Second-Screen AI Practice
Build AI fluency by routing everyday digital work through a dedicated screen
- Difficulty
- Easy
- Time to result
- ~weeks to results
- Steps
- 5
- Confidence
- 98%
Galloway's concrete advice for career resilience is to keep a second screen or persistent window devoted to AI and route suitable digital work through it. The mechanism is environmental and experiential: constant visibility prompts repeated use, while real tasks teach where models save time, where prompting matters, and where judgment is still required. He applies the idea to contract review by asking one model to analyse and redline an agreement, then repeating the exercise with another model. The reusable practice is broader than that example: submit appropriate inputs, test a useful transformation, compare outputs, verify the result, and retain what works. Galloway says he has to force himself to use new technologies, making deliberate exposure central to the method rather than assuming fluency appears automatically.
Origin
Extracted from The Diary of a CEO
Core principles
- 01AI fluency grows through repeated use on real work
- 02A visible tool is easier to remember than a hidden one
- 03Small experiments reveal where AI helps and where it fails
- 04Human judgment remains responsible for the output
- 05Skill updates are part of staying employable
How to run it
- 1
Make AI visible
Keep an approved AI tool open on a dedicated screen or persistent workspace. Reduce the friction between receiving digital work and testing AI on it.
Pro tip Use a pinned window if a physical second screen is unavailable.
Watch out Visibility does not override data-handling or confidentiality rules.
- 2
Choose a real input
Select a suitable document, question, draft, or process from current work. Define the result you need before prompting.
Pro tip Begin with reversible, low-risk work where comparison is easy.
Watch out Do not upload private material to an unapproved service.
- 3
Run a focused task
Ask the model to analyse, redraft, classify, or challenge the input against explicit criteria. Iterate when the first result misses the requirement.
Pro tip Include the role, task, constraints, and desired output format.
Watch out A confident response can still be incorrect.
- 4
Cross-check consequential work
For important work, compare another model, source, or qualified reviewer with the first output. Resolve disagreements instead of averaging them blindly.
Pro tip Use an independent check that did not generate the original answer.
Watch out Two models agreeing is not equivalent to expert approval.
- 5
Keep the useful pattern
Record prompts, checks, and task types that reliably help. Reuse the pattern while continuing to verify each output.
Pro tip Save the workflow, not sensitive source content.
Watch out Re-test patterns as tools and job requirements change.
In the wild
Galloway describes asking a project owner to give an agreement and the intended terms to an LLM, request a review and redline, then perform the same exercise with another model. He presents this as a way to reduce some routine external legal review, not as a claim that qualified lawyers are always unnecessary.
→ The team obtains a first-pass comparison before deciding what professional review remains necessary.
Common mistakes
Using AI only for demonstrations
Occasional novelty use does not build the repeated judgment required for practical fluency.
Skipping verification
Routing work through AI is a learning method, not permission to trust outputs without checking them.
Exposing restricted data
Convenience can create confidentiality or compliance failures when tools are not approved for the material.
Is it for you?
Best for
Knowledge workers who need practical AI fluency but do not yet use models consistently in their workflow.
Not ideal for
It is unsuitable for confidential, regulated, or high-stakes material unless approved tools and independent review are in place.
From the transcript
“Always have a second screen open that has nothing but AI on it, LLMs.”
“Anything you get digitally, port it into your AI screen and start playing with it.”
From the episode
Scott Galloway: AI Wasn’t Built For You. The Rich Don’t Need You Anymore!