Users default to chat and never find the structured mode

If users mostly choose chat while structured workflows receive little use, compare the tasks and outcomes in each mode.

Investigate whether the structured path offers relevant value, is discoverable, accepts the required inputs and fits the user’s task. A structured interface does not by itself produce better output.

Summarize with:

Key facts

  • Most activity happens in the free-form input.
  • Compare output quality and completion cost for the same tasks before deciding which mode is better.
  • General assistants have made an empty text box the default way to ask software for something.
  • Choosing a mode requires knowing which one fits a problem the user has not yet described.
  • Mode names taken from internal architecture or job titles make users map their situation onto the product’s categories.

Why it happens

Chat is familiar. General assistants have made an empty text box the default way to ask software for something. It requires no decision up front.

Workflows ask for a choice first. Choosing a mode requires knowing which one fits a problem the user has not yet described. Typing the problem first feels safer.

Mode names describe the system. Names taken from internal architecture or job titles make users map their situation onto the product’s categories.

Design consideration
Boundaries are invisible

Users cannot see what a workflow will do, what it will refuse or how long it takes. Chat has no visible boundary to misjudge, so it wins by default.

How it shows up

  • Most activity happens in the free-form input
  • Users describe the product as an AI chat
  • The mode with the best output has the lowest use
  • Interview users about their choice of mode and compare it with observed tasks
  • Sales demos show the workflow, and recorded sessions show customers skipping it

Options and trade-offs

What helps:

  • Detect when a chat request matches a workflow and offer the switch inside the conversation
  • Name modes after the question they answer
  • Show the structure of the output before the user commits
  • Show the quality difference once, on the user’s own input

The tension: routing from chat into a workflow adds a step to the fastest path, and some users will resent it. The step is justified only if the workflow output is clearly better.

Common mistakes

  • Removing chat to force users into workflows
  • Naming modes after internal components
  • Hiding the structure of a workflow’s output until it runs
  • Demonstrating workflows in sales without checking how customers use them

Where we’ve worked on this

5.0
Overall review rating on Clutch.co
How can we help you with your project?
Let’s talk