# A question without an answer is work

16 September 2026 · Field note · 5 min read

By [Zohaad Fazal](https://axiomatic.digital/en/about#zohaad), Co-founder, deployment and markets

When the sources do not cover a question, a general model guesses. Inora says so, and the topic becomes a task for the owner of that knowledge, without the question and without the name of the person who asked. Then the owner writes a test question for the next version. That loop makes Inora better, not the model.

I have led the implementation of Inora since February. If I had to name one thing I have come to value on the floor, it is an answer that is not an answer.

## What a general model does with a gap

A general model always has an answer. When something is not in the documents you give it, it fills in from everything it has ever read. Sometimes it makes something up, and then it sounds just as sure as when it is right.

For a senior care assistant in the middle of the night, that is the worst answer they can get. A wrong answer given with confidence gives them no warning. An honest “this is not in there” does.

## What Inora does

When the organisation’s sources do not hold the answer, Inora says so. It says where it looked: in the organisation’s own sources and in the regional agreements it has adopted. And it names who in the organisation owns the subject. It does not guess, and it never answers a question about care, the organisation, the law, money, numbers or a person from what its model learned elsewhere.

Take a question such as: how do we use the new donning aid that came this week? If that is nowhere in the sources, the senior care assistant does not get an average of the internet. The senior care assistant hears that it is not there, where Inora looked and who owns the subject.

How do we use the new donning aid that came this week?

- Your own documents no passage
- Designated protocols no passage
- Adopted regional agreements no passage

This is not in your sources.

Searched: your own documents, the protocols those documents designate and two adopted regional agreements.

This topic falls under aids; owner: the quality adviser.

Related, without an answer to your question: your work agreement on compression stockings. Work agreement: compression stockings, North district team, §2

The topic goes to the owner, without your question or your name.

When it is not in the sources, Inora says so, and where it looked.

## Where the topic goes

It does not end there. A language model classifies every message to Inora: answered or not, whether knowledge was missing, and what kind of knowledge the question needed. No one reads the conversation to do so.

So every question without an answer gets a cause. If the knowledge is missing, two sources contradict each other or a document has expired, a task lands on the list of the person who owns that area of knowledge. The task sits in the document system the organisation already uses, with the topic and the source that was missing; the question itself never goes with it. The same happens when a care worker reports that an answer falls short.

Nor does the name of the person who asked. That is not a detail. Care workers have to be able to ask Inora anything, including what they would rather not ask a colleague. If they do not know something, they must be able to say so without it going on their record. The gap is work for the owner of the knowledge, not for them.

## What is ours

Not every gap is in the sources. Sometimes it is in Inora itself, and then it comes to us.

Two things kept coming back in implementation: an answer that was in the quality handbook and did not come, and abbreviations only the organisation itself uses. Explaining it to a team does not fix that; the next team runs into it again. So it is now in Inora. The organisation’s list of abbreviations is a source of its own, and both cases are in every organisation’s test. What one team runs into is fixed for every team after it.

## Where it began

The idea of counting gaps is older than my time at Axiomatic. On Friday 28 November 2025 Ahsan built the first measures of the quality of the knowledge library. The first measured how well what Inora retrieves fits the questions users ask, judged by a language model, and it was to run over every question asked until then. That made the gaps in the knowledge library countable.

Counting was the start. Since 21 January 2026 every organisation has had [its own usage report](https://axiomatic.digital/en/updates/per-team-never-per-employee), with the topics that found no good answer yet, each with the number of times it was asked. That report is the agenda for its knowledge.

## The test for the next version

Closing a gap is half the work. The other half is knowing it stays closed.

A mistake you only fix comes back. A mistake the system prevents does not.

On Monday 9 February 2026 Ahsan read through a stress test of ten cases, written to catch an assistant out. He wanted one more test: does the system retrieve the right knowledge from the knowledge base?

That test exists now. Every topic on the list, once its source is there, gets a test question that the owner writes personally. It joins the organisation’s test set, with the source that now answers it. Every new version of Inora goes through those test questions before the organisation works with it, even when only the language model changes. With every version the organisation sees what changed and how the test went.

## Not the model

A language model learns nothing from the conversations with Inora. No one trains a model on an organisation’s documents, questions or answers: not us, and not any model provider. So the model does not get better from what a senior care assistant asks. Inora does.

What makes Inora better is the loop: the topic without an answer, the cause, the owner, the source, the test. When the model changes, the new version goes through the same test questions. The sources the organisation adds in that loop stay its own.

## What the board sees

Every month the board sees how Inora is used, summed across all teams: which topics staff ask about, which topics found no answer yet, which gaps were closed and how much time it gave back. Never by employee, and never a question as the care worker asked it. So a gap is not only on the knowledge owner’s list. The board also sees whether it was closed.

A question without an answer says nothing about the person who asked it. It says where the organisation’s knowledge is not yet in order. That is work, it has an owner, and every month shows whether it was done.

## Postscript, 2 November 2026

So a question without an answer goes to whoever owns that knowledge as a topic: the topic without a source, rewritten without identifying detail, with the number of times it was asked, per protocol and per team. The question itself, an example question and the name of whoever asked never go with it. The monthly report shows, per team, the topics staff ask about and the topics that found no answer yet, with a count, only for teams of at least five people; the board and quality receive the sum across all teams ([Privacy](https://axiomatic.digital/en/trust/privacy)).

## Do you work in care and use Inora?

Sign-in and help go through your own organisation: your team’s project lead and ambassadors can help you.

## Read on

- [The gap is not the model. It is the knowledge.](https://axiomatic.digital/en/updates/the-gap-is-not-the-model.md): The knowledge care runs on lives in the protocols and in the heads of the few people in every team who quietly do the work better. A general model does not bring the two together, not even with your documents added. A knowledge system does.
- [Per team, never per employee](https://axiomatic.digital/en/updates/per-team-never-per-employee.md): From 21 January 2026 every organisation that works with Inora receives its own usage report. It is drawn from the system’s own log, counted per team and never per employee. The reason: in South-East Brabant the impression of use and the log were far apart.
- [The map is not the territory](https://axiomatic.digital/en/updates/the-map-is-not-the-territory.md): A care organisation’s knowledge is four maps of one territory, the moment of care. It stays true as long as the maps are read at that moment and redrawn by it. Inora keeps both directions running and never reads the territory itself.
