LLenny's Podcast
← All frameworks
StrategyPaul Adams (CPO of Intercom)

Map Your Product Against What AI Can Do

Start from your product's core job, then classify each part as AI-replace, augment, partial, or not-yet

Difficulty
Moderate
Time to result
~weeks to results
Steps
4
Confidence
95%

A first-principles method for setting AI product strategy that deliberately starts from the customer problem rather than the shiny technology. You go back to basics on why people use your product, enumerate the concrete things AI can now do, then map one against the other to see which parts of your product sit 'in the path of the meteor.' The output is a per-capability verdict — replace, augment, partial, or can't-yet — that tells you where to bet.

Origin

Paul Adams, Chief Product Officer at Intercom, developed this as the lens Intercom used when they 'ripped up our strategy almost entirely' after ChatGPT launched (Nov 2023) and rebuilt from first principles, leading to their AI chatbot Finn.

Core principles

  • 01Don't start from the technology — it's seductive and it sidetracks you (taking a photo of your dinner is cool but not a strategy).
  • 02Start from the product's core premise: what problem does it solve and why do people love it.
  • 03AI capability is a concrete, enumerable list (write, summarize, answer queries, find facts, scan text/images, listen to voice, and increasingly take actions) — map against that list, not a vague notion of 'AI.'
  • 04Any product with a workflow (almost all B2B SaaS) or multimedia is in the meteor's path.
  • 05The verdict per capability is not binary — it's replace / augment / partial / not-yet.

How to run it

  1. 1

    Go back to basics on the core premise

    Ask what your product actually does, what problem it solves, and why people use and love it. Strip it to the fundamental job before thinking about AI at all.

    Pro tip Frame it as a customer question ('why do people love this?'), not an internal feature inventory.

    Watch out Resist starting with a cool AI demo — the tech will sidetrack you away from what actually matters to customers.

  2. 2

    Enumerate what AI can concretely do

    List the specific capabilities: write text, summarize text, answer queries, find facts, scan text and images, listen to and repeat voice, reason/build rules, and take actions. Include what it 'will be able to do' soon, not just today.

    Pro tip Treat 'take actions' (AI actually doing things, like changing a flight) as the next big unlock to plan around.

  3. 3

    Map product capabilities against AI capabilities

    Lay your product's jobs against the AI-capability list. For each, ask: can AI do this fully, partially, or not yet? Classify each as replacement (AI just does it) or augmentation (AI helps the human do it).

    Watch out If your product is in the path — has workflow or multimedia — a foundational strategic change may be needed, not a bolt-on feature.

  4. 4

    Decide what you do today

    Turn the map into action: where AI replaces, rebuild around it; where it augments, add the co-pilot; where it can't yet, hold. Be willing to bet the farm on the parts most exposed.

    Pro tip Watch for a competitor's confirming signal (Intercom acted partly because Sam Altman named customer service as a first industry to be disrupted).

    Watch out Don't 'bolt it on' with a siloed AI team — the strategy has to reshape the core product.

In the wild

Intercom rebuilds strategy around Finn

After ChatGPT, Intercom mapped its core job — customer support — against AI capability, saw that answering customer queries was directly replaceable, and 'literally ripped up our strategy almost entirely and started again from first principles,' building the AI-first chatbot Finn as the first line of defense before a human.

Some customers now resolve up to 50-70% of inbound questions with Finn; Intercom treats AI as the future of the product with senior leadership 'all in.'

The reporting-software thought experiment

Adams applies the map to a Tableau-style reporting product: its core job is helping people see and decide from data. Run against AI, the whole UI of charts, filters and queries could collapse into 'just a box' you ask 'who was our top rep in January' — meaning the product is ripe for a newcomer with a different paradigm.

Illustrates that even non-obvious categories (reporting, project management) sit in the meteor's path once mapped to the core job.

Common mistakes

Starting from the technology

Getting sidetracked by cool demos (photographing food, GPT Vision) instead of the customer problem produces novelty, not strategy.

Bolting AI on

Creating a separate 'AI team' that adds AI to everything, rather than integrating AI thinking into every product team, repeats the mistake companies made treating mobile as a side project.

Is it for you?

Best for

Product leaders and founders in established SaaS categories deciding whether AI is a feature or a foundational strategy shift

Not ideal for

Products with no workflow or multimedia surface where AI has little leverage, or teams that need incremental optimization rather than a strategy reset

From the transcript

I wouldn't start there I'd start with uh the thing your product does like what's the core premise behind it why do people use it

21:30

you kind need to map like what your product does against what AI can do

22:00

for some of it it'll be like kind of replacement AI will replace it'll just do it and you know another places it be augmentation…

23:00

we literally ripped up our strategy almost entirely and started again

24:00

From the episode

What AI means for your product strategy

Paul Adams (CPO of Intercom)