Key Takeaways

  • Ramp treats internal employee inquiries as programmatic API calls rather than manual interruptions for product managers.
  • Their internal AI agent, Gadget, automatically answers 85% of inbound questions directed at PMs, indexing company knowledge to draft launch collateral and status updates.
  • When Gadget cannot answer a query, a human steps in, and the resolution feeds back into the model to close knowledge gaps permanently.
  • Autonomous loops resolve 60% of reported micro-UX issues within 24 hours without engineers triage-tagging tickets.

The Method

Fast teams run into a wall that code generation cannot fix: coordination drag. When engineers ship ten updates a week, sales reps, customer support agents, and designers flood product channels with questions. Who owns this feature? Why did this button move? When is the rollout reaching enterprise clients?

Geoff Charles saw this tax stall Ramp. “Too much process slows down the builders,” Charles says. “Too little process and everyone is confused. And so human attention becomes that bottleneck. How do you solve this? Well, every question is an API.”

To build this API layer, Ramp deployed an internal tool called Gadget. The system relies on a two-part operational loop.

First, make the company legible to software. “The principle here is simple,” Charles explains. “For you to empower an agent, the agent needs to be able to understand and read the organization. And the organization needs to be legible to your agents.” Ramp structured PRDs, release notes, Slack channels, and code repositories so that Gadget can query live project states instantly. Today, “85% of questions that are being asked to PMs now are fully answered with AI and those that are not are answered and fed back into the system.”

Second, automate small reactive fixes end-to-end. Product teams waste hours triaging tiny UI bugs: misaligned text, confusing tooltips, broken padding. Instead of filing tickets that sit in backlogs for three sprints, Ramp created autonomous resolution loops. “You need to automate your way out of these small loops,” Charles says. “For most small things, an AI fully runs the loop.”

When a salesperson or customer flags a minor interface flaw, the system diagnoses the issue, opens a pull request, runs tests, and preps the fix. “Thanks to these autonomous loops, 60% of UX issues that are identified by either a customer or a salesperson or a CX person or ourselves are fixed within 24 hours.”

Where This Breaks Down

This architecture fails if your company runs on tribal knowledge trapped in private direct messages. If decisions happen in 1-on-1 calls without written summaries, an agent like Gadget has nothing accurate to parse. It will hallucinate feature statuses and mislead your sales team.

Automated bug fixes also create risk in regulated flows. A micro-UX tweak in a checkout funnel or ledger balance display can alter compliance logic. You cannot let an autonomous loop push code to production without strict sandboxes and test suites.

What to Do With This

Audit your product team's Slack channels from the past two weeks. Categorize every inbound message sent to your PMs into three buckets: status checks, documentation requests, and bug reports. If more than half are simple lookups, create a dedicated public knowledge channel and connect a search agent to your documentation hub before hiring your next product manager.