Latent Demand Mining
Watch for people jumping through hoops to make your product do something, then make that the smooth path.
- Difficulty
- Moderate
- Time to result
- ~months to results
- Steps
- 4
- Confidence
- 88%
Anthropic uncovered coding and knowledge-work markets by watching latent demand — behavior users exhibit before there's a first-class feature for it. Fiona Fung's method: notice when people use the product in ways you didn't intend, especially when they jump through hoops to make something work, form a hypothesis about the underlying job, and build a smoother, better experience for it. This is how Claude Code's non-coder usage became Cowork, and small-business usage became a bundled Claude for small business.
Origin
Fiona Fung, at Anthropic; she notes the pattern served her across products long before Anthropic and ties it to 'customers use your product in ways you did not intend.'
Core principles
- 01Customers will use your product in ways you did not intend, for good or bad
- 02Jumping-through-hoops behavior is a signal of latent demand for a smoother experience
- 03Form a hypothesis about the underlying job before building
- 04Stay close to feedback and iterate — the best path is learning, not planning
How to run it
- 1
Watch for unintended usage
Keep an eye out for use cases popping up that you didn't design for, especially adjacent to your core product.
- 2
Spot the hoop-jumping
Notice specifically where users are working hard or improvising to make something work — that friction marks latent demand.
- 3
Form a hypothesis about the real job
Articulate a hypothesis for what the user is actually trying to accomplish, rather than just copying the behavior.
- 4
Build the smoother experience
Make the improvised path a first-class, smoother, better experience and keep iterating on feedback.
Pro tip Bundle scattered steps into one toggle or flow when users are manually stitching them together.
In the wild
The team noticed many people who were not coders were using Claude Code, asked whether they could make that experience better, and leaned into it as latent demand. Separately, small-business owners kept asking which plugins were included, which became a bundled 'Claude for small business' toggle inside Cowork.
→ Latent-demand signals turned into new products (Cowork) and a bundled small-business experience, ahead of explicit requests.
Common mistakes
Fixating only on intended use
Being all fixated on the use case you designed for (e.g. PDFs and invoices) blinds you to the unexpected, higher-value ways customers actually use the product.
Copying the behavior without a hypothesis
Building the improvised workaround as-is, instead of hypothesizing the underlying job, misses the chance to make the experience genuinely smoother.
Is it for you?
Best for
Product leaders and founders looking for the next opportunity adjacent to a product people already use in surprising ways
Not ideal for
Pre-launch products with no real usage data yet to reveal unintended behavior
From the transcript
“we also keep an eye on like latent demand.”
“co-work we noticed hey, a lot of folks that were not necessarily coders were were using Claude Code. Can we make that experience better?”
“because actually when you see people jumping through hoops to make something work, can you actually make that an even like a smoother and and…”
“customers will use your product in ways that you did not intend for good or for bad.”
From the episode
What happens after coding is solved?
Fiona Fung (Manager of the Claude Code and Cowork Teams)