Enterprise AI shifts to economics

Enterprise buyers are spending less time chasing the newest artificial intelligence model and more time on price, speed, and controls they can deploy across a company. (futurumgroup.com) The Futurum Group said 67% of organizations already use generative artificial intelligence in production, and 75% plan to increase spending in 2026. Its analysis said buyers now weigh latency, operating cost, and governance alongside raw model performance. (futurumgroup.com) OpenAI’s recent ChatGPT Business updates fit that shift. Release notes published in April 2026 highlighted connectors for Azure Boards, Basecamp, and Zoho Customer Relationship Management, while earlier business-plan updates added record mode, internal-tool connectors, and flexible pricing. (help.openai.com, openai.com) OpenAI has also kept adding the kinds of controls information-technology departments usually ask for before a broad rollout. Its business and enterprise documentation now centers on single sign-on, System for Cross-domain Identity Management provisioning, role-based access controls, and audit and compliance logging. (help.openai.com, help.openai.com, openai.com, help.openai.com) That marks a change from the first wave of generative artificial intelligence adoption, when much of the attention went to benchmark wins and public model launches. In large companies, the harder question is whether a tool can plug into identity systems, data sources, and billing rules without creating a new security problem. (openai.com, help.openai.com, futurumgroup.com) Other surveys point in the same direction. Deloitte said in its 2026 State of AI in the Enterprise report that companies are moving from ambition to activation, while a separate 2025 Deloitte analysis said artificial intelligence is taking a growing share of digital budgets and forcing executives to defend returns more clearly. (deloitte.com, deloitte.com) Deloitte’s year-end 2024 generative artificial intelligence survey also found companies were focused on scaling deployments and overcoming barriers to value, not just experimenting with pilots. That is the phase when procurement teams start asking how many users need access, what each workflow costs, and how fast answers arrive. (deloitte.com, deloitte.com) OpenAI’s April 9, 2026 enterprise release notes captured the same tradeoff in product form. The company introduced GPT-5.3 Instant Mini as a fallback model after rate limits, and on April 2 it rolled out a new Codex seat with usage-based pricing instead of a fixed monthly charge. (help.openai.com) The center of gravity in enterprise artificial intelligence is moving from “best model” to “best operating model.” The winners will be the vendors that can make artificial intelligence cheaper to run, easier to govern, and simple to fit into existing work. (futurumgroup.com, openai.com)