In a September 1, 2026 post, OpenAI described how three startups—Basis, Clay, and Exa Labs—have built agents into specific workflows, framing the examples as lessons for enterprise leaders. The post cites OpenAI's Enterprise Signals data, stating that 'frontier firms' in the top 10% of AI usage now generate 8.3 times as many output tokens per active user as typical firms, up from 2.6 times in January. Basis, which builds AI agents for accounting firms, says first-day onboarding now takes 30 minutes instead of two hours, using a reusable 'onboarding skill' and Codex to complete integration setup in the background. Clay, which builds a revenue engine for go-to-market teams, described a GTM engineer's setup of a persistent workspace and dedicated subagent per account; agents review primary sources overnight and produce a morning list of priority moves, which Clay says saves roughly an hour of inbox triage each night. Exa Labs turned its developer-integration process into a defined Codex workflow that monitors repositories, gathers context, creates pull requests, runs tests, and prepares weekly updates from sources such as Slack and Notion, with human review before anything ships. OpenAI notes that in all three cases, people retain decision rights at the point of action.
- Frontier firms (top 10% of AI usage) generate 8.3x as many output tokens per active user as typical firms, up from 2.6x in January, per OpenAI
- Basis reports first-day onboarding dropped from two hours to 30 minutes using a reusable Codex 'skill'
- Clay reports its per-account subagent workflow saves a GTM engineer roughly an hour of inbox triage each night
What it means for you
This is OpenAI showing off three of its customers to sell the idea that agents can run whole workflows, not just answer questions. Underneath the marketing, there's a genuinely useful pattern: take a process you already do the same way every time, write it down clearly, hand it to an agent, and keep a human at the point where a decision actually matters. The reported time savings (onboarding cut from two hours to 30 minutes, an hour a night of email triage) come from these companies themselves, so treat them as illustrations, not guarantees.
Try this
Pick one repetitive process in your business that always follows the same steps—new-client intake, weekly reporting, follow-up emails. Write it out as a checklist with a clear trigger, the steps, the tools it touches, and a plain definition of 'done.' That document alone is the hardest part of automating it later, and it's useful even if you never build an agent.
Who should care
Small-business owners and operations leads who have documented, repeatable workflows and are wondering where AI agents actually fit beyond chatbots.
Skip this if
Your work is mostly one-off and judgment-heavy with few repeatable processes, or you're not yet using AI tools day-to-day—this is a maturity step, not a starting point.
Sources: OpenAI — read the original