What we answer for
Your name on it.
We answer for it.
What we answer for, in writing, before you have to ask. Last updated 18 October 2026.
Our mistake is never your team's.
Nothing leaves until the person you name says yes.
If we stop, it is all still yours.
Tested on twenty of your own past jobs before it goes live, and kept working as AI changes.
Your yes covers what you saw.
We answer for the rest.
Ours to answer for
- Rates it gathered
- Sums it did
- 1 flag raised
When something goes wrong: stop, tell, fix.
AI agents get things wrong. Ours stop themselves, and that is in writing.
- Stop
After three failed tries it stops itself, and your team works the old way.
- Tell
The person you name is alerted at once. India's CERT-In directions of April 2022 give six hours to report.
- Fix
Proved on the failed case and twenty past jobs before it runs again. It is on us.
- Write up
With your yes we publish it as our error, never as “the AI did it”.
Our error14 October 2026
- What happened
- One quote went out at last month's rate
- Whose
- Ours. Our AI agent read an old rate card
- What we fixed
- It reads only the card your team marks current
- What changed
- A new check, proved on 20 past quotes
Sample data · Powered by Agentec
Your firm is never named.
Quote agentStopped itself
- Stopped
- After three failed tries
- Told
- The person you name, at once
- Fixed
- Proved on the failed case and 20 past jobs
- Meanwhile
- Your team works the old way
- Your stop
- One tap, any time
Sample data · Powered by Agentec
People, by name.
- Who works on your systemEach person is named in your proposal, with what they have built.
- Who signs, and who you escalate toEashan Gupta, founder. He signs the agreement and every note that needs an answer.
- Support hoursWritten into your agreement. Out of them, a failing AI agent still stops itself, and you hold the stop switch.
Our own AI policy.
It binds the AI agents we run for you and our own. As it stands on 13 October 2026.
May do alone
- Read
- Sort
- Draft
- Check
- File
Always waits for a person
- Spending money
- Signing
- Filing with an authority
- Deleting a record
- Any message outside the rules
- How we testOn past jobs with known answers, before it touches live work and after every change.
- Your dataIt trains your own model only, with your written yes, and never reaches another client.
What we won't build.
Five things we turn down, whoever asks.
No system that watches your staff.
No fake reviews.
No calls that pretend to be a person.
Nobody's data sold or shared.
No AI agent that sends on its own.
The terms, in plain words.
The termsAgentec Intelligence Private Limited
- 1
AI, said plainly
Our AI says it is AI. So does any AI agent that speaks for your firm, at the start of every conversation.
- 2
The price, in writing
One fixed price before any work, with the scope and the test it must pass. Nothing starts without your yes.
- 3
If you stop, or we do
Every account and line of code is in your firm's name from day one. Your data comes back as plain files on the date your agreement sets.
- 4
Your files
Where they sit and what leaves the building: your files and security.
Eashan Gupta
Founder, Agentec. Published 18 October 2026.
In every agreement we sign.
Bring us the problem, and hold us to this page.
It starts with 45 minutes on a video call with Eashan, with nothing to sign. This page comes with the written proposal that follows.
Eashan Gupta, founder. You meet him on the first call.