Enterprise AI buying shifts

Enterprise buyers are shifting from “best benchmark” models to models that are fast, cheaper to run, and easier to manage inside existing software stacks. (futurumgroup.com) Futurum Group said April 12 that 67% of organizations already have generative artificial intelligence in production and 75% expect to increase budgets over the next year, based on its first-half 2026 survey of 838 decision-makers. The same survey put OpenAI GPT adoption at 61%, ahead of Azure OpenAI at 50% and Google Gemini at 47%. (futurumgroup.com) OpenAI’s March 3 release of GPT-5.3 Instant fit that demand by emphasizing faster responses, better web-search answers, and fewer refusals or warning-heavy detours in everyday use. OpenAI said those changes target “tone, relevance, and conversational flow,” areas it said do not always show up in benchmark tables. (openai.com, openai.com) The buying shift is also showing up in workplace software, where the product work is increasingly about controls and connections rather than raw model novelty. OpenAI’s ChatGPT Business release notes were updated April 13 and list a steady stream of admin and workflow changes through late March and early April. (help.openai.com) On April 8, OpenAI added Outlook shared mailbox and shared calendar actions for ChatGPT Business, including reading shared mailbox messages, moving messages, and sending plain-text email on behalf of a shared mailbox when Microsoft permissions allow it. The note also told workspace owners and admins to review action controls before enabling the new actions. (help.openai.com) On April 2, OpenAI introduced usage-based Codex seats for ChatGPT Business with no fixed monthly seat cost, cut subscription-based ChatGPT seats by $5 per month, and said eligible workspaces could earn up to $100 in Codex credits per added seat, capped at $500 per workspace. (help.openai.com) The week before that, on March 27, OpenAI rolled out updated Box, Notion, Linear, and Dropbox apps with new actions, including write capabilities where supported, and said admins could review app actions and required scopes in workspace settings. (help.openai.com) That is the language of information technology procurement: permissions, scopes, billing, seats, and connectors to systems companies already use. OpenAI’s business pricing page markets ChatGPT Business at $20 per user per month on annual billing and says Enterprise adds governance features including custom data retention policies, encryption at rest and in transit, and no training on business data by default. (openai.com) The model lineup is also being sorted by workload instead of one-size-fits-all prestige. OpenAI’s developer documentation lists GPT-5.3 Chat, which points to the GPT-5.3 Instant snapshot used in ChatGPT, at $1.75 per 1 million input tokens and $14 per 1 million output tokens, while the Business and Enterprise rate card treats GPT 5.3 Instant as an unlimited message model and charges credits for heavier tools such as deep research, agents, and higher-end reasoning models. (developers.openai.com, help.openai.com) OpenAI has also been pruning older ChatGPT choices while steering paid workspaces to the newer stack. Its rate card says GPT-4o, GPT-4.1, GPT-4.1 mini, OpenAI o4-mini, and GPT-5 Instant and Thinking were retired from ChatGPT on February 13, 2026, with GPT-4o in Custom GPTs fully retired after April 3. (help.openai.com) The result is a market where “better” increasingly means faster answers, lower operating cost, and tighter administrative control, not just a higher score on a lab test. Vendors still compete on model quality, but the money is moving through seat pricing, connectors, and governance settings that make artificial intelligence look more like standard enterprise infrastructure. (futurumgroup.com, help.openai.com, openai.com)