Univentures · holding Contact
← All projects
Live Energy certificates · B2C and B2B

Certyfikatomat

An energy certificate ordered like a product in a shop. The document is still signed by a licensed human.

Visit the project site
24/7 ordering without a call
0 emails sent without approval

What it is

Certyfikatomat turns a service that normally takes several phone calls and a visit into an online order. The client submits the property data, pays and receives the finished document.

Underneath sits an ordinary shop and a set of automations that carry the order through its stages: collecting technical data, the auditor’s work, quality control and delivery.

The point of the project is that the owner does not have to sit inside the process. An order can travel the whole way without him, and he steps in only where something deviates from the norm.

Business model

  • Selling individual documents through the shop at a repeatable price, without per-case quoting.
  • Traffic comes mostly from search. It is a sale where nobody chases anyone: the client is looking, because the document is needed for a transaction.
  • Profitability comes from the cost of handling one order barely growing with the number of orders. Only the auditor’s work grows, and that is settled per document.

Where AI wins

  • Client correspondence. Questions about status, deadlines and scope come back in the same dozen or so variants. The model recognises the case and drafts a reply in the team’s voice.
  • Order in the inbox. Messages land in the right case categories, and the ones needing a reaction do not get lost between newsletters.
  • Data consistency between systems. A status change in one place propagates where it should, without retyping.
  • Content that ranks. Dozens of pages answering real buyer questions appear faster than a team could write them.

Where the humans are

  • A licensed auditor. The certificate is signed by a person entered in the register, who takes responsibility for it. That cannot and must not be automated.
  • Quality control. Before the document goes out, someone checks that the data matches. The model can prepare, but it does not bear the consequences of an error.
  • Approving replies. Every AI-drafted message waits for approval. Nothing reaches a client on its own.
  • Unusual cases. Complaints, corrections and situations where the property documentation fits no template.

What it taught us

The biggest gain did not come from the model but from separating two statuses. As long as “ready at the auditor” and “sent to the client” were one state, clients received a completion message with no file. Automation amplifies a well-described process, and amplifies its flaws too.

This describes how we work, not commercial data. We do not disclose client names, terms of cooperation or financial figures.

Want a system like this?

We roll out the same way of working in companies and industry organisations.

AI training and implementation