Cohere has launched North 2, an enterprise AI platform that lets businesses use models from rival providers and set spending limits for staff and teams. It also offers pre-built agents for workplace tasks, giving companies a way to organise AI work without making Cohere’s own models the only option.
Cohere Watch analysis
What happened
North 2 is designed to help employees run AI agents for longer, more complex assignments. The Logic reports that the platform retains more details of those tasks to improve automations over time, and lets businesses measure the return of AI-powered work within the platform.
Customers can connect models from Anthropic, OpenAI and Google, as well as open-source models from providers including Alibaba, DeepSeek and Z.ai. Teams can use pre-built capabilities and agent libraries aimed at functions such as finance and marketing. Cohere says businesses can set budgets, monitor spending and automatically switch simpler tasks to cheaper tools. The company has not built its own model router into North 2.
Why it matters
Enterprise AI buyers are increasingly weighing model performance against cost, data control and the risk of being tied to one provider. North 2 puts model choice and spending controls in the same workplace platform, while offering ready-made tools for staff who are not developers. That makes the launch a practical bet on helping businesses move beyond scattered experiments and into broader use.
The trade-off is worth watching: a platform that makes it easier to mix models could give customers more flexibility, but it also asks them to trust one provider to organise work across several others. Cohere’s own models remain part of the offer, but North 2’s openness means customers can choose alternatives where they suit a task better.
Our read
This is a meaningful change in enterprise AI packaging: choice and cost management are part of the product, not a spreadsheet bolted on afterwards. The pitch is especially relevant to organisations trying to expand AI use without handing every team an unmetered tab. The harder proof will be whether businesses can measure useful returns and manage costs as agents take on more work.
What to watch
- Whether customers adopt models from multiple providers within North 2.
- How its spending limits and automatic switches work in practice.
- Whether Cohere adds its own model router, which the company says it has not yet built.
- What evidence customers publish about savings and results from agent-based work.
Discussion spark: Would you rather have one AI platform that lets you switch between competing models, or choose each model and tool separately for yourself?
Sources and evidence
- Cohere opens up its AI platform to give businesses more control – The Logic (5 October 2026, 13:00 UTC)
not affiliated with or endorsed by Cohere