AI and transparency: towards a more explicit approach

In short

AI is pushing in everywhere, sometimes cleverly, sometimes not. Over the past year it has settled right into the heart of our processes at Klaro Cards, often successfully. After a year of that, what was left to do was to make it all more transparent.

A user close to us pointed me to the DALIA scale, which struck me as useful, simple enough to adopt, and informative enough. So we wasted no time putting it in, in two places:

• A page dedicated to AI, which joins our legal pages, and on which we set out where and why we use AI.

• A DALIA visual indicator, on our articles, our series and our white paper for now, which tells you how the content was written, co-written, reviewed, translated. The indicator itself links back to the page above.

Hoping this may inspire others — and above all, give us your views: is it useful? What could be better? What other initiatives of this kind do you know?

Looking back on a year of bringing AI in

Generative AI landed at the end of 2022; we had been using it on and off ever since. It was only in 2025 that the Klaro Cards team took the full measure of what AI now made possible, both for software development and inside the product itself.

AI in and for the product

• As early as July 2025, we had grasped the value of an MCP server, one that lets our users connect their favourite AI agent to their projects. The first version of that plumbing was in production in November 2025, and has been improved ever since.

• In January 2026, Korben covered our work on the expression language Elo, cut to the measure of what Klaro Cards needs, and whose compiler was largely written by Claude Code using a black-box method (a harness technique, a word that has seen a lot of use since).

• Through 2026, AI also settled naturally into our development and editorial processes, which let us move faster on some features, improve the user experience, extend the guides and the translations, and so on.

The pivot the new website carried

We rebuilt the Klaro Cards website recently, around an editorial choice: to name operational friction and put it front and centre — the thing SMEs fight every day.

Nothing we had written before made a transparency process urgent. Most blog posts up to then were written and translated 100% by hand (D0 on the DALIA scale); only the feature announcements were drafted by AI and then reviewed (D3 or D4), which was easy enough to explain and to defend (it is never more than a changelog). Our guides fell into much the same category — though that one was already starting to itch.

The white paper on operational friction (link above) is another matter. The core of the argument is ours, but a large part of it was co-written with AI, then reviewed, rewritten and improved by eight hands (Thomas, Bernard, Marc, Stuart), in two languages. A genuine D3 on the DALIA scale, and that deserved to be said somewhere.

The series on Jira

More recently still, we decided to spend some time on Jira. Klaro Cards was born in 2016, as an alternative to Jira and Trello for its founders. We have (oddly) rarely discussed Jira here, preferring to focus on the processes of industrial SMEs — risks, non-conformities and the like — rather than those of the software industry, scrum and agile.

But Atlassian, who publish Jira, decided to remove every option for keeping your data in Europe. Having realised over the years that Jira was heavily used for something other than software development, we made the strategic choice to take the subject back.

Out of that came our series on the end of Jira Data Center, written to see more clearly ourselves, and to cover the subject in the hope of catching the attention of the European public bodies and SMEs still on Jira who are either looking for an alternative for 2029, or have moved to Jira Cloud in the United States rather more because they had to than because they chose to.

Long story short, that Jira series is co-written with AI. We chose the table of contents and the subjects to cover. The AI searched for us, sourced its claims, handed in drafts. We read them, reworked them, sometimes rewrote them. We read the sources, checked the claims, argued our own corner (obviously), and said where we stand.

All of that work deserved to go out in full transparency. The DALIA scale arrived at just the right moment — it is D3 — thank you, Jeanne, the user mentioned above.

A last word

Bringing AI in raises questions that are societal, environmental, ethical, and sometimes personal.

Rather than avoiding it at all costs, or embracing it wholesale, you will read on our page that we are looking for a balance...

A balance of that kind is only possible in transparency, and in conversation. Carry that conversation on with us, through any of the ways you have of reaching us.

About this article

Since that is what it is about: the French original of this article was written by hand. An AI then read it over — spelling, style, consistency with our other pages — and a few turns of phrase come word for word from what it proposed. It also checked that every link led somewhere. That edition is rated D1.

The English version you are reading is a different piece of work: written entirely by an AI from the French, then reviewed by a human. It is rated D4. From one language to the other, the same article carries two different badges — that is deliberate, and it is exactly what the scale is for.

This last section was written by Claude AI, on my request. Easily to catch, isn't it 😉 ?