Start with a Press Release: What Amazon Taught Me About AI and Product Development

Over the past couple of years, I’ve been deep in the weeds of product and services development. And one of the most important lessons I’ve learned, the hard way, is this:

If you can’t clearly explain who your product is for, what problem it solves, and why someone would pay for it… it’s probably not worth building.

That’s where Amazon’s “Working Backwards” approach comes in. It starts not with code, architecture diagrams, or even a roadmap—but with a press release.

Why Start With a Press Release?

Amazon’s founder Jeff Bezos championed this method to keep teams focused on what really matters: the customer. The idea is simple. Before you build anything, write an internal press release that announces your new product or service.

It needs to explain:

• The problem your customer has
• The solution your product offers
• The benefits and outcomes for the customer
• Quotes from imagined customers or stakeholders
• And why it’s better or different than anything else out there

It sounds simple, but it’s surprisingly hard. That’s the point.

“We start with the customer and work backwards. That becomes the touchstone for how we invent.” — Jeff Bezos

Writing Is Thinking

When I first came across this idea, I thought it was a bit of a gimmick. But I’ve since come to appreciate how powerful it is. Writing a good press release forces you to think clearly—and simply—about what you’re trying to build. If you can’t do that, chances are your customers won’t understand it either.

Once the press release is done, the team can go further:

• FAQs – to anticipate technical, business, and customer questions
• User manuals – to visualise how someone will actually use the product
• Mockups – to bring the experience to life before any code is written

Only after all this is approved does development begin.

This is the part that is most frustrating to engineers who can’t wait to get stuck into the technical part, so I always take time to explain why this phase is critical to the commercial success of the service or product. Without it, the chances of seeing it in production and actually adopted, are slim.

My Own Experience: What Happens When You Don’t

In 2023, I was part of a team that took six months building a service. It looked promising on paper. But we dove straight into engineering and documentation without taking enough time to sit with the customer problem, validating and testing it—and testing our hypothesis on how much someone would pay for the solution.

You can guess how that ended. It never took off. We overcooked it. The price point wasn’t right—and it was built for a customer segment we didn’t have access to at that time.

Looking back, we didn’t have a clear line of sight to customer value. And honestly? That’s still something I’m learning every day. Especially in engineering-led roles, it’s so easy to get caught up in the “what” and forget the “why.”

What I’ve Learned About Product-Market Fit (Especially Under $20M)

Another hard lesson: Don’t over-document when you’re still in search mode.

I used to think we needed all our processes nailed down early. But when you’re still trying to find product-market fit—typically under the $20M revenue mark—your time is better spent selling, meeting customers and experimenting, not perfecting processes.

That phase is all about hustle: talk to customers, learn their needs, test ideas, meet some more customers…and iterate fast.

Only around the $20M+ mark does portfolio planning really start to matter. That’s when tools like the BCG Matrix, market segmentation, and customer persona work come into their own.

AI Is No Different

The same thinking applies to AI. Before jumping into models, pipelines, or data wrangling—ask yourself:

• Who is this for?
• What problem does it solve?
• How will we announce its success?
• What kind of customer (or segment) will care?
• In that segment, who are the personas that care – is it the CIO, COO, Head of Product or Head of Finance you’re building for? How will it make their life easier?
• How will we sell it—not just to the end user, but to their boss, their procurement team, and anyone else who needs convincing?
• What will be needed to successfully launch it?