Mykola Vorobiov

Case study · Real estate

KIMA: a CRM-agnostic intelligence layer for German real-estate agencies.

Each system after SignalsAPI carried the same method into a different industry and a different technical shape. KIMA is the transfer proof: paired event and command queues underneath, in the agencies' hands 2025 to 2026.

Problem

The problem: routine CRM work on different systems.

German real-estate agencies run different CRM systems with different APIs and different sales processes, and every agency's workflow looks unique on the surface. Underneath the surface variation, salespeople waste hours on routine CRM operations: logging calls, moving deals between stages, sending follow-ups, closing out tasks. The work is the same in shape; the surface is what's different.

Insight

The insight, twice over.

The diagnostic line:

Domain experts make a consistent mistake: they translate their manual process directly into software. "Automate my spreadsheet." This overconstrains the problem because humans and software think in fundamentally different categories.

The second insight is positioning: KIMA is an AI employee for real estate agents (it connects to CRMs), not a real estate CRM. The public hears "real estate software" and pictures another CRM; KIMA sits above the CRM, doing the routine work the salespeople would otherwise do by hand.

Architecture

The architecture: paired queues, one invariant in the middle.

Schematic — shape only
KIMA Real estate
CRMs inbound adapters event queue business rules command queue CRM events to standardised signals outbound commands act back

The invariant underneath the chaos is a CRM-agnostic intelligence layer on two queues: inbound adapters turn CRM events into standardised signals, business rules decide what should happen, and outbound commands act back on the CRM. Different CRMs plug in via adapters. Different agencies configure via business rules. The intelligence layer in the middle never changes, which is why this is a platform rather than a set of bespoke integrations.

The whole system is Python, event-driven. The German name carries the positioning: Künstliche Intelligenz Mitarbeiter, an artificially intelligent employee. The system is "CRM-independent and agency-independent" precisely because the adapters and the rules absorb what changes, and the middle does not.

Proof

The proof, and what it shows.

Two months from the idea to a working system in the agencies' hands. SignalsAPI's method was learned on a self-healing scraping pipeline against hundreds of sources that change daily; KIMA is the proof that the same move transfers to an entirely different industry and an entirely different technical shape, event-driven CRM orchestration. The hidden invariant is the same; the surface variation is not.

Credit

The line under the work.

Architect & Sole Developer.

KIMA: a CRM-agnostic intelligence layer for German real-estate agencies, on paired event and command queues, in the agencies' hands 2025 to 2026, as Architect & Sole Developer. The transfer proof: the same invariant-under-the-chaos move as SignalsAPI, in a different industry and a different technical shape.