To hire an AI agent, name one process for it to own, one number that proves it worked, and the exact point where a person takes over. Then check that it writes to your CRM, agree who answers when it fails, and ask the vendor what they will not promise. Sign a pilot, not a transformation.
An agent that does everything does nothing measurable. Before you talk to a vendor, write down the single process you want off your team's plate: inbound WhatsApp enquiries from the website, the first call to every new lead, the morning LinkedIn outreach, appointment reminders for tomorrow's clinic list.
Then write the process as it happens today: who receives it, what they ask, where they record it, what happens when they are busy. A vendor who can repeat that back to you has understood the job. One who starts with the technology has not.
Pick the metric before go-live, and pick one. It should be a number you already track, or could track with one extra field in the CRM.
Avoid what the agent can inflate on its own, such as messages sent or conversations started. Those measure activity, not outcome, and a vendor who leads with them is showing you what is easy to count.
The most expensive mistake in an agent project is a vague answer to when a human steps in. Decide before you sign: which topics the agent must never handle alone, what hour it stops trying, and who receives the conversation when it escalates.
An agent that works but writes nowhere has moved the work rather than removed it. Before signing, ask where a conversation lands, in which fields, and what happens when the same person messages twice.
Every agent fails sometimes: it misreads a message, quotes an old price, promises something you do not offer. What matters is that the failure has an owner, in writing.
That owner is you, not the vendor. The agent speaks with your brand, so your team replies to the customer; the vendor's accountability is the system, the review and the alerting.
The questions that separate a partner from a reseller are the negative ones. A vendor who can say where an agent is the wrong tool is telling you the truth about where it is right.
A sensible first month is one process, one channel, one number to judge it by, and a review at the end of it. That is what a rental model is for: you can stop. Standard cover for one agent on one channel starts at ₹999 per month, and Pro at ₹14,999 is for several agents or channels working together. See pricing for what each includes.
At the end of the pilot, decide with the number rather than the feeling. If the agent did not move the metric, change the process or stop. If it did, add the next channel, usually the one where the same conversations already happen. If you would rather talk it through first, contact us with the process you have in mind.
Rental pricing is the common model: one agent on one channel from ₹999 per month on Standard, and ₹14,999 on Pro for several agents or channels. Setup and the CRM connection are included; taxes and channel usage charges are not.
The build is usually days rather than months: an audit, configuration, testing on real conversations, then go-live. Approvals outside our control, such as WhatsApp business verification, sit separately and should be started on day one.
That is an ownership question, not a technology one. Agree before signing who replies to the customer, how fast you are told, who can pause the agent, and whether the vendor reviews transcripts with you.
No. A working agent writes into the CRM you already use, whether that is HubSpot, Zoho, Salesforce, Google Sheets or your own backend, and the field mapping should be checked before go-live.