Why we came in through work

Every consumer AI company in India starts with chat, and the reason is obvious. Chat is the largest possible front door and it costs nothing to open.

We started somewhere much narrower. Somebody applying for a job.

That looks like a smaller ambition and it was not. It was the only entrance we could find that a company could actually survive walking through.

The problem with the obvious door

Indian consumer software has one arithmetic problem and everything else is downstream of it.

A frontier model stack costs the same to run in Delhi as it does in Denver. The bill does not know where the user is. But what an Indian consumer will pay for software is set by the Indian market and it moves slowly, over years, in response to incomes rather than to product quality.

Put those two facts next to each other and you get an industry conclusion that has been repeated for a decade: India cannot be monetised. Build the audience, take the losses, hope advertising arrives before the money runs out.

The conclusion is wrong, but only slightly. The correct version is narrower. India cannot be monetised at that cost base. What a user pays is a market fact. What it costs to answer them is an engineering decision, and it is the one variable in the equation anybody actually controls.

That reframing is most of our strategy. But it still leaves the question of where to start.

Why work

We wanted a category with a second payer, and in India that is a very short list.

Hiring is on it. When somebody finds a job through Redrob, two parties benefit and both of them have money attached. The candidate pays a small consumer subscription. The employer pays for the worker, every month, out of a hiring budget that already exists and is already being spent somewhere less useful.

That second payer changes what kind of company you are allowed to be. It means consumer revenue does not have to carry the whole business. The consumer side can be priced at what people will actually pay rather than at what the model bill demands, because it is not the only thing paying for the model bill.

Almost nothing else in Indian consumer software has that shape. Food delivery has a second payer and terrible unit economics. Payments have volume and no margin. Education has a payer and no repeat behaviour. Work has all three: a payer on each side, a reason to come back, and a question urgent enough that people will tell you the truth about themselves to get it answered.

What a job application tells you that a chat does not

This is the part we underestimated.

A job application is the highest-stakes thing most people will ever ask software to do for them. And because the stakes are high, people supply information they would never volunteer to a chatbot. Verified identity, confirmed by documents rather than self-reported. Real intent, because there is no reason to lie to a system that is going to be checked. A history of what they have actually done, and evidence of what they can do.

Nobody has put those on the same person before, at scale, in India. Job boards have intent and no verification. Social platforms have identity and no intent. Assessment companies have capability and no relationship.

The dataset that comes out of the other end is not a recruiting asset. It is the foundation for everything that comes after, which is why the destination was never recruiting.

What happened next

An agent that gets a job right earns the right to be asked about everything after it.

That is not a hypothesis we are waiting to test. A meaningful and growing share of what people ask our agent has nothing to do with jobs any more. They arrived for work, they stayed for everything else, and every surface we have shipped since is a response to a question people were already asking the agent that used to only do hiring.

Work was not the destination. It was the only door we could afford to come in through.