Pattern 01Every answer cites its source
Problem. Teams will not act on an AI answer they cannot check.
Pattern. Each finding links to the exact page and line, one click away.
Inspiration · 13 patterns
Patterns that makeAI products trusted.
Small product decisions decide whether people use an AI product or work around it. These are the ones we keep coming back to, each taken from something we shipped. Open any to see it working.
Pattern 01Pattern 02“Not in the document”
Problem. Models would rather guess than admit they don't know.
Pattern. An explicit, honest answer for questions the source does not cover.
Seen inSyphon AI →
Pattern 03The human signs, then it locks
Problem. Medical and legal outputs must be exactly what a person approved.
Pattern. Drafts stay drafts until signed. After signing, nothing changes.
Pattern 04“Not sure” is a real answer
Problem. Forced yes-or-no questions turn doubt into a quiet No.
Pattern. A third answer that becomes a follow-up task for the team.
Seen inPreOp Screen →
Pattern 05Safe with the AI switched off
Problem. Safety that depends on a model is hard to defend.
Pattern. Critical checks are the expert's own rules. AI only helps with wording.
Seen inPreOp Screen →
Pattern 06Scores that explain themselves
Problem. A bare score can only be trusted, never checked.
Pattern. Every score comes with its reason in one line and the evidence behind it.
Pattern 07A delay you can cancel
Problem. Automation that acts in your name must never surprise you.
Pattern. Actions queue for 24 hours, and any one can be stopped before it goes.
Seen inWorkApply →
Pattern 08Remind before the date that matters
Problem. Everyone remembers the renewal date, not the last day to cancel.
Pattern. Pull both dates out and set reminders before the real deadline.
Seen inContract Blitz →
Pattern 09One style per user
Problem. Generic AI drafts get rewritten, and the time saved is lost.
Pattern. Each user's assistant learns from their own examples.
Seen inEsperwise →
Pattern 10Rank by what matters
Problem. The biggest customer is not always the most profitable one.
Pattern. Put costs next to revenue, then let people flip the ranking.
Seen inRetail Insights →
Pattern 11Agents draft, owners approve
Problem. Agents that act alone lose the owner's trust on the first mistake.
Pattern. Every plan arrives as a draft waiting for approval.
Pattern 12Swap providers by config
Problem. Per-minute platforms and single vendors lock the business in.
Pattern. The same agent runs on two realtime providers. Switching is a setting.
Seen inVeraVoice →
Pattern 13Many brands, one core
Problem. Local brands carry trust that a national brand would throw away.
Pattern. Each portal keeps its name and colours. The product underneath is shared.
Seen inRecruiting NOW →
Want the reasoningbehind them?
Common questions
What is an AI product design pattern?
A repeatable product decision that solves a common problem with AI, such as showing the source of every answer so people can check it in seconds.
Can I see the patterns working?
Yes. Each pattern links to the case study where it runs, and most case studies have an interactive demo.
Which pattern matters most?
Usually the human check: deciding where a person approves, signs or can cancel. Our human-in-the-loop guide goes deeper.