OpenAI

An agent that closes out every sales call

Cut post-call admin to a quick review: Salesforce tasks, follow-up emails and internal Slack updates are drafted from the call

Bryant McCombs Account Director, OpenAI LinkedIn

The problem

After every customer call an account director has to write notes, update Salesforce, draft follow-up emails and route context to their manager, billing or deal desk in Slack. Each step is small, but across 20 or 50 accounts it piles up, and Salesforce hygiene suffers.
What they built

An AI agent setup Bryant uses for his own selling, connected to Salesforce, Gmail, Slack and internal documentation. Its main skill takes a raw call transcript, works out which account and opportunity it belongs to, creates follow-up tasks, drafts follow-up emails and sends internal Slack messages. In one run it also attached a customer proposal built from the account's Salesforce data without being asked. A second skill reviews his priority emails each morning and drafts replies for him to approve. He drives most of it from his phone.
What changed

Bryant says his post-call work went from a series of manual steps to a few reviews and approvals each day.
So what

Salesforce gets updated and follow-ups get drafted after every call without relying on the rep's memory. The rep still approves anything a customer will see.

Under the hood

Codex drives the Salesforce CLI and is connected to Gmail, Slack and internal docs. The transcript-to-closeout skill runs for two to five minutes as one long task across those tools rather than a back-and-forth chat. A separate automation checks his Slack every three hours for anything urgent and drafts responses before notifying him. He uses Codex remotely from his phone with access to his computer.