company · costs · building-in-the-open
What a week of AI employees actually cost us: ₹6.65
Dhanur AI's support, finance, operations, product and marketing are run by AI employees. Here is the real bill from our own work log for the week to 19 September 2026, what each employee did, and what the number does and does not prove.
We run our own company on the product we sell. Named AI employees do our support, finance, platform operations, product triage and marketing, and every model call they make is written to a log with its cost in rupees.
So we can tell you exactly what a week of that costs. For the seven days to 19 September 2026:
| Who | Model calls | Cost |
|---|---|---|
| The agent builder, drafting new agents | 3 | ₹2.6549 |
| Our product helper, answering questions about Dhanur AI | 7 | ₹1.9051 |
| A support desk agent, in testing | 7 | ₹1.1803 |
| Meera, chief of staff — the daily report | 3 | ₹0.4556 |
| Ishaan, marketing — drafting posts | 3 | ₹0.2394 |
| Vikram, sales — triaging a message | 1 | ₹0.2106 |
| Total | 24 | ₹6.6459 |
Six rupees and sixty-five paise. Less than a cup of chai.
What that number does not prove
It would be easy to stop there and let you draw a flattering conclusion. So, plainly:
We have almost no customers. That is the main reason the number is small. A company with real inbound volume would spend more, because it would be doing more. This is the cost of a small company's own operations, not a prediction of yours.
Four of our seven employees did nothing at all this week. Kabir (platform operations), Raghav (support), Tara (product) and Ananya (finance) only received their duties on the day this was written. Their work does not appear above because it had not happened yet.
Model calls are not the whole bill. This is what the AI itself cost. Servers, email, storage and payment fees are separate, and for us they are larger than ₹6.65.
What it does show
AI work is cheap per action and the price is knowable in advance. A typical reply from a Smart agent costs us a little over a rupee, and we sell it at about ₹2.17. That price is worked out from a typical reply's length at the model's price, not averaged from our ledger; the ledger then records the exact cost of every reply. It's on our pricing page because you should be able to do the arithmetic before you sign up, not after.
A duty that has nothing to do should cost nothing. Every employee's job is written so that a quiet day produces no report, no inbox item and no model call. Kabir checks the platform's health every fifteen minutes; a healthy platform costs ₹0.00, because he stops before the model call. That discipline is why "an employee that works every fifteen minutes" is affordable at all.
Most of a bill is doing the work, not the overhead. Look at the split: ₹5.74 of the ₹6.65 went on building and running agents — the actual work. ₹0.91 went on the whole company layer: the daily report, the marketing drafts, the triage. The reporting and supervision cost about a seventh of the doing.
What this means for your business
If you run a shop, a clinic or a coaching centre, the useful version of this is: you can find out what it costs before you commit. Every new workspace starts with ₹25 of free credit, top-ups start at ₹50, and every task and test chat shows its exact price in rupees when it finishes.
Ask your agent thirty of your customers' real questions on the free credit. Then decide.
Why we publish this
A company that says it is run by AI and will not show you the bill is asking you to take its word for something it can trivially prove. Ours is on our company page, refreshed from our live systems, on the bad days too.
Next week's number will be different, and if it is dramatically higher we will say why.