Agents are always slightly wrong when first built. They require hundreds or thousands of conversations to get close to reliable. The people who know what is wrong are not the developers who built the agent. They are the end users who interact with it every day.
An insurance adjuster knows when the agent is handling a claim incorrectly. A salesperson knows when the outreach tone is off. A support lead knows when the agent is closing tickets too aggressively. None of them have access to the codebase. None of them want it.
If the only way to correct the agent is to file a ticket with engineering and wait for a code change, the iteration cycle is too slow to ever converge on quality. The platforms that win will let users correct agents the way they would correct a direct report — conversationally, immediately, without writing code. The correction has to take effect on the next run, not the next sprint.
This is also why the engineering agent platform is quietly becoming the enterprise agent platform. The primitives are the same. A sandbox for execution. Tool access via MCP or APIs. A prompt that defines the role. Model routing. A conversational refinement interface. These are what engineering agents need, and exactly what a sales agent or an insurance adjuster agent needs. The difference is cosmetic, not architectural.
Companies that constrain their platforms to engineering workflows are creating an opening. The mid-market CTO wants one tool, not five. I've argued separately that you need a directory of agents — the same directory has to be the place where the user files the correction, not a Jira queue staffed by overworked platform engineers.
Sources
Related Essays
You Need a Directory of Agents
Companies have directories of employees. They have no equivalent for the agents doing real work. Every agent should be inspectable, auditable, and correctable.
The IKEA Effect Is Real for Agent Tooling
People pay 63% more for furniture they assembled themselves. The same dynamic governs agent platforms — developers who build their own agents care about them and keep improving them.
The Coordination Crisis AI Tooling Created
When everyone can build, nobody knows what has already been built. Five teams independently spin up the same Slack bot, and the duplication goes undetected for months.
Key takeaways
- Agents are slightly wrong when first built. They need hundreds or thousands of conversations to converge on reliable behavior.
- The people who know what is wrong are the end users, not the developers. An insurance adjuster knows when a claim is being mishandled. A salesperson knows when the outreach tone is off.
- If the only way to correct an agent is a ticket to engineering, the loop is too slow to ever converge. Platforms that win let users correct agents conversationally, immediately, without writing code.
FAQ
Why is the user the right iteration loop?
Users have ground truth about whether the agent is right. Developers do not see the agent in context — they see logs and dashboards. The user sees the actual claim, customer, or outreach and can tell instantly whether something is off.
How does this change the platform design?
The platform needs a conversational refinement interface where users can correct the agent the same way they would correct a direct report. That makes the engineering agent platform and the enterprise agent platform structurally identical.