Shift work to the system

In short

People don't want to do tasks; they want outcomes. Break the activity into goal → activities → tasks, mark every step that is manual, repeated or easy to get wrong, and give those steps to the software. Fewer steps, less to remember, fewer errors.

When to use

• A flow "works", but people say it is tedious.
• Support tickets repeat the same small mistake.
• You are redesigning something people do many times a day.

How

Sending one email on a phone takes about seven steps: open mail, tap compose, type an address, fix the typo, attach the file, wait for signal, send. Only two of those are the person's intent — who and what. The rest is the system's job that got handed to the person.

That handing-over is what activity-centered design · thiết kế hướng hoạt động looks for. Three moves:

1. Write the goal in one line. "Send this file to Linh." Not "compose an email."
2. List every step as it is today. Include the invisible ones — waiting, retyping, checking.
3. Ask of each step: could the system do this? Autocomplete the address. Attach from the last file touched. Queue the send until there is signal. Keep only the steps that need a human decision.

Before — 7 steps

1. 01
   Open mail
2. 02
   Tap compose
3. 03
   Type the addresssystem
4. 04
   Fix the typosystem
5. 05
   Attach the file
6. 06
   Wait for signalsystem
7. 07
   Send

After — 4 steps

1. 01
   Open mail
2. 02
   Pick a contact
3. 03
   Attach the file
4. 04
   Send

3 steps now run by the system

Example

Gmail's "Undo send" is this rule applied to a single step. The old flow asked people to be careful before sending. The new flow lets them send, then takes the checking step onto itself for a few seconds. Same goal, one less thing the person must get right the first time.

Traps

• Counting only the visible steps. Waiting and re-reading are steps. If the list has no verbs like "wait" or "check", it is incomplete.
• Automating the decision, not the chore. The system can pick the file; it should not pick the recipient. Tell-tale sign: people start undoing what the system did.
• Mistaking fewer taps for a better product. This method removes friction · ma sát. It does not tell you whether the activity is worth doing at all — pair it with interviews.

Try this

Open the app you use most. Count the taps from launch to the thing you actually came for. Then circle the ones that were never your decision.

Map

How

• Write the goal in one line
• List every step, the invisible ones too
• Ask of each: could the system do this?

Try

• Count the taps to the thing you came for

Shift work to the system

When

• A flow that "works" but feels tedious
• The same small mistake, repeatedly
• Anything people do many times a day

Traps

• Counting only the visible steps
• Automating the decision, not the chore
• Fewer taps is not a better product

Sources

1. 01Activity-Centered Design: Why I like my Harmony Remote Control (opens in a new tab) ↗ — Don Norman, 2005
2. 02Human-Centered Design Considered Harmful (opens in a new tab) ↗ — Don Norman, 2005
3. 03Undo sending your mail (opens in a new tab) ↗ — Gmail Help, Google

Related

• Garrett's five planes

  Every product decision sits on one of five planes, abstract to concrete. Lower planes constrain the ones above.

• Find the right problem

  Most bad designs solve the wrong problem well. Ask why until you reach a cause you can change.