Forrester Outlines Sovereign AI Strategy At TalosCon 2026
Forrester emphasizes preserving choice in AI architectures, citing data residency as insufficient for sovereignty. This approach matters as enterprises seek to maximize AI value while maintaining control.
Original reporting · forrester.com
Original: “TalosCon 2026: The Winning AI Strategy Is To Preserve Choice”
Read “TalosCon 2026: The Winning AI Strategy Is To Preserve Choice” at Forrester →
How this was written. Message Verified wrote this article with a language model from Forrester’s reporting (model: @cf/meta/llama-3.3-70b-instruct-fp8-fast); it describes that reporting rather than adding to it. The original, linked above, is the full story.
Key points from Forrester’s report
- Forrester emphasizes the importance of preserving choice in AI architectures to achieve sovereignty.
- Technology leaders should prioritize freedom of choice across five key layers: data, models, compute infrastructure, deployment environments, and governance controls.
- A portfolio approach to model selection can help enterprises match capability to workload requirements and avoid unnecessary costs.
- Enterprise data remains a key differentiator in AI, and factors like governance and workflow integration can matter more than model performance.
- Preserving the freedom to choose will be central to the future of enterprise AI, rather than simply selecting a winning model.
The conversation around sovereign AI often focuses on data residency and model location, according to Forrester. However, these factors alone do not guarantee sovereignty. An organization must be able to change models, move workloads, and replace infrastructure providers to truly achieve sovereignty.
Forrester reported that technology leaders should prioritize AI architectures that preserve freedom of choice across five key layers: data, models, compute infrastructure, deployment environments, and governance controls. The ability to make changes without rebuilding applications is a crucial test of sovereignty.
At TalosCon 2026, the growing maturity of open-weight models was a notable theme. These models may not match the capabilities of frontier models, but they offer viable alternatives for various enterprise workloads, particularly when paired with retrieval and task-specific tuning.
As noted by Anant Yardi, “you do not need a frontier model to approve invoices.” This observation highlights a disconnect in enterprise AI, where the focus on benchmarks and foundation models may not align with actual workload requirements. Forrester suggests adopting a portfolio approach to model selection, using smaller models for constrained tasks and open-weight models for workloads requiring transparency and customization.
Enterprise data remains a key differentiator in AI, as model performance is often secondary to factors like retrieval-augmented generation, governance policies, and workflow integration. For many use cases, a smaller model connected to trusted enterprise content can generate more useful outcomes than a larger model without relevant context.
Forrester advises technology leaders to resist treating AI strategy as a model-selection exercise and instead focus on architectural decisions that affect portability, governance, and long-term strategic flexibility. As foundation models mature, control, portability, and dependency management will become increasingly important evaluation criteria for enterprise AI investments.
Ultimately, the enterprises that create the most value from AI will be those that can combine enterprise data and processes with the freedom to adopt new models and evolve their architectures as the market matures. Preserving the freedom to choose will be central to the future of enterprise AI, rather than simply selecting a winning model.
Questions and answers
What is sovereign AI?
Sovereign AI refers to an organization's ability to control its AI operations, including changing models, moving workloads, and replacing infrastructure providers, without being dependent on a specific provider or model.
Why is data residency insufficient for sovereignty?
Data residency alone does not guarantee sovereignty, as an organization must also be able to change models, move workloads, and replace infrastructure providers to truly achieve sovereignty.
What is a portfolio approach to model selection?
A portfolio approach involves using different models for different workloads, such as smaller models for constrained tasks and open-weight models for workloads requiring transparency and customization.
