How we envisioned, user tested and completed a custom application in 4 weeks using AI

Background

At TeamPilot.io we specialize in software product development, connecting top freelance talent with clients in the Nordic region. Most clients are product companies looking for experienced developers and teams, either to expand existing teams, or to get a full service delivery. They employ freelancers both in Scandinavia and across the globe, with the focus on providing high quality, trustworthy talent within product development.

Turning visitors to leads, considering diverse buyers

The goal with rebuilding our public app was to turn site visitors into actual leads and to improve brand perception.

With a fairly unknown brand, we wanted to better showcase the experience and expertise availble in our network, so potential customers could see the breadth of expertise available and easily find the right person or team for their need. Our hypothesis was that different buyers search in different ways, and we wanted to learn more about that.

The Project Timeline

The application was actually completely developed by one product manager, in a few weeks working about half-time, in a fully AI-native development process. The focus was on the customer, and business development and finding the right service for our customers rather than on the programming.

We can do real agile business development in real time while building the service, all due to the new AI-driven business development process Johan developed for us. - John Cannerborg, COO

We planned for a four week first phase of the project, and met the timelines very well. What did take some more time was the final tweaks and preparations for going live, so we monitored and did minor tweaks a little bit longer.

• Week 1: First customer session and a clickable front-end prototype

• Week 2: UI improvements and admin interface built

• Week 3: User tests, backend integration, and feedback implementation

• Week 4: Stabilization, automation, bugfixes

• Post-development: Final tweaks and go-live preparations

Was it easy.... Well, while I am an experienced product manager and engineer, I had not done production grade programming and deployments in many years, and not in the programming languages and frameworks we chose. So I had a steep learning curve for the gaps that the AI-tools didn't fill, and catching up with modern development practices hands-on. This was however caught up along the way during this project - within the timeline.

Proof in the pudding

Within three weeks we could actually present a working beta to a mix of CTOs and purchasing managers, providing valuable feedback within the time that traditional projects barely have had their kickoff.

I was thrilled to get customer feedback on a working beta already the third week - John Cannerborg, COO

The user tests showed that all used the site in completely different ways. Some want to see past experience in their industry. Others care more about specific skills, roles, or companies the freelancer has worked with. With immediate feedback on their way of thinking, what they missed and what they liked the most, we could immediately adjust the direction of the product.

A product based on customer needs and perspectives

Our end product was a public website that enables customers to get an immediate taste of the capabilities and find the right experts within TeamPilot’s network.

Below, the front page developed based on the customer feedback received already in the third week:

In a few weeks we had an application that provided:

• An interactive public interface, where the customers can look at the data and content from many different perspectives, all based on customer feedback.

• A content management tool for staff to update the site, including case studies, blog posts and videos.

• Integration with the backoffice Talent Management API to access live data with contractor profiles, industries, and skills

• Ability to route customer requests into the CRM

Key Learnings

While both the customer and I was thrilled about the speed of business development with AI tools, there were both pros and cons. Most likely the cons will be solved by the tools very soon.

What worked:

• Fast feedback: I showed working concepts to the customer within days, not weeks.

• Time to market: Real users tested a working product in the third week. We could easily fine tune and adopt the service based on feedback.

• Small team - no wait, no handovers: AI-generated design and UX were “good enough” to get started without waiting for specialists.

What we were challenged with:

• 80/20: While the design, ux and programming were automatically generated very fast by the AI, it lacks the finesse you get a professional team.

• Random AI edits: Small change requests could break things anywhere. Prompting and testing became the main tasks. More time had to be added to testing when letting the AI develop.

• Weak deployment support: Our main tool, Lovable, didn’t help with proper deployment outside of their own service. We handled that separately with traditional CI/CD scripts, which still requires engineering skills.

• SEO gap: The AI developed site was a pure front-end application, and needed extra work to become SEO-friendly. The tool we chose didn't support this out of the box.

Was It Worth It?

Absolutely. I estimate that using fully AI-driven programming was 2–4x more efficient than a traditional system development approach. In addition to that, we saved a lot of time with AI driven design and UX.

But the real value wasn’t in programming efficiency! It's in fast and agile business development, being able to user test and adapt for new ideas immediately. We could go from discussion to a working version in hours, not days. Sometimes we built a new feature or fixed something even during a meeting.

We will now continue to evaluate the business impact, iteratively improving, before polishing something with an unknown value.

Final Thoughts

If your priority is getting to market fast and testing with real customers early, AI-assisted tools like Lovable and Cursor can make a huge difference. This way of working isn’t for everyone though, and if you need strict code quality and long-term maintainability, traditional methods still win.

Need help to speed up business development?

Please feel free to reach out to me. I would also be happy to educate your leadership, teams, or help implementing AI-driven agile business development in your organization.

Johan Eliasson

Johan Eliasson is an independent product and technology leader. Please feel free to reach out for any questions or opportunities for collaboration.