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

On September 25, 2026, OpenAI published a customer story about Proaction, a startup that builds fleet management software. The headline numbers are simple. With Codex, Proaction saves 40 to 60 engineering hours per month, saves its founders about 33 hours per month, and reports a 60% increase in sales. Proaction is based in North America. It sells software to businesses that manage fleets of vehicles, from cars and trucks to construction machinery. The company uses Codex and the API, and its product relies on GPT-Live-1, GPT-6 Astra and ChatGPT-5.6 Sol. Start with the problem. Every fleet operates differently. The mix of vehicles, the workflows and the habits of each customer vary. So showing a prospect how the platform fits their business is an essential part of selling it. But personalized demos need engineering time, and a small startup does not have spare engineers. Before Codex, the founders relied on conversations and slide decks to explain what was possible. Colin Knudsen, co-founder and COO, is not an engineer. Each time he wanted a demo, he had to loop engineers in. Codex changed that workflow. After a sales call, Colin points Codex at the Granola recording, the prospect's email threads and any spreadsheets the prospect has shared. Codex uses that context to customize an HTML demo environment. The environment mirrors Proaction's product, but it is filled with the prospect's own fleet. When Colin shares his screen, the prospect sees their own cars, trucks or equipment, organized around how they work. They can point to what needs adjusting and help build the solution. In Colin's words, the two sides generate the end solution together, without engineering involved at all.

Now the numbers. Colin builds four to six customized, interactive demos a month. Each takes 30 to 45 minutes. He estimates that an engineer would need about 10 hours for a comparable demo, so the company avoids 40 to 60 hours of engineering work every month. On conversion, he estimates that the share of deals moving from first contact into solution development, rather than nurture, has risen by 50% to 60% with the custom demos. A caution is needed here. These are Colin's own estimates. They are not the result of a controlled experiment. Treat them as an informed operator's judgment, not as a benchmark. The value of the demo continues after the sale. When a prospect becomes a customer, Colin hands the customized demo to engineers as a visual reference. That reduces questions and back-and-forth about what to build. Proaction also used Codex to build a customer solution center. Prospects can log in, explore workflows tailored to their business and review sales materials. This helps customers explain what they need. It also gives non-engineering teammates a way to turn those conversations into clearer requirements. By the time engineers get involved, they have a concrete picture of what to build. The second thread is Codex as a daily workspace. Colin's work spans sales, customer support and product management. Through Codex plugins for tools including Granola, Gmail, Slack, Linear, GitHub and HubSpot, he brings customer context together and acts on it in one place. He pulls call transcripts and email history to prepare follow-ups, creates Linear issues and updates HubSpot opportunities. He also set up a scheduled automation that reviews recent calls and prepares sales updates for the team. Before, a heavy workload meant jumping between tabs and copying information from one tool to another. Now he describes what he needs, and Codex gathers the context and executes the next step. He estimates 15 to 20 distinct tasks a day, and believes Codex saves him 25 to 33 hours a month. He says that everything he does is centered around working in Codex.

The third thread sits inside the product. Proaction uses OpenAI models across its platform. When customers submit photos with vehicle issue reports, ChatGPT-5.6 Sol helps identify damage. With GPT-Live-1, the company is building agents that handle more of the day-to-day work of running a fleet. It calls this its Managed Execution Layer. Colin says the goal is for Proaction to execute work for its customers, beyond helping them manage and track it, and that advances in OpenAI voice are a big reason this is possible. Customers can ask specialized agents to handle things like tolls or service, or set up workflows that put the right agent to work automatically. The agents use OpenAI models, including GPT-Live-1 and GPT-6 Astra, to make voice calls and to review documents and images. The source text available to us is cut off after this point, so we do not describe further details. What is the mechanism? The key idea is that context acts as the interface. Codex does not invent a demo from nothing. It is pointed at real material, such as recordings, emails and spreadsheets, and it produces an interactive HTML file. An HTML demo is light. It needs no backend deployment, and it is easy to change live during a screen share. The approach merges requirements gathering and prototyping into one conversation. It shortens the distance between the customer's words and a visible product. There are three lessons for enterprises and developers. First, users of coding agents are moving beyond engineers. When a non-technical founder builds the prototype, scarce engineering time goes to real product work. Second, the value comes from workflow integration, not from single outputs. Plugins connect the CRM, the issue tracker, email and the code repository, and what they save is switching cost. Third, voice agents move software from a system of record toward a system of execution. This matters most in industries that run on phone calls, such as fleets, logistics and maintenance.

Some caution is also in order. The gains are mostly self-reported estimates from one early-stage company and one heavy user. Letting agents place calls or handle tolls and service for customers raises questions about accountability, error correction, compliance and audit trails. The source does not describe how Proaction handles them. Building demos from customer data also requires clear rules about data access and privacy. Any enterprise that copies this pattern must answer these questions first. Looking ahead, similar patterns will likely spread to other verticals. Sales staff will build custom prototypes with coding agents. Operations staff will hand repetitive outreach to voice agents. For OpenAI, the story shows its products working together: Codex as the workspace, and the API and voice models as the parts embedded in a product. For Proaction, the real test is whether the Managed Execution Layer can run reliably and in an auditable way in the messy conditions of real customers. If it can, competition in this software category will shift from the number of features to the amount of work actually completed.