AI web design approach

We built a set of templates and set up automation for AI-generated content. We extract the client’s industry from LinkedIn, generate a custom site for them, and send it via cold outreach. More details in the video.

Customer development

We initially focused on freelancers and digital contractors using a “digital accountant” concept. After 20 exploratory and 12 solution interviews, we learned they still rely on a human accountant for personalization and trust.

Spreadsheets

the result of the work is two tables. We worked out how to easily find the necessary information in it. And most importantly, how it gets there. Next steps — dashboards and integration management.

Mobile Version

We realized that the only type of receipt not already covered by the system was paper receipts. To process them, we needed a scanner — which turned out to be the only real use case for the mobile app. That led us to a more rational decision: we integrated this functionality into Telegram and WhatsApp bots.

Kulud.ai

2024

Bookkeeping

UX/UI

Product

About

Kulud.ai automatically finds, extracts, and categorizes receipts and invoices from all your inboxes, and then matches them with the corresponding transactions with the help of AI.

About

Kulud.ai automatically finds, extracts, and categorizes receipts and invoices from all your inboxes, and then matches them with the corresponding transactions with the help of AI.

About

Kulud.ai automatically finds, extracts, and categorizes receipts and invoices from all your inboxes, and then matches them with the corresponding transactions with the help of AI.

Team

My Impact

I simplified the proof-of-concept design, improved UX logic, regularly ran UX and product research (mainly customer development interviews), and set up a feedback loop with early users after the soft launch.

My Impact

Lessons

Personal Touch — Users still want a human feel, so we add team quotes, friendly faces, and a personalized interface.
Social Trust — Trust is critical when handling sensitive data. Real reviews, testimonials, and friend recommendations build credibility.
Optimization — To speed up collecting testimonials, we draft and summarize feedback, then confirm it with users — a trick that also works for external reviews.
Working with AI — Our main challenge is minimizing token usage to improve unit economics, crucial for AI startups post-product-market fit.
No Logo — The kulud.ai interface is intentionally unbranded. In the "0 to 1" phase, we designed it white-label-ready to stay flexible for different growth paths.

Lessons

Personal Touch — Users still want a human feel, so we add team quotes, friendly faces, and a personalized interface.
Social Trust — Trust is critical when handling sensitive data. Real reviews, testimonials, and friend recommendations build credibility.
Optimization — To speed up collecting testimonials, we draft and summarize feedback, then confirm it with users — a trick that also works for external reviews.
Working with AI — Our main challenge is minimizing token usage to improve unit economics, crucial for AI startups post-product-market fit.
No Logo — The kulud.ai interface is intentionally unbranded. In the "0 to 1" phase, we designed it white-label-ready to stay flexible for different growth paths.

Lessons

Personal Touch — Users still want a human feel, so we add team quotes, friendly faces, and a personalized interface.
Social Trust — Trust is critical when handling sensitive data. Real reviews, testimonials, and friend recommendations build credibility.
Optimization — To speed up collecting testimonials, we draft and summarize feedback, then confirm it with users — a trick that also works for external reviews.
Working with AI — Our main challenge is minimizing token usage to improve unit economics, crucial for AI startups post-product-market fit.
No Logo — The kulud.ai interface is intentionally unbranded. In the "0 to 1" phase, we designed it white-label-ready to stay flexible for different growth paths.

Go to the Next

Project

Go to the Next

Project

Go to the Next

Project