OpenAI published a customer story on September 25, 2026, describing how Proaction, a company that builds software for businesses managing vehicle fleets, uses Codex and other OpenAI models. According to the account, co-founder and COO Colin Knudsen, who describes himself as non-technical, builds four to six customized interactive HTML demos a month using Codex, each taking 30 to 45 minutes. Knudsen estimates this avoids 40 to 60 hours of engineering work monthly, based on his figure of roughly 10 hours per demo if produced by engineers. Knudsen states the share of deals moving from initial contact into solution development has increased by 50% to 60% with the custom demos. He also uses Codex plugins for tools including Granola, Gmail, Slack, Linear, GitHub, and HubSpot to prepare follow-ups, create issues, and update opportunities, estimating this saves him 25 to 33 hours a month. Proaction additionally uses ChatGPT-5.6 Sol to identify vehicle damage from submitted photos, and is building voice agents on GPT-Live-1 and GPT-6 Astra for what it calls a Managed Execution Layer. One agent, Marty, coordinates vehicle maintenance. All figures are estimates attributed to Proaction staff and reported by OpenAI, which markets the products described.
- Knudsen builds 4-6 custom demos monthly in Codex, each taking 30-45 minutes, estimating 40-60 engineering hours saved per month
- Proaction reports a 50-60% increase in deals moving from initial contact to solution development with custom demos
- Knudsen estimates Codex saves him 25-33 hours a month across ~15-20 daily tasks; published by OpenAI on September 25, 2026
What it means for you
A vendor-published story showing a non-technical founder using an AI coding tool to build custom sales demos himself instead of pulling engineers off other work. The specific numbers are self-reported estimates from the company being featured, so treat them as illustrative rather than proof. The genuinely interesting part is the pattern: connecting an AI tool to your existing apps (email, calls, CRM) so it can gather context and do the next step for you.
Try this
Pick one repetitive task you do across two or three tools this week — for example, turning a sales call recording into a follow-up email and a CRM update — and try walking an AI assistant through it once, feeding it the transcript and prior emails as context.
Who should care
Small-business founders and salespeople who currently rely on engineers for custom demos, and anyone spending hours copying information between tools like email, Slack, and a CRM.
Skip this if
You don't do custom demos or heavy multi-tool workflows, or you're allergic to vendor case studies where every metric is supplied by the happy customer.
Sources: OpenAI — read the original