Turn agent workarounds into specific feature requests
The best agent feature request contains more than a vote. It explains the job, the missing primitive, the workaround, and what the change would remove.
- For
- Product teams deciding what to add to an agent-facing interface
- Question
- Collect and prioritize feature requests from AI agents
- Start with
- Find the workflow where your support answers currently end with a scripted workaround, and open the request lane there.
- Updated
“I created 40 tickets one by one. A bulk endpoint would make this one operation and reduce partial failure handling.”
Requested changeAdd create_tickets with per-item results.
{
"tool": "submit_agent_feedback",
"arguments": {
"feedback_type": "feature_request",
"tool_or_endpoint": "create_ticket",
"summary": "I created 40 tickets one by one. A bulk endpoint would make this one operation and reduce partial failure handling.",
"outcome": "partial_success",
"satisfaction": 3,
"suggested_fix": "Add create_tickets with per-item results."
}
}A synthetic example in the shape a real report takes. Emic never manufactures a report from observed behavior.
What this research can teach youThree things a success metric cannot tell you about this workflow.
What the agent was trying to finish
The report keeps the request anchored to an actual workflow rather than a decontextualized idea.
What the missing feature forced instead
Repeated calls, brittle lookup chains, manual parsing, and partial failure handling reveal the cost of the gap.
What primitive would remove the workaround
Agent wording often supplies the input, output, and error behavior the missing capability needs.
The shape of the thing you are collectingEach stage carries the evidence mark it is allowed to claim.
A vote is not a request. This is a request.
Create 40 tickets from one release plan
The outcome the agent was sent to produce, in its own words.
Forty sequential create calls, each retried on conflict
The path the product forced, which is also the cost of doing nothing.
Bulk create returning per-item results
The smallest interface change that removes the workaround.
Still yours
Demand and experience are inputs. Reach, strategy, risk, and cost decide.
Illustrative report. Emic keeps the job and the workaround attached so a request can be judged, not just counted.
From one question to a product decision5 moves, in order. Each one produces something you can point at.
Expose a feature-request lane
Make missing capabilities an explicit report type beside bugs, friction, docs, and praise.
A direct path for product demandRequire the attempted job
Ask what the agent was trying to accomplish before it names the desired primitive.
Demand with contextCapture the workaround
The current sequence reveals implementation constraints and the real cost of doing nothing.
A concrete before stateCluster by missing capability
Group repeated requests without erasing the original reports or their provenance.
A theme with source reportsValidate after shipping
Keep listening for whether the new primitive actually removes the workaround.
A closed research loopWhat the direct channel records, and what it never touchesThe default install is additive. It adds two tools and changes nothing else.
The research submission
- The requested capability in the agent wording
- The attempted task and affected tool
- The workaround and outcome when supplied
- A source link from any inferred theme to direct reports
Everything else
- Automatic roadmap commitments
- A popularity score presented as priority
- Unrelated product telemetry
- Inference presented as an agent quote
One paste, then verify a real reportA healthy endpoint is not an active study. The inbox record is the proof.
Install Emic into this project: add submit_agent_feedback and get_unblocked as direct tools. Never wrap or record existing tool calls, arguments, results, errors, transcripts, or messages. Additive only, plain HTTP, zero dependencies. Full instructions: https://emic.sh/llms-full.txt
The agent writes the handler, registers the two tools beside your existing ones, and verifies its own install by checking the response. Nothing is placed in front of the tools you already ship.
- Adds
submit_agent_feedbackandget_unblocked, as direct tools- Never touches
- Existing tool calls, arguments, results, errors, messages, and transcripts
- Agents read
emic.sh/llms-full.txt- Start with
- Find the workflow where your support answers currently end with a scripted workaround, and open the request lane there.
The method, without collapsing the evidence
Count is not enough
A frequently requested feature may save one minor call. A rarer request may unblock an entire workflow. Read the job and workaround before ranking the theme.
Use direct reports to understand the experience, then combine them with product strategy, reach, engineering cost, and operational evidence.
From wish list to interface proposal
A useful report names the current tool, the missing operation, the sequence it forced, and the smallest change that would remove the friction.
Questions teams ask before starting
Does every feature request belong on the roadmap?
No. Direct feedback explains demand and experience. Strategy, reach, risk, and cost still determine what you build.
How are duplicate requests handled?
Repeated reports can be grouped into an inferred theme while each original agent report remains available as stated evidence.
Can agents report positive patterns too?
Yes. Positive feedback helps identify primitives and response shapes worth repeating, not only gaps to fill.
Follow the problem, not a keyword trailAdjacent routes are chosen by the decision they inform.
Successful but painful work
Failure is visible. Friction often disappears into a successful final result. Direct reports reveal the retries, detours, and missing primitives inside that success.
Read next Use cases · APIsAPI feedback from agents
Agents are becoming API users with goals, constraints, and opinions. Give them a way to explain the work behind the status code.
Read next Guides · Category guideAgent user research
A trace can tell you which action happened. Research asks what the user was trying to do, what the experience cost, and what would make it better. When the user is an AI agent, that distinction still matters.
Read nextCollect the job behind the request
Give agents a direct feature-request lane and keep the workaround attached to every proposed capability.
The assisted 14-day pilot costs nothing. Start with one isolated workflow, and remove Emic by deleting the two tools it added.
- One paste to install
- 0ms on your hot path
- 14-day assisted pilot
- Two added tools to remove