From idea to MVP in weeks: what AI actually sped up, and what it did not

Two years ago a working first version of a product took us months. Today it takes weeks. That compression is real, and it has changed what a small team like ours can take on.

It has also created a quiet misunderstanding. When people hear that AI makes building faster, they assume everything got faster. It did not. Some parts of a project collapsed from weeks to hours, and some did not move at all. Knowing which is which is most of the skill now.

What genuinely got faster

• First versions. Getting from a clear plan to something clickable used to be the long part. Now it is often the shortest.
• Variations. Exploring five directions for a screen, a slide or an illustration costs about the same as exploring one.
• Repetitive production. Applying a brand across dozens of assets, writing tests, documenting components, resizing campaigns for every format.
• Research groundwork. Summarising interviews, organising notes, comparing competitors. The reading still matters, the sorting is fast.

A good example is the deck engine we built for Caya. It uses Claude together with Caya’s detailed brand guidelines to turn presentation content into designed slides. Deck production went from weeks to days, and the quality stayed at the level of bespoke agency work, because the brand foundation behind it was done properly first.

What did not get faster at all

The tools got superhuman. The judgement did not come included.

• Deciding what to build. Which problem matters, for whom, and what the first release has to prove. No model makes that call for you.
• Getting decisions made. The slowest part of most projects is still waiting for a yes, a piece of content or a partner’s approval.
• Earning trust. Insurers, banks and hospitals do not move faster because your code was generated faster.
• Quality control. Someone with experience still has to look at every screen and every line and decide whether it is right. Fast output that nobody checks is just fast debt.

When building takes weeks instead of months, the bottleneck moves. For most of our clients it has moved to decisions and to sales.

The bottleneck moved

This is the part that surprised me most. When the product is no longer the slow part, everything around it becomes the constraint. Teams that used to have a year to figure out positioning, pricing and their first customers now have a working product in a quarter and nobody to sell it to.

That is why our work rarely stops at the product anymore. On WTE Companion the same team that built the app also created the brand, the website and the first paid social campaign. That campaign brought in 321 leads in a single day. A faster build only paid off because the go-to-market was ready when the product was.

How to plan a project now

• Plan the go-to-market at the same time as the product, not after it. Expect the product to be ready sooner than your launch.
• Put your senior people on the decisions and the reviews. That is where their time has the most leverage now.
• Be suspicious of timelines that only got shorter in the build phase. If discovery and approvals take as long as before, the project will too.
• Ask any partner what they check by hand. The answer tells you more about their quality than their tool list.

AI has made us faster at almost everything we produce. It has made the thinking, the deciding and the selling more valuable, not less.