As of 13 August 2026, AI can only partly design a better sign-up flow for your service.

What to actually do

1. Hand it to a person

   The route this page recommends

   A person who owns the outcome does this end to end, worth it when the failure is dear.

2. Use a tool built for this

   Second choice

3. Do it yourself

   The distant third

   A chat interface, chat-fluent skill, and roughly 1 hour until you can act on the result.

   How to actually do it

   1. Open your current sign-up journey, product requirements, analytics, support tickets and user research, then gather the existing screens, required fields, eligibility rules, account states and known abandonment points.
   2. Paste those materials into the prompt, removing passwords, payment details and unnecessary personal data, and mark unknown facts as unknown rather than asking the model to fill them in.
   3. Ask the model to produce the proposed flow, alternatives, wireframe specification, copy, assumptions and test plan in the requested order.
   4. Build the selected flow as a clickable prototype in Relume or your existing design tool, including every branch, validation message, error state and recovery route.
   5. Compare each drafted field, rule, label and consent step against your current product rules, technical limits, privacy requirements and accessibility standard, then correct anything the model assumed.
   6. Ask a colleague who understands the service to try the prototype without explanation, record where they hesitate or fail, and compare their behaviour with the test hypotheses.
   7. Run the revised flow with representative users, compare the results with the existing journey where possible, and only publish the change after the product owner, developer, accessibility tester and privacy lead approve their parts.

   Prompt

   Design a sign-up flow for [service] aimed at [target users] in the UK.
   
   Use the information below and do not invent product rules, user research findings, legal requirements or technical capabilities.
   
   Service and value:
   [describe what the service does and why someone signs up]
   
   Current sign-up flow:
   [list every screen, field, decision and error state, or say that there is no existing flow]
   
   Known user evidence:
   [research findings, analytics, support tickets or user quotes; label anything unknown]
   
   Business constraints:
   [required data, eligibility rules, account types, verification steps, payment requirements and success measure]
   
   Technical and operational constraints:
   [platform, authentication options, integrations, accessibility requirements, languages and support capacity]
   
   Privacy and security constraints:
   [what data must be collected, retention rules, consent requirements and any sensitive data involved]
   
   Produce:
   1. The proposed end-to-end flow as numbered steps, including branches, validation, errors, recovery and the point at which the account becomes usable.
   2. A screen-by-screen wireframe specification with purpose, fields, labels, primary action, secondary action, help text, error states and accessibility notes.
   3. Three materially different alternatives, explaining the trade-off of each.
   4. Draft plain-English interface copy in UK English. Do not make claims about compliance unless they are supplied above.
   5. A list of assumptions and unanswered questions, separated from facts.
   6. A test plan covering task success, abandonment, accessibility and comprehension, with hypotheses and measures but no invented targets.
   7. A launch checklist stating what must be confirmed by the product owner, designer, developer, accessibility tester and privacy lead.
   
   Prefer the smallest amount of information needed at each step. Do not call the flow better unless the test evidence supports that conclusion.

   Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.