Inspiration

Having worked in adtech for over a decade and handled large sets of data, one thing I noticed is there is huge amount of data and it takes weeks to finally sort it and get the answers needed, but it still it won't tell you what to do. This whole process again would take many meeting, discussion and finally an action plan. That's how I came up with Tech Rasa, that not only sorts the data but also tells you what to do, where are leaks, what is strongest point. It helps in better decision for your business.

What it does

Upload your business data in CSV, Excel, or JSON. In under 90 seconds, Rasa runs a 12-module cognitive analysis across your entire company and delivers one Ground Truth sentence, what is actually wrong, what to do about it, and what not to do. Every finding is validated through a 5-layer epistemic engine that checks its own reasoning before surfacing anything to you.

How we built it

Rasa Intelligence was built from scratch during the hackathon window as an AI-native business diagnostics platform. The architecture has six distinct layers, each solving a different part of the intelligence problem. The analysis engine, epistemic validation layer, Gemini integration, memory layer, department intelligence router

Challenges we ran into

there were many challenges, turning data into one sentence and not just a report. Hallucination of AI which can totally kill the business diagnotic and could cost alot. Running epistemic validation at inference time within a 90-second window required splitting the validation layers across Claude Haiku and Claude Sonnet by complexity.

Accomplishments that we're proud of

After launching our product was featured in 2 different magazines Kripitech and Startupfortune(https://krispitech.com/from-data-overload-to-decision-ready-how-rasa-intelligence-delivers-a-business-verdict-in-90-seconds/ and https://startupfortune.com/founders-have-enough-dashboards-what-they-want-now-is-a-diagnosis/). We are an active member of AWS Tel Aviv.

What we learned

One of important things we learned is problem with AI products is output and not the AI. Grounding can not be a feature to be added on top, it has to be the architeture. As wrong number in business disgnotic can drive wrong decision. Another valuable learning is with AI building has become faster nad easier, however, challenge is to market and get users, which is not impossible but the intent has changed alot. As a marketing person myself, I am seeing it in many different channels. As the user buying intent is heavily based on AI suggestion, the funnel looks different for marketing now.

What's next for Tech Rasa Intelligence - AI Business Diagnostic

The vision for Rasa Intelligence is simple and large: every small and medium business on the planet should have access to the same quality of strategic intelligence that today only large companies can afford through consultants, analysts, and strategy teams. The next phase of Rasa moves from a platform you log into to a permanently embedded Company Brain. An always-on intelligence layer that connects directly to every data source a business uses. CRM, accounting, product analytics, support tickets, marketing data all of it flowing into Rasa continuously, so the Ground Truth updates in real time as the business changes. No uploads. No manual triggers. The Company Brain simply knows. We are also building MCP server for Rasa, a protocol that let's any AI tool, agent, connect to Tech Rasa directly. It will be embedded within the organization.

Built With

Share this project:

Updates