One product, set up for each organisation
In July 2025 I wanted no custom implementations; by November I saw the value of implementing for each organisation. Both hold at different levels: one codebase and one knowledge model, and for each organisation its own sources, order, answer policy and formats. Every implementation makes the next one smaller.
Written by
Founder and CEO
Essay8 min read
I have put two sentences of mine side by side, and they contradict each other.
On 24 July 2025 I wrote down what Inora’s first implementation had to prove. One line read: “No custom implementations […] Minimal overhead by design”. Nothing built specially for one organisation.
On 20 November 2025 I wrote: “I see all the potential in the AI market right now in boutique consulting and implementation of AI per organisation: unique, and aligned with the context of that organisation.”
Four months apart, and both written with the same certainty. For a long time I thought I had to choose. I no longer do. The two sentences are about different things. The first is about the code; the second is about what an organisation knows and how it works. Below is where the line between them runs, who works on which side of it, and the measure that holds it in place.
What July got right
On 24 July no care organisation was working with Inora yet. The same list asked for one codebase for every care organisation. I meant to translate each organisation’s own terms and procedures into the tasks of the profession, so that every new organisation would add almost no work.
The first half still holds, and I have become stricter about it. Custom code is a separate version. An organisation with its own version of the software gets its own faults, its own releases and its own maintenance. An improvement in one version never reaches the other. The people who build the software then spend their time keeping versions alive instead of making one product better.
The second half was wrong. A quarter of an hour earlier that night I had written down the principle behind it: the profession’s knowledge outranks an organisation’s own procedures. I saw those procedures as a translation layer. That is not how a care organisation works.
What November saw
By 20 November Inora had been in care for eleven weeks: since 4 September at the first care organisation, in Maastricht, and since 15 September in two regions, South Limburg and South-East Brabant. Two minutes before that sentence I had shared a first sketch of what I saw for the next phase: workspaces of its own for each organisation, which its own administrators could set up themselves.
What I saw was that an organisation is not a copy of the profession. It has a quality handbook with its own exceptions, its own report formats, its own abbreviations and its own view of good care. Someone who asks on a ward at night how something is done here wants their own organisation’s answer. An answer that is the same for every organisation is fully right for none of them.
What care workers need is a library of knowledge for their own field, in the context of their own organisation.
But November overshot. “Unique” sounds like building something different for every organisation. “Consulting” sounds like a service beside the product, sold by the hour. Both lead back to the separate versions July rightly warned against.
On 23 February 2026 I wrote Inora’s product architecture: one platform, knowledge on five levels from the profession to the ward, and where two levels disagree, the most specific source prevails. That turned July’s order around. What I had not yet written down clearly is where the one product ends and the organisation begins. I do that here.
One product
This is the same for every organisation.
The code. Every organisation works with the same Inora. No customer has a version of its own, and an improvement reaches every organisation in the same release.
The knowledge model. The same layers: the organisation’s direction, its own sources, the regional agreements it adopts and the national references it uses. The same registers: every source with an owner, a version and a review date. The same rule when two of its sources disagree: the answer follows the source the organisation gives precedence to, says in one sentence that the other differs and marks both, and Inora never decides for itself which one is right. A source from outside the organisation never reaches the care worker.
The line. Inora does not diagnose, calculate a dose or a score, or decide urgency. Those rules are fixed in its core, and no organisation’s setting switches one off.
Set up for each organisation
This differs from one organisation to the next, and it belongs to the organisation.
Its sources. Its quality handbook, its protocols and work agreements, the regional agreements it adopts, the national references it chooses.
Its order. The organisation decides which source leads. One handbook says it in two sentences: the national protocols lead, and the organisation’s own exceptions are in the handbook. An organisation can also let a regional agreement lead on one topic until its own protocol is revised.
Its answer policy. Its goals, vision and methods set the order of the steps, the questions put to the care worker and the words of the answer. Where an organisation works towards clients’ independence, the answer starts with what the client can do themselves. A method never changes what a protocol says.
Its formats. SBAR, SOAP, the goals of the care plan as headings, or a format from its own handbook. The headings sit in a register its quality team keeps. The code sets the headings, never the language model, which writes only what goes beneath them.
Its environment. Every organisation has an environment to itself, with its own sources, users and conversations.
The rule that separates the two lists is short. What differs between organisations is content or setting, and so it is the organisation’s. None of it is code. And what an organisation puts into Inora makes Inora better for that organisation, never for another.
Who sets it up
The people who build Inora set it up, team by team. That is not a detail of the schedule. Write the answer policy with the policy adviser and you see where the language model misreads a method. Put a discipline’s formats into the register and you see where a heading does not fit. That is exactly where you learn what has to become product. If someone else does the setting up, the builder hears it second-hand, or not at all.
We stay until the team works with Inora without us, and we make no team depend on us. After that the organisation maintains its own setup: the quality team the headings, the owner of each source that source, the policy owner the answer policy. We answer for whether Inora is used.
That was also what the sketch of 20 November was after: an organisation that sets up its own Inora. It just needs no code of its own for that. Its own content is enough.
The measure
The difference between setting up and custom work is not always obvious at once. So we hold one measure: every implementation makes the next one smaller.
If we do by hand at the second organisation what we did by hand at the first, it is no longer setup. Something is missing from Inora, and we build it into Inora itself, once, for everyone. An explanation is forgotten; a change to the product stays.
Three examples:
- The language model sometimes made up the headings of a report. A stroke-care chain handover, asked for by name, came back in its handbook structure 0 times out of 7. Since 13 August that structure comes from the handbook itself, not from the model. After the change it came back right 40 times out of 42.
- A new version of the handbook took months to reach Inora. Now the register keeps step with the organisation’s own document system, and Inora answers from the new version within minutes, without anyone at Axiomatic having to act.
- A region keeps the agreements it shares once. Each organisation in the region adopts them source by source and loads none of them itself. The second organisation starts with what the first one needed.
What passes from one implementation to the next is what we learned about the machinery: how Inora sets headings, keeps up with a handbook, adopts a regional agreement. Never an organisation’s knowledge. One organisation’s handbook never enters another’s Inora. What gets smaller is our work, not what makes the organisation its own.
Why not by the hour
We hold ourselves to fewer hours per implementation, never more: each organisation needs less of us than the one before. That only works if we do not earn from hours. A company paid by the hour earns from every implementation that stays as large as the last one, and from every organisation with a version of its own. That says nothing about the people in such a company. A company does what it is paid for.
So we do not bill hours.
The question for every request
For every request from an organisation I now ask one question: is this the organisation’s, or does it belong in Inora? If it is the organisation’s, we set it up and the organisation keeps it. If it belongs in Inora, we build it once, for everyone. A third answer, code for one customer, does not exist.
In July I was right about the code. In November I was right about the organisation. It took me more than a year to see that I was talking about two different things.
Postscript, 2 November 2026
As it stands: implementation has a fixed price per team, beside the annual licence per location or team, and is never billed by the hour (Implementation).
Inora is not a medical device. Inora supports triage: the professional triages, scores and decides. Inora puts the organisation’s own protocol beside the professional, flags what is missing and records the source and version of every answer; it does not diagnose or decide on treatment. Its intended use keeps it outside the medical device rules (The professional decides).