Proaction boosts sales 60% and saves 75+ hours with Codex

Proaction builds software for fleets of cars, trucks, and construction equipment. Personalized demos used to need engineers the team did not have, so founders sold with conversations and slides.

With Codex, co-founder Colin Knudsen says he builds four to six customized HTML demos a month in 30–45 minutes each, sparing an estimated 40–60 engineering hours monthly. He puts the lift in deals moving from first contact into solution development at 50–60%.

That is the quiet story under the model race: useful AI showing up in the sales loop and the ops layer, not just the changelog.

We’re watching Codex leave the IDE and sit in the founder’s day job. Knudsen feeds Granola recordings, emails, and prospect spreadsheets into Codex and gets a demo that mirrors the customer’s own vehicles and workflows. Prospects see their fleet on the screen. Engineers later inherit that demo as a visual brief.

He also runs Codex as a hub across Granola, Gmail, Slack, Linear, GitHub, and HubSpot—estimating 25–33 hours saved a month across 15–20 daily tasks. That is the benefit of the doubt useful AI earns when the work is concrete.

The sharper turn is the Managed Execution Layer: agents on GPT-Live-1 and GPT-6 Astra that make voice calls, read documents and images, and handle tolls or service. One agent, Marty, talks to a driver, calls shops, arranges service, and helps get an estimate approved. Humans step in for review.

If this scales, fleet software stops being a dashboard and becomes a dispatcher. The upside is real time back for managers who are drowning in coordination. The downside is familiar: voice agents calling vendors with imperfect judgment, and more customer data sitting inside the model stack. We’re for the productivity win. We’re also watching how tightly Proaction keeps a human in the loop when Marty starts dialing.

OpenAI published the case study on September 25, 2026, as part of its startup customer series. Proaction is a North American software startup in fleet management.