In short

Customers ask for new features in support tickets, sales calls and passing emails, and the requests are lost or remembered only by the loudest voice. The improved process finds feature requests in those conversations, records them against the customer who asked, groups similar requests together and shows the product owner the full picture.

The system collects and groups. The product owner decides what is built, and any message to a customer about a request is approved by a person.

  • Sector Software and technology firms
  • Role Product owner
  • Job Recording who asked for what

The job today

Feature ideas are noted in a spreadsheet when someone remembers, or added to the development board with no link to who asked. Sales say a feature is essential, support say customers want something else, and the product owner has no clear way to tell. Customers who asked are never told when their request is delivered.

Where it breaks

  • Requests are scattered across tickets, calls and email.
  • There is no link between a request and the customers who asked.
  • Priorities are set by whoever argues hardest.
  • Customers are not told when the feature they asked for arrives.

The improved process, step by step

  1. Find requests in conversations AI preparesThe system reads new support tickets and saved call notes and picks out anything that looks like a feature request.
  2. Record who askedEach request is recorded against the customer and the conversation it came from, with the customer's own words.
  3. Group similar requests AI preparesRequests that describe the same need in different words are grouped, so the product owner sees one item with every customer behind it.
  4. Show the full pictureEach group shows the customers who asked and the context from each conversation. The product owner adds the commercial weighting.
  5. A person decides and approves updates A person approvesThe product owner decides what goes on the roadmap. When a feature ships, update messages to the customers who asked are drafted and approved by a person before sending.
  6. Close the loopDelivered and declined requests are marked as such, so the record stays useful.
Tracking feature requests at a small software company: process diagramSteps: 1. Find requests in conversations (AI prepares); 2. Record who asked; 3. Group similar requests (AI prepares); 4. Show the full picture; 5. A person decides and approves updates (a person approves); 6. Close the loop 01 AI PREPARES Find requests in conversations 02 Record who asked 03 AI PREPARES Group similar requests 04 RULES CHECK Show the full picture 05 A PERSON APPROVES A person decides and approves updates 06 Close the loop
The process at a glance. Teal steps are prepared with AI help; the amber step is where a person approves.

Where a person must approve

A person approves

The product owner decides what is built. Any message to a customer about a request is approved by a person before it is sent. Anything that sends, changes or cannot be undone waits for that approval.

The system never promises a feature or a date to a customer.

The data boundary

  • The system reads tickets and call notes chosen for this purpose. It cannot change the development board or the roadmap on its own.
  • Only the customer name, contact and their words about the request are recorded.
  • Customer data is not used to train AI models.
  • Every request, grouping and message is logged.

Checklist before you start

  • Our tickets and call notes are kept somewhere the system can read.
  • We have one place where feature requests should live.
  • We know who owns the roadmap and makes the decisions.
  • We know how we weigh requests, such as by customer size or strategy.
  • We are happy to tell customers when their request is delivered.
  • We have agreed what we will never promise in writing.

This is an illustrative walkthrough of a common pattern. It is not a client case study, and the details will differ in your organisation.