Find the signal

Filter AI noise through the work that actually needs to move.

There is always another model, feature, demo, or executive request.

The pressure is not only to learn it. It is to decide whether it changes the work already in motion.

Without that filter, designers spend more time reorganising their AI workflow than improving the experience in front of them.

See it

Relay's designer receives three things on Monday: a request to “add an AI assistant,” a new prototyping feature, and a thread about autonomous agents.

What went in: Three developments with no shared customer problem.

What AI produced: A list of use cases broad enough to justify all of them.

What the designer decided: Only the prototyping feature could answer a current question: whether a blocked handoff should interrupt the owner or wait in a queue. The other two ideas went into a watch list with conditions for revisiting them.

The signal was not the most impressive technology. It was the change that could improve a decision this week.

Why it works

  • New capability is not the same as a useful product opportunity.
  • Current work supplies a stable centre of gravity when tools keep changing.
  • A bounded experiment produces evidence; broad exploration mostly produces more possibilities.
  • Saying “not now” preserves attention without pretending the change does not matter.

Try it

Try it · 5 min

Run the work test

Take up to five AI developments or requests competing for your attention. For each, write the current design problem it changes, the customer or team value it could create, the smallest test, and the condition that would make you ignore it for now. Choose one test.

KeepA signal list with one experiment to run now, a short watch list, and reasons for leaving the rest alone.

Check it

  • Does the chosen experiment begin with a real problem rather than a tool?
  • Will it improve a decision, an experience, or a useful work product?
  • Can you learn something within one piece of current work?
  • Have you made the cost of distraction visible?
Human checkpoint

Which change deserves one hour of your attention this week?

Compare its connection to current work, likely customer value, learning speed, reversibility, and the opportunity cost of pausing something else.

Keep it

AI SIGNAL FILTER

CHANGE OR REQUEST
-

THE WORK IT CHANGES
-

VALUE IF IT WORKS
- Customer:
- Business:
- Design practice:

SMALLEST TEST
-

DECISION
- Try now / Watch / Ignore for now

REVISIT WHEN
-

Before you use this

Do not use urgency from a launch or an executive message as evidence of customer value. The designer takes over when choosing what current work to interrupt and what result would justify wider adoption.

Can you use it?

Question: What turns an AI development into signal?

Answer: A clear connection to real work, meaningful value, and a small test that can change a decision.

Next

Once you can see the signal, decide what design is prepared to stand for.