We all know discovery matters.

Talk to your users. Understand their problems. Validate before you build. Every product leader has heard this a hundred times.

But the most critical gap I see in high-growth product teams today is not whether they do research. It is how they build it into the way they work.

Last week, DoorDash launched a new app called "Tasks."

It pays their couriers 8 million Dashers across the US to record videos of everyday activities. Handwashing dishes while wearing a body camera. Photographing restaurant menus and hotel entrances for delivery drivers… etc

All of this data feeds AI and robotics models. DoorDash sells it to partners in retail, insurance, hospitality, and tech.

Now, this is obviously an AI story. But I believe the deeper lesson here is about discovery.

Nobody on the product team needs to remember to schedule anything.

To understand why this matters, we need to talk about how most product teams actually do research today.

In 2021, Teresa Torres published Continuous Discovery Habits.

Her core idea was simple.

Stop treating research as a big quarterly project. Make it a weekly habit instead.

She found that most teams only talk to customers at the beginning of a project, if at all. Then they disappear into sprints, build for months, and hope they got it right.

Her recommendation: minimum weekly touchpoints with customers. The product manager, designer, and engineer — what she calls the "product trio" — should talk to real customers every single week. Even 30 minutes is enough.

When teams do this, something shifts. They stop guessing. They catch problems earlier. They build things people actually want.

But here is the thing. Most teams still struggle to actually do it.

In my experience running design sprints and leading design teams for more than a decade, I have seen this pattern again and again.

A team commits to weekly customer conversations.

- It works for a few weeks.
- Then a sprint gets busy.
- A deadline moves up.
- The PM is in meetings all day.

The designer is asked to deliver prototypes while also running interviews.

And the weekly touchpoint? It is the first thing that gets dropped.

Teresa Torres herself says it only takes 30 minutes a week. She is right. But even 30 minutes requires someone to schedule it, prepare for it, show up, and synthesize what they heard.

Most teams think they are doing discovery. But there are levels to it. Here is how to tell where your team actually sits.

You pause everything. Hire researchers or an agency. Run a study. Wait for findings. Go back to building. Research happens at the beginning or not at all.

According to my experience, this is where most companies still sit.

Weekly customer touchpoints. Teresa Torres' model. It works but it depends on human discipline. And human discipline inside a busy sprint is not always reliable.

This is what DoorDash just demonstrated. The product itself captures the data. Discovery is built into how the business operates. No one needs to schedule an interview because the system learns on its own.

Ask yourself honestly: where does your team sit on this ladder?

You cannot turn your users into AI trainers tomorrow. But you can start moving toward Level 3 with two steps.

Sit down with your team and list every place your product already touches a user. Support tickets. Onboarding calls. Feature usage data. Session recordings. NPS responses. Checkout flows. Error messages.

Most teams are surprised by how long this list is. You are already generating signal everywhere — you are just not capturing it.

I have done this exercise with teams and the reaction is always the same: "We had all of this sitting there and never looked at it."

Look at your list and find the one touchpoint with the most signal. Then ask: how can we capture what users are telling us here — without anyone on the team having to schedule a call or run a study?

This is the part most teams overlook. You do not need a research team or a big budget to start automating discovery. You need AI and a human in the loop.

I am building Lyyvora right now, and I run experiments like this with interns I hire for 60 hours of work. One intern. One AI tool. One touchpoint to analyze.
That is enough to get started.

Here is what that looks like in practice:

The point is this: AI handles the processing. A human keeps it grounded. You don’t always need a senior researcher.

You need someone curious, an AI tool, and a clear question to answer.

That is it. One audit. One person with AI. You are not at Level 3 yet, but you are no longer depending on your senior team remembering to schedule an interview.

💬 Claude - CoWork Feature — Anthropic's Here are projects available in co-work, and update that helps the AI better remember how you work. (FYI: this article was created with 2 days of back and forth with co-work inside a project)

🤖 Manus — I am debating between Gamma and Manus.ai, and
Manus.ai is still winning, creating great slides for presentations.

Tech moves fast, but you're still playing catch-up?

That's exactly why 200K+ engineers working at Google, Meta, and Apple read The Code twice a week.

Here's what you get:

All delivered twice a week in just 2 short emails.

Join 200K+ engineers