Beleggio (former Costpilot)

Take back precious time + nerves by AI-powered cost intelligence: Automates receipt collection + tracking of recurring (monthly + yearly) costs, gets clarity about sufficient liquidity year-round.

Comment

Inspiration

As the owner of an online business, I spent years manually collecting, downloading, renaming, and filing receipts from dozens of software subscriptions, vendors, and service providers.

Over time, I started procrastinating the process because it was repetitive, time-consuming, and offered little immediate value. Instead of staying on top of it, a growing digital pile of receipts accumulated across email accounts, vendor portals, and cloud folders.

The longer I postponed the task, the harder it became. Months later, I often had to reconstruct which receipt belonged to which service, payment, or accounting period. At the same time, I never had a reliable overview of my average monthly costs, recurring subscriptions, or future payment obligations.

To compensate, I built and maintained manual Excel spreadsheets to track recurring expenses and estimate future costs. While this helped, it required constant manual updates as well and quickly became outdated. Because again, I was the bottleneck.

In 2025, I built my first automation using GPT and Make.com to organize receipts and expenses automatically. The concept showed great potential, but the available technology was not yet reliable enough for real-world business operations.

Rather than forcing a solution, I decided to wait until AI agents became capable of handling more complex workflows with greater accuracy and reliability.

When agent-based AI systems matured, the opportunity became clear: build the solution I had wanted for years; not only for my own business, but also for my clients and fellow entrepreneurs who struggle with the same administrative burden and therefor lack of real facts in financial visibility.

What We Learned

One of the biggest lessons was that receipt management itself is not the real problem. Turns out, that having a realtime overview of recurring costs offers better decisions in the daily life of a CEO.

I wanted to offer this according to the receipts we collect. So the real challenge is automating the process while providing visibility into realtime recurring costs, future payment obligations, and upcoming cash requirements. So that we are really able to make the best decisions instead suddenly having liquidity problems, because we "forgot" yearly payments or other obligations.

Most accounting systems are designed to document the past. Entrepreneurs, however, need visibility into the now for clever decisions concerning the future.

We also learned that full automation is not always the goal. When financial records are involved, trust matters. AI should automate as much as possible, but ask for human feedback when it comes to sensitive data, or when confidence is low. To learn from corrections fast, and continuously improve.

Another important lesson was the value of system independence. Rather than building Beleggio on top of a single existing agent platform or even using pre-existed agents offered by the global big players, our goal is to create an independent agent ecosystem that can operate on the client’s own computer and infrastructure. But external AI models help us build Beleggio, but Beleggio itself is designed to remain flexible, portable, and independent of any single AI provider and even independent of internet for a short amount of time (if necessary, like times of war or other tec issues).

This approach also strengthens data privacy and security, helping clients maintain control over sensitive financial- and accounting information while building greater trust in the system.

We also discovered that AI can dramatically reduce support and maintenance effort. When users report a problem, it is often more efficient to let an AI copilot analyze logs, reconstruct the issue, and suggest or implement a solution than having humans spend hours manually tracing the root cause. This insight inspired the development of an internal AI support copilot that helps diagnose, repair, and continuously improve the system.

How We Built CostPilot (Workingname) aka Beleggio (final name)

CostPilot is being developed as a team of specialized AI agents working together.

The system automatically collects receipts from multiple email accounts and sources, identifies vendors, tracks recurring expenses, prepares accounting records according to the owners preferences, and forecasts future payment obligations.

Our architecture combines email scans, document processing, AI-powered classification, workflow orchestration, and human-in-the-loop validation.

Rather than creating another document storage system, our goal is to transform financial administration into actionable cost intelligence that helps business owners make better decisions fast (instead of waiting until the accountant presents "the numbers"), improve financial visibility, and maintain sufficient liquidity throughout the year.

Challenges We Faced

One of the biggest challenges was teaching AI to reliably distinguish between invoices, payment confirmations, recurring subscriptions, and irrelevant emails (like very confidently written spam mails) across many different formats and providers.

Another challenge was moving beyond simple receipt management. We wanted to build a system that not only organizes documents, but also creates meaningful financial visibility and predicts future obligations.

The biggest challenge and the most time eating was agent 4ths task: to download missing vendors (after the data synch with the banking analyzer, which is a bonus reality check of beleggio). This touches very sensible data exchanges with different login procedures and many portals are not even meant to let in a bot... we lost too much precious time (about 6 weeks) with that. I then decided to postpone this service and not to include it in the foundingmember-package, again trusting that the fast AI-Development around us will suddenly offer easy solutions.

That been said, the fast upgrades of Codex, and coming up with the model 5.6 Sol, was a magical relief within a short amount of time! Really impressive.

Collecting receipts and offering this overviews remains a core focus of the project. Our approach is to keep humans in control while allowing AI agents to handle the repetitive work with a better promise of care and execution than human could ever offer.

This work then is just done. It reliefs many CEOs or assistants of CEOs head and deletes not only a wrong monetary assessment of the current situation, but definitifly gives a lot of time and nerves back. Speaking of nerves: this concerns those always thinking of the document stuff not to be done yet, but also the better clarity for financial decisions to invest through the daily life, reliefs today and the tomorrow of the company.

Accomplishments That We're Proud Of

We transformed a real operational problem from daily business life into a scalable AI-native solution. We were first surprised that almost no one did offer this service. Now we know why: Not everything can get automated caused of the sensitiv and legal data and accesses.

We successfully moved beyond traditional receipt management and designed a system focused on future financial visibility, recurring cost intelligence, and liquidity planning, letting work AI-agents hand in hand with specific human tasks.

Most importantly, we built the foundation for a platform that is designed to help not only our own businesses, but also thousands of online entrepreneurs facing the same challenges.

What's Next for Beleggio

The first prototype for us and a first client is built, installed and works beautifully.

Our next milestone is completing the team of agents, we plan to offer 6 agents in the complete package. But there are already more ideas for further agents, such as to support also offline entrepreneurs and local shops with paper receipts to process them fast and the most automatically possible.

We are expanding the agent ecosystem to include intelligent vendor portal access, bidirectional feedback loops, automated issue resolution, and AI-assisted, self learning customer support.

Long term, our vision is to help startup online businesses to have a better chance to survive, thanks to automated financial overviews; because they can lead their energy and nerves to things, which attracts clients and brings more revenue.

But our vision is also to support big, big online players thanks to this automation, to focus their human team power into spaces, where only human services are needed, with clients and other humans, related to the company. And of course, to safe several salaries when it comes to collect and interpret those documents to fulfill the control systems of governments and to safe a ton of money because of the realtime visibility of their costs.

Let's move from receipt chaos to cash flow clarity through intelligent automation, financial visibility, and proactive liquidity planning!

We plan to build a partner ecosystem through a combination of software licensing, certified implementation partners, training programs, and ongoing service subscriptions, we aim to create a scalable model that delivers value to entrepreneurs while ensuring sustainable growth of the Beleggio ecosystem.

Built With

Share this project:

Updates

Submission history