Why AI sovereignty is now a board-level question

The three market trends no executive team can afford to ignore

Reflections following the AI Strategy Summit 2026

At the beginning of June, I had the privilege of sharing the stage at the AI Strategy Summit in Stockholm, hosting a fireside chat together with IBM’s Vahid Zohali, General Manager IBM North region & CEO IBM Sweden and a breakfast seminar session with Dave Pemberton, IBM WebMethods Hybrid Integration Product Management Lead.

The conversations on stage – and perhaps even more so in the breaks and exhibition area afterwards – confirmed something that has been building for months: the AI conversation in the boardroom has shifted. We are no longer asking whether or how to adopt AI. We are asking on whose terms.

That question matters more than most organisations yet realise. Because beneath the surface of the AI market, three trends are overlapping and reinforcing one another – and together they are reshaping the strategic agenda for every forward-leaning company.

Trend one: the real cost of tokens is arriving

The era of subsidised intelligence is ending. As hyperscaler-based LLM and platform providers adjust pricing to reflect the true, consumption-based cost of token generation and inference, customers are responding rationally: they are looking elsewhere. Smaller language models – fit for purpose in relation to each task, process and complexity level – are proving that bigger is not always better. Local hardware and edge-based infrastructure are making it viable to generate tokens independently of cloud-based AI services altogether.

The lesson is not “abandon the cloud.” The lesson is that token economics is now a discipline. Not every process needs a frontier model, and matching model capability to business task is becoming a genuine competitive cost factor.  This follows the pattern from the preceding era of cloud infrastructure subsidized go to market approach with a certain level of ”overmigration”, and current rebalancing or reversal to on-prem seen for some types of workloads.

Trend two: trust is the new bottleneck

At the same time, model diversification and differentiated token value are colliding with a very human reality: your employees are already using AI tools – inside and outside the governance of your IT organisation. Shadow IT has found its most powerful expression yet.

This was a central theme in our breakfast seminar ahead of the summit, and I will repeat the point here: the answer is not prohibition. The crowd-driven adoption happening in your organisation right now is exactly the innovation, creativity and efficiency energy you need to stay competitive in what I would call the disruption economy – the successor to the digital economy and the attention economy. But that energy only compounds into advantage when data ownership, control, security and reliability – in a word, trust – are designed in, not bolted on. Governance, done right, is an accelerator, not a brake.

Trend three: sovereignty moves to the top of the agenda

Add to this the unpredictability of cloud infrastructure costs, hardening regulatory resilience requirements, and an increasingly uncertain geopolitical situation around data residence and independence – and it becomes clear why cloud sovereignty and data sovereignty are climbing rapidly towards the top of the strategic agenda.

The logical next step is — AI sovereignty — : the ability to autonomously train, refine and operate fit-for-purpose, business-specific models that support your competitive differentiators and your token cost position – without vendor or platform lock-in carrying practical or prohibitively expensive consequences. Businesses need to own and control their data. A measure of independence and portability must now be a standing item in risk and strategic planning, not an afterthought.

What this means in practice

If the three trends share a common thread, it is this: AI value does not come from the model. It comes from how the model connects – to your data, your systems, your processes and your people.

All AI is integration driven. Integration enables AI, and AI empowers integration.

My strongest recommendation to executive teams is not to start with specific tooling or model selection at all, but with a unified, holistic governance framework spanning three domains that have traditionally been managed apart:

1. Data – ownership, residence, quality and lifecycle, as the foundation for both trust and model performance.
2. Identity – increasingly including non-human users; agents and automations need access governance just as employees do.
3. System integration – the connective tissue that determines whether AI-driven automation scales safely or sprawls uncontrollably.

Treated separately, each becomes a bottleneck. Treated as one framework, they become the platform on which sovereign, cost-efficient and genuinely trusted AI capability is built.

Organisations need to turn integration, data and governance into the enabler of their AI ambitions rather than the obstacle to them. The companies that get this right will not just adopt AI – they will own it.

If these themes resonate – or provoke – I would genuinely enjoy continuing the conversation. That, after all, is what summits are for.

Author: Fredric Travaglia is Business Architect and Offering Owner for Enterprise Integration at Epical.