IDC AI Insights
You Don't Choose a Model, You Govern a Constellation
Model selection stopped being a one-time technical choice. See why enterprises are naming an owner instead of chasing a winner.
The Reality
The Model Landscape Isn't Consolidating. The Org Chart Is.
IDC tracked 66 open language models released across CY24 alone, with 34 more in the second quarter of 2025, and the release pace hasn't slowed. (IDC #US52237124; #US52237325)
Where This Fits
Three pillars carry the model governance story, from fit to ecosystem to ownership.
Model selection should be use-case-driven, not brand-driven. The right small, domain-specific, or regional model frequently outperforms a general-purpose frontier model on cost and fit for a given task.
Key takeaway: The best model for the job is rarely the biggest one.
The real capability enterprises need isn’t picking a model once, it’s an ongoing evaluation and orchestration discipline that governs a portfolio of models as that portfolio keeps changing.
Key takeaway: You don’t choose a model anymore, you govern a constellation.
A new role, Chief AI Officer, is emerging specifically to own model, data, and governance decisions that used to be nobody’s full-time job. Enterprises without a named owner are accumulating app and model sprawl no one is accountable for.
Key takeaway: A model portfolio without a named owner is a liability accumulating quietly on the balance sheet.
As enterprises enter a multi-model, multimodal, multiagent era, the real question is: how will you navigate the expanding constellation of AI models to unlock business value?
Tim Law, Research Director, AI and Automation, IDC
Continue the Series
More on Enterprise LLM Selection
- Video: Govern the Ecosystem, Not the Model
- Quote Card Series: Someone Has to Own This
Name the Owner Before the Constellation Grows
Subscribers get direct access to the MarketScape and Chief AI Officer research. Not yet a subscriber? Start with IDC Quanta.