Despite all the hype around AI, most companies aren’t ripping out their existing technology stacks to make way for it. Instead, a more pragmatic approach is emerging, one that’s less about replacement, more about adaptation and one that hedges bets around which system(s) will “win out.”

At the centre of this shift is a changing philosophy around build versus buy. Yes again (or maybe it never went away).

Historically, many media organisations invested heavily in custom-built systems — from identity solutions to CMS platforms — often as a way to maintain control or differentiation. But as Florent Daudens highlighted, “We should compete on journalism, not tooling.” And AI is exposing the limits of the in-house build approach.

Perhaps more accurately: Not every layer of the stack is worth owning.

Commodity systems, particularly those that are widely solved, are increasingly being replaced with external solutions. The logic is straightforward: Internal resources are better spent on areas that directly contribute to unique value rather than maintaining infrastructure others can provide more efficiently. 

This is leading to more selective and more strategic, in-house development, which may include adapting an off-the-shelf or open source system for your own needs.

At the same time, the structure of the stack itself is evolving. At the last INMA Product and Tech Advisory Council meeting, we discussed the separation of CMS and front-end layers.

In many cases, legacy front-ends have become bloated, impacting performance and limiting flexibility. Rather than replacing everything at once, many organisations have started by simplifying and modernising the front-end, improving speed, reducing complexity, and creating a cleaner foundation for AI integration.

Others are taking a more modular approach to CMS development, breaking systems into components that can be iterated and enhanced independently. This creates more flexibility, particularly important as AI capabilities continue to evolve rapidly.

What’s notable in all of this is the sequencing of these changes. Before layering in AI, teams are focusing on simplification. Cleaning up the front-end, reducing technical debt, and clarifying system boundaries are becoming prerequisites for effective AI adoption.

Without that groundwork, AI risks amplifying existing complexity rather than solving it.

Another emerging tension sits around standardisation.

As organisations move toward more federated models — with teams operating more independently to maintain individual brand integrity — maintaining consistency in tools and platforms becomes harder. 

At the same time, there is a reluctance to lock into specific vendors, especially at the foundation model level, too early. This makes sense given how quickly the AI landscape is shifting.

The result is a more cautious approach to standardisation when it comes to newsroom and product tooling.

Some organisations are adopting multiple AI tools in parallel, experimenting with different models and licensing structures, while deliberately avoiding long-term commitments to the core foundational models. While this may create short-term complexity and cost, it preserves long-term flexibility.

AI isn’t replacing the stack. It’s forcing organisations to become much more intentional about what they keep, what they rebuild, and what they let go. 

If you’d like to subscribe to my bi-weekly newsletter, INMA members can do so here.

Banner photo: Adobe Stock By Chayada.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .