AI skills that outlast the next model release

The AI Product work group on weekly practice loops, written standards, delegation and checking whether AI improves the whole task.

Updated October 13, 2026

Practice on a task you already do every week. The skill that carries over is giving AI the right context, a bounded job and a clear standard, then judging what comes back.

Turn a weekly task into a workflow. Shishir Mehrotra suggests writing down the information you check, the judgment you make and what a good result must include, then letting an agent draft while you review, for a month. Mark Hull would run a weekly loop: state the user problem, draft with AI, test the result with someone who would use it and note where it failed.
Write your standards down. Sachin Rekhi gives Claude Code reference files with his frameworks for judging a strategy. He would have a PM pick one deliverable, write down the criteria for judging it and keep them in portable files for the next tool. Ami Vora would ask a colleague to find a case those criteria miss.
Make the goal concrete. Ami would write the AI goal into a performance review and protect time for it, the way she reserves an hour or two late on Friday afternoons for one growth goal. Aparna Chennapragada suggests three habits: use an approved tool on a recurring task, turn one task into a small prototype, and practice delegating a research brief and checking it.
Judge the whole task and refine the instructions. Mark would track rework, whether the result met the user's need and how long the review took, since a faster first draft counts only if the whole task improves. Shishir suggests describing intent instead of copying old rules, as when he replaced nearly 800 email filters with four instructions. Aparna would have a colleague try your AI helper without instructions and watch where they hesitate.

Group chat with Ami, Sachin, Aparna, Shishir, Mark — the AI Product work group

Question:

How can a working professional build AI skills that will last?

Shishir MehrotraCEO of Superhuman (formerly Grammarly). Co-founder of Coda and former Chief Product Officer of YouTube

I think the lasting skill is designing a workflow that uses AI well. Pick a task you do every week, such as preparing a customer update. Write down the information you check, the judgment you make, and what a good update must include. Then let an agent gather the context and draft a first pass while you review the facts and decide what to send. In an interview with MIT Sloan Management Review, I described a customer memo where sales, product, and support agents each bring relevant information into the document where someone is already working.

sloanreview.mit

Run that workflow for a month. Keep examples of where the agent helped and where your review caught an error, then change the instructions or source material. The skill you build is knowing how to give an agent useful context, assign it a bounded task, and check its work. I’ve written that knowledge workers are becoming workflow designers as agents take on assignments alongside human teammates.

Substack

Mark HullCo-founder and CEO of Exceeds. Former CPO of GoodRx

Pick one real task in your job and run a weekly loop: state the user problem, use AI to make a draft or prototype, test the result with someone who would use it, and record where it failed. Repeat with a better prompt, better context, or a different model. That builds judgment alongside tool fluency.

In an INSEAD session, I had 25 product and engineering leaders build prototypes and pitch them. The exercise exposed both the speed of the tools and their limits: the builders still had to know which user problem was worth solving. I think three skills travel across model releases: understanding where a user's expectations break, knowing how models fail on your tasks, and specifying inputs and outputs clearly enough to check the work.

Keep a small record of the task, the result, the correction, and the time or quality gained. After a month, you'll know which uses actually improve your work.

Aparna ChennapragadaVP of Wearables at Meta. Former CPO at Microsoft

I’d build three habits around work you already do.

Use an approved AI tool on a recurring task, such as preparing for a meeting or drafting a difficult email. Compare its output with what you would have done, and notice where it saves time and where it misses context. AI use is a contact sport; you learn its limits by putting it to work.Spotify
Turn one task into a small prototype. I built a Chrome extension that asks, each time I open a new tab, how AI could help with the task at hand. Building it took about ten minutes with GitHub Copilot. That tiny prompt changed how often I tried AI in my own workflow.Apple Podcasts
Practice delegation and review. Ask an agent to prepare a research brief for a customer meeting, then check its sources, missing context, and recommendations before using it. The lasting skill is specifying the job, judging the result, and deciding what stays in your hands.Spotify

Ami VoraCPO of Anthropic. Former CPO of Faire and VP of Product and Design at WhatsApp

Shishir’s workflow example gives someone a place to practice. I’d make the learning goal equally concrete, because a working professional’s calendar will otherwise eat it. I’ve found it useful to choose one specific growth goal, define what progress looks like, and reserve an hour or two on late Friday afternoons.

Substack

For an AI skill, I’d write the goal into a performance review and ask a colleague what capability would actually help the team. In my career planning, I’ve shared growth goals with my manager and team, included deliverables, and checked every few months what I’d learned.

Substack That feedback loop matters when AI tools change quickly: in a conversation about AI careers, I said I try to stay curious, learn the underlying patterns, and keep applying general problem-solving skills.

YouTube

Sachin RekhiFounder and CEO of Notejoy. Former Head of Product for Sales Solutions at LinkedIn

A repeatable workflow needs a clear standard for good work. In my Claude Code strategy-critique workflow, I give the agent reference files with my frameworks for evaluating a target audience, value proposition, and differentiation. On a personal finance product strategy, it flagged an audience definition that was too broad and a problem statement that was too shallow. Those critiques were useful because the underlying standards were explicit.

sachinrekhi

For a working PM, I’d start by selecting one deliverable, such as a product strategy or customer interview synthesis. Write down the criteria you use to judge it, run AI against an actual draft, and compare its feedback with customer evidence and your own review. I organize PM work by deliverables across vision, strategy, design, and execution, so there are clear artifacts on which to practice.

sachinrekhi

Keep your standards and examples in portable files so you can apply them to the next tool. Your ability to assess the work should improve each time you revise an artifact.

Ami VoraCPO of Anthropic. Former CPO of Faire and VP of Product and Design at WhatsApp

Sachin’s point about writing down standards is useful, and I’d include a way to challenge those standards. I once told my manager I could think of ten reasons a team’s risky idea would fail. She asked me to think of ten reasons it could work. I realized my strength at spotting risks had become a reflex, and I’d stopped running experiments that might change my mind.

Substack

For an AI workflow, I’d write down the criteria for a good result, then ask a colleague to find one case where my criteria miss something. Try that case with the tool and with your usual process. You learn more about the tool and about the judgment you brought to the task.

Mark HullCo-founder and CEO of Exceeds. Former CPO of GoodRx

A skill can feel useful while making the work worse downstream. So I would add an outcome check to the practice loop: for each AI-assisted task, track how much rework it creates, whether the final result meets the user's need, and how much time the review takes. A faster first draft counts only if the whole task improves.

We studied 3,407 engineers over 30 days in open-source projects and found that heavier AI use was associated with more output and higher code acceptance rates. The same study found a slight tilt toward smaller tasks, and the results cannot establish cause or tell you whether the pattern holds in a different workplace. Builders should compare similar tasks over time and keep review standards steady. That teaches you where AI actually helps your craft, instead of rewarding sheer volume of use.

blog.exceeds

Shishir MehrotraCEO of Superhuman (formerly Grammarly). Co-founder of Coda and former Chief Product Officer of YouTube

A useful next exercise is to describe a task in terms of intent rather than copying every old rule into a new tool. When I moved my email, I had nearly 800 filters. I replaced them with four Superhuman auto-label instructions that described the kinds of messages I wanted grouped.

YouTube

For a working professional, the practice is to write a short instruction for a recurring decision, review which cases the AI sorted incorrectly, and refine the instruction. That builds a skill you can carry across products: explaining your judgment clearly enough for a system to apply it and knowing when to override the result.

Aparna ChennapragadaVP of Wearables at Meta. Former CPO at Microsoft

One more skill worth practicing is product judgment in a natural-language interface. If you build an AI helper for your team, prototype the conversation itself: What does it ask before starting? When does it show its plan? How does the user correct it? In my conversation with Lenny, I argued that conversational products still need design, including the timing and usefulness of follow-up questions.

Apple Podcasts

A useful exercise: take the workflow Shishir described and let a colleague try the prototype without instructions. Watch where they hesitate, change the interaction, and try again. That tests whether your AI skill improves someone else's work, beyond your own ability to write prompts.

Ask a follow-up

AI Product

516

AI Product

3M+ insights

AI product leaders from Anthropic, OpenAI, Midjourney & 8 more

by SuperMe

Frequently Asked Questions

What should a first eval setup for an LLM product include?
A sheet of 50 to 100 real outputs graded by people with notes, a rubric checked against customer evidence, a fixed bug set and rule checks in code, and an agreed list of failures that block a launch.
How do you know your LLM evals measure what users care about?
Build eval cases from failures users hit on real tasks, then compare scores with what people do next: rephrasing the question, editing heavily, abandoning the answer or spending time checking it.
Which parts of a product should become agentic first?
A frequent, bounded job inside a workflow people already use, where they can review and correct the result before it goes out. Prototype it on real inputs before committing to the full agent.
How is AI changing the product manager's job?
Execution and coordination get cheaper, so more of the job is choosing the problem, judging output and supervising agents. Sachin Rekhi still wants PMs talking to customers themselves, and Ami Vora wants them close to those conversations.
Which operations work can AI agents handle today?
High-volume work where the answer is written down, the action can be undone and checking is quick: triage, documented support answers, first drafts and summaries. Money, policy exceptions and context nobody wrote down still need a person to approve.
How will AI change services businesses?
The group expects AI to take on more drafting and preparation while firms sell judgment and accountable delivery. Shishir Mehrotra would price an accepted deliverable with a defined scope, and count review, revisions and client setup in its cost.
Why doesn't AI adoption show up in business results more often?
Usage is an input. Results come from changing one workflow and measuring its outcome, including the time spent checking and fixing AI output and whether a decision changed.
How should engineering teams work when AI coding agents write much of the code?
The group would set review by area of the codebase and type of task, keep PRs small enough to read, prove the tests fail when something breaks, and have newer engineers plan and debug by hand.
Where does AI speed up design and content work?
Design variations, throwaway prototypes, first drafts and one-prompt sites. Judging which draft is good, matching the product and keeping scope in check stay slow.

All building with ai questions →

Frequently Asked Questions

AI skills that outlast the next model release

AI Product

Frequently Asked Questions

What should a first eval setup for an LLM product include?

How do you know your LLM evals measure what users care about?

Which parts of a product should become agentic first?

How is AI changing the product manager's job?

Which operations work can AI agents handle today?

How will AI change services businesses?

Why doesn't AI adoption show up in business results more often?

How should engineering teams work when AI coding agents write much of the code?

Where does AI speed up design and content work?

Your ideas, super powered.

AI skills that outlast the next model release

AI Product

Frequently Asked Questions

What should a first eval setup for an LLM product include?

How do you know your LLM evals measure what users care about?

Which parts of a product should become agentic first?

How is AI changing the product manager's job?

Which operations work can AI agents handle today?

How will AI change services businesses?

Why doesn't AI adoption show up in business results more often?

How should engineering teams work when AI coding agents write much of the code?

Where does AI speed up design and content work?

Your ideas, super powered.