The first question every AI founder should ask — is NOT about AI !!
Every other day I meet founders who tell me they are building an AI product. My first question is never — “Which LLM are you using?”
Instead, I ask something much simpler — “Where does your data come from?”
That one question often determines whether the product will become truly intelligent or remain just another AI wrapper. Every successful AI product is built on a journey that starts long before AI enters the picture. It usually looks like this:
Physical World –> Digital Data –> Structured Information –> Intelligence –> AI-driven Decisions
Most founders immediately jump to the last step and very few invest enough time in the first three. Take any industry as example:
– doctor examines a patient.
– warehouse receives a shipment.
– salesperson meets a customer.
Everything begins in the physical world and until that interaction is captured digitally , AI has nothing meaningful to work with.
The next challenge isn’t collecting data .. it’s collecting the right data. Every business has its own raw material but raw data rarely creates intelligence — it has to be “organised , connected , cleaned , labelled , related .. in short enriched.
Only then does it become information that AI can understand. This is where many startups underestimate the effort. They spend months choosing models but very little time designing their data architecture. IMHO changing an AI model is easy .. re-building years of poorly collected data is not.
One of my favourite ways to think about it is this: “look at data to AI .. what fuel is to an aircraft is.” People admire the engine and yet a few realise the quality of the journey depends on the fuel.
However , the objective isn’t simply to collect data .. it’s to create context. AI doesn’t become intelligent by seeing more data. It becomes intelligent by understanding better context.
Founders often ask — “When should we introduce AI?”
I think the better question is — “Have we built a data foundation that deserves AI?”
Equation – good data creates better intelligence .. poor data creates faster mistakes.
As AI becomes cheaper and more accessible , data will become the real competitive advantage. Not because companies have more of it but because they understand it better.
So if you’re building an AI startup today , resist the temptation to start with prompts , models or agents. Start with looking into your data.
My 2 cents – “AI isn’t the foundation of an intelligent product. Data is. AI simply reveals the quality of the foundation you’ve already built. So , before you build an AI product , build an AI-worthy data foundation.”
Until next time .. happy thinking !!
Author – Sumit Rajwade, Co-founder: mPrompto
