The door Atlassian closed

A date that has not yet arrived in many IT departments

Some deadlines only take effect with a delay. Atlassian has set a hard end date for Jira Data Center that received a fair amount of attention in the trade press, but that has not really landed as a concrete deadline in many of the affected companies: since 30 March 2026 there are no new Data Center subscriptions for new customers. Existing customers can still buy until 30 March 2028. On 28 March 2029 the licences expire for good, and the products become read-only.

Anyone running Jira Data Center today therefore faces a choice that can no longer be deferred, even though the end date is still some years away: migrate to the cloud, or carry on with an expiring product until the last deadline has passed as well. A standard migration to the cloud takes between 12 and 16 months according to Atlassian's own figures — not a decision that can be taken and implemented within a quarter, but an undertaking that brings its own planning, its own budget and its own risks.

Why an interesting interim period is opening up right now

Between the end of new sales and the final expiry lies an interim period of several years in which a large number of organisations continue to work with a platform for which there is no longer any official way to connect a modern AI assistant. Atlassian's own MCP server for connecting AI models is explicitly built for Atlassian Cloud only — for Data Center the vendor itself offers no equivalent route, neither officially nor unofficially.

Anyone who nevertheless wants to connect AI to their Jira during this interim period — and the pressure to do so grows with every new generation of models that gets tried out inside the company — therefore has only two realistic options: to use community-built connectors without any redaction and without an approval mechanism, or to forgo AI in this area entirely while their own workforce has long since been experimenting with other tools in parallel. Both are unsatisfactory, for different reasons.

The demand is not assumed, it is documented in surveys

That this demand is real and not merely an assumption from within the software industry is shown quite clearly by recent surveys from Germany: the share of companies actively using AI has risen from 17 per cent to 41 per cent within a single year — the fastest increase since comparable surveys began. At the same time, 70 per cent of the companies surveyed state that they have already stopped at least one innovation project because of data protection rules or legal uncertainty, and 54 per cent say explicitly that data protection is holding back the use of AI in their own organisation.

Taken together, these two figures produce a picture that goes beyond mere curiosity: usage is growing faster than internal control over it. 41 per cent already use AI actively, but only 21 per cent have any formal AI strategy for it at all. The difference between those two figures is, at its core, uncontrolled use — employees who have long since been experimenting without their IT department or their data protection officer knowing the full extent of it.

Why this is not an argument for new Atlassian customers

At this point a clarification is worth making that is easily skipped over: this window addresses existing customers exclusively, those who already have Data Center in use — it is not an argument for winning new customers for Atlassian, because that route has not existed since 30 March 2026. Anyone not running Data Center today can no longer buy it new in order to end up in this situation. The target group for this interim period is therefore clearly bounded from the outset — and for exactly that reason more manageable and more concretely graspable than a general "AI security market" that wants to address everyone and everything.

Who else is moving into this gap

It would be naive to assume that a gap this obvious would stay unoccupied for long. The Atlassian Marketplace already contains a number of add-ons that deal with personal data in Jira tickets — on closer examination, however, it turns out that most of them tidy up and clean the existing Jira store itself, rather than securing the handover of content to an AI model. That is a related but clearly different task: a tool that searches old tickets for personal data and flags it does not help with what happens in the moment an assistant reads exactly that ticket and passes it on to a language model.

Community projects that connect an AI assistant directly to Data Center exist as well — mostly without any redaction before the model and without a mechanism that holds writing actions for a human approval. They solve the technical problem of connection, not the actual problem of control over what happens in the process.

What this interim period means for the affected companies

For organisations that cannot migrate at short notice for regulatory reasons or because of the sheer size of their installation — and that affects heavily regulated sectors such as financial services, healthcare or public administration disproportionately — this interim period is not a theoretical footnote but the real period in which daily work takes place. They are simultaneously under migration pressure and under the pressure of expectation from their own workforce, who know AI tools from their private lives and would like to use them at work too.

A tool like JiCo is built for exactly this phase — not as a replacement for the cloud migration that is actually due, but as a way of not letting the intervening years pass unused or uncontrolled. The architecture itself does not stand in the way of a later cloud migration, because the redaction and the approval check sit inside the JiCo process itself, not in a fixed dependency on Jira Data Center. A finished cloud connector has not been built yet, though — an honest limitation that belongs to the same attitude with which the product's own boundaries are otherwise discussed.