Editorial analysis: Enterprise AI practitioners should treat base frontier models as a broadly available input rather than the primary source of competitive advantage. Capture and orchestration of proprietary data, robust integration with workflows, and product-level engineering typically determine whether AI initiatives produce measurable ROI.
What happened
In a June 29, 2026 post titled "Are Frontier Models Becoming A Commodity?", Josh Bersin frames the current market as one where major providers - OpenAI, Anthropic, Google, and Microsoft (MAI) - supply powerful base models while open-source projects such as GLM, Deepseek, Kimi, Mistral, and IBM's Granite broaden options for adopters. Bersin writes that his research for an Enterprise AI Playbook, surveying 200+ companies, found only about 8% are building real enterprise applications. He also writes that much early experimentation has been subsidized by roughly $1.5 trillion of forward-looking investor capital. Bersin references a recent article by Satya Nadella as sharing the same basic premise.
Industry-pattern observations: Companies that move from experimentation to production typically invest in data plumbing, access controls, observability, and UI/UX for specific workflows. For practitioners, this means engineering effort usually outweighs model selection once baseline capability is available. Open-source models lower vendor lock-in but increase the burden on teams to manage updates, safety, and inference cost.
What to watch
adoption metrics (percentage of teams deploying production apps), cost-per-inference after optimization, and the emergence of standardized connectors for common enterprise data sources. Observers should also track whether increasing availability of capable open models materially reduces model licensing costs versus the operational expenses of productizing AI.
Key Points
- 1Frontier models are becoming a widely available input; competitive differentiation shifts to proprietary data and integration.
- 2Bersin's survey of 200+ companies found only 8% building production enterprise apps, highlighting an adoption gap.
- 3Lower-cost open-source models expand options but increase the engineering burden to manage safety, updates, and inference costs.
Scoring Rationale
This piece is a pragmatic industry synthesis rather than a technical breakthrough. It highlights adoption and ROI patterns relevant to practitioners, but it does not introduce new models or benchmarks.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems