Enterprise AI goes multi‑model
Microsoft 365 Copilot and related product moves are shifting enterprise AI from a single ‘best model’ strategy toward multi-model workflows that route tasks to specialised agents and log model choice as part of the product. OpenAI also updated GPT‑5.4’s context handling and rolled out a ChatGPT Pro plan with expanded Codex access, underlining a market split between higher‑intensity tooling and lighter consumer tiers. ( )
Microsoft is starting to treat artificial intelligence models less like one all-purpose employee and more like a team with different job titles. In Microsoft 365 Copilot’s Researcher tool, OpenAI’s GPT can draft an answer and Anthropic’s Claude can review it in a feature called Critique. (geekwire.com) That is a break from the first phase of the artificial intelligence boom, when companies spent two years arguing about which single model was “best.” Microsoft’s new Council feature instead shows side-by-side outputs from different models so a worker can compare them inside the product. (computerworld.com) Microsoft is building this into its office software, not leaving it as a hidden engineering setting. In its March 9 Wave 3 announcement, the company said Microsoft 365 Copilot was adding “multi-model intelligence” alongside new agent features for business users. (microsoft.com) The practical reason is simple: writing a memo, checking a claim, searching company files, and planning a project are not the same job. A model that is fast at drafting can be weaker at review, so Microsoft is separating generation from critique inside one workflow. (computerworld.com) That also changes how companies keep records. GeekWire reports that model choice is becoming part of the product itself, which means enterprises can see not just the answer but which model produced it and which model checked it. (geekwire.com) OpenAI is moving in a parallel direction from the model side. Its ChatGPT release notes say GPT‑5.4 Thinking now maintains context better on questions that require longer reasoning and gives users an upfront plan while it works, which is meant to reduce extra back-and-forth on complex tasks. (help.openai.com) OpenAI is also splitting its pricing by intensity of use instead of pretending one subscription fits everyone. On April 10, Business Today and TechCrunch reported a new $100-a-month ChatGPT Pro plan with 5 times higher Codex usage limits than the $20 Plus tier. (businesstoday.in) (techcrunch.com) Codex is OpenAI’s coding agent, so that new plan is aimed less at casual chat and more at people who run artificial intelligence on real software work all day. OpenAI’s March 4 release notes also added a Windows app for Codex that can run multiple coding agents in parallel with isolated workspaces and reviewable code changes. (help.openai.com) Put those two moves together and the market starts to look different. Microsoft is turning enterprise software into a dispatcher that routes work across models, while OpenAI is charging more for heavier-duty agent use and keeping lighter tiers for everyday chat. (geekwire.com) (businesstoday.in) The old question was “which model wins.” The new question inside companies is closer to “which model does which step, under whose supervision, and with what audit trail,” and Microsoft 365 Copilot is starting to answer that inside Word, Excel, and the rest of the work stack. (microsoft.com)
