Chat-or-Click Delegation Rule
Delegate parallel or context-heavy browsing and click through simple tasks yourself
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 4
- Confidence
- 96%
The Chat-or-Click Delegation Rule separates tasks that benefit from a conversational agent from those better served by direct interaction. Manual clicking remains appropriate when an action is simple, immediate, and faster to complete than explain. Agent delegation becomes valuable when the task has many steps, can run in the background, spans multiple tabs, or benefits from accumulated context. The core metric is not whether the agent moves faster than a person; it is whether delegation reduces attention and permits useful parallel work. Users should define the goal, let the agent execute, and then verify the result. This avoids converting every interface into chat while still capturing the leverage of autonomous browser execution.
Origin
Extracted from Marketing Against The Grain during a discussion of when Atlas's chat interface improves browser work.
Core principles
- 01Use direct interaction when it is faster than describing the task
- 02Use agents when work can continue unattended or in parallel
- 03Favor delegation for multi-step, repetitive, or context-heavy tasks
- 04Judge automation by total attention saved, not raw execution speed
How to run it
- 1
Estimate manual effort
Assess the time, attention, and number of actions required to finish the task yourself.
Pro tip Include context switching and data transfer, not just clicking time.
Watch out Do not automate merely because automation is available.
- 2
Test for delegation leverage
Look for repetition, multiple tabs, background execution, or extensive context that makes agent work advantageous.
Pro tip Tasks that can run while you do something else have higher leverage.
Watch out Sensitive or irreversible actions require closer control.
- 3
Define the finish line
Tell the agent the desired output, constraints, and conditions that indicate completion.
Pro tip Ask for a summary of actions and unresolved uncertainties.
Watch out Vague delegation makes verification difficult.
- 4
Verify and refine
Inspect the result and improve the instruction if the same task will recur.
Pro tip Save proven instructions for repetitive work.
Watch out Do not assume successful navigation means the underlying result is correct.
In the wild
A user delegates a statewide Zillow search for inexpensive land suitable for campsites, then continues other work while the browser explores listings. Manually opening one known listing may be faster, but searching and screening many candidates creates enough parallel leverage to justify delegation.
→ The agent returns candidate properties while the user avoids sustained browsing attention.
Common mistakes
Automating trivial clicks
Describing and verifying a tiny task can take longer than completing it manually.
Ignoring attention savings
An agent can still be valuable when it is slower if it allows several tasks to proceed in parallel.
Is it for you?
Best for
Knowledge workers choosing whether to delegate a browser task to an AI agent.
Not ideal for
Tiny one-step actions that are faster and safer to complete directly.
From the transcript
“There is gonna have to be the when chat is a good UX experience versus when just clicking it yourself as a good UX experience.”
“Like I don't think everything gets instrumented into a chat UX, but to your point, where I think this is valuable is when you can…”
“Now it's not wicked fast. Like you could argue maybe I could have found it just as fast or faster myself.”
From the episode
OpenAI Just Launched a Web Browser (and It’s Smarter Than Chrome)