including comma space separated words).

Hi ​@Diane Hilton,

Thanks for this request. I will put it on our improvement list.

We want to build a better overview of all the work that needs to be done for a knowledge article. Like replying to feedback, reviewing and publishing an article. For this we also, of course, want to show the date the review or the edit has last been performed. Here is already a ideation/future prototype on how we want to make this work visible.
 

Hi ​@Bayron,

Unfortunately this is currently not possible. Changes are directly saved and published. Therefore we want to introduce a draft state so any change being made first needs to be reviewed and published (we will also take into accounts the permissions!!)
With regards to versioning this is something we want to support in the future, we currently are working on the draft state and after that most likely the permissions side. Versioning would be something that comes after that, but might still take a while.

Hi ​@Diane Hilton,

In the near future (upcoming weeks) we will add a new overal "health" dashboard for knowledge management zooming in Knowledge utilization (%calls with KI's), Self service deflection (views in the SSP) and Knowledge lifecycle (freshness and up to date KI's). This is an intermediate step and will be extended along the way. Biggest metrics we still want to get right is zooming in on views and proper usage.

Hi ​@Thomas Pastoors ,
I'll share the answer I gave at the top of this thread which was based on other rich text capabilities:

Good question, and honestly, one we hear a lot. We know there's strong demand for embedded video in knowledge items, it's a request we've seen repeatedly and take seriously.

At the same time, with limited resources we've had to make tough calls on where to focus first. Video support hasn't been the priority over the past period, other areas have needed our attention more urgently.

That said, it's still very much on our radar for the future, not something we've shelved. Part of why it's not ready yet is that we first need the right supporting architecture in place, for example in incident management and the self-service portal, so that video actually works well end to end rather than being a bolt-on feature. That foundational work is something we're actively building toward.

So the short answer: not yet, but not forgotten either. As soon as there's concrete progress or a timeline to share, we'll give an update.