AI Automation Cost in India: Building and Running It

Full disclosure: we build AI automation. The sellers quoted below include firms we compete with; they are listed alphabetically, every figure is one they published, with the date we read it, and nobody paid to appear. Our own starting prices are on our pricing page (a logged workflow from ₹25,000, a WhatsApp chatbot from ₹50,000, an AI voice agent from ₹2 lakh, before GST); the other numbers here are sellers' published prices or arithmetic from assumptions we state, which you can change.

This guide is for owners and operations heads who have been quoted anything from a few thousand rupees to several lakh for "an automation" and want to know why.

What are you paying for when you buy an AI automation?#

Every automation carries three bills, and quotes often show only the first:

• The build, once. Designing the rule, connecting the systems, testing it on real cases and handing it over.
• Upkeep, every month. APIs change, templates get rejected, a model is retired, the team wants one more step. Someone has to watch the runs and fix what breaks.
• Usage, per run. Model calls, WhatsApp messages, voice minutes and hosting, billed by the providers whatever the build cost.

What you are buying also varies more than the word "automation" suggests. Our guide to AI agents for business sets out how a rules-based automation, an assistant and an agent differ. For pricing, five scopes cover most requests, from a rules-based workflow with no AI in it to an agent working across several systems; the estimate table below defines each.

What do Indian sellers publish for AI automation?#

Few of the agency pages we read on 4 October 2026 publish prices for this work. Four that do, two on their own price pages and two as market ranges in their guides, with the figures as published:

All read on 4 October 2026. Dollar prices converted at ₹95.99, the FBIL reference rate for 1 October 2026 shown on the Reserve Bank of India's website↗ (opens in a new tab). Each seller defines its scopes differently, so compare what is included, not only the figure, and check whether a price includes GST.

The structure is the same everywhere: a one-time price per workflow, bot or agent; a monthly retainer for upkeep; and usage billed separately at the providers' rates. The spread inside each band is the scope, which the next section turns into numbers.

How do you estimate an AI automation build yourself?#

The same way as any software: effort in person-months multiplied by the cost of one person-month. Our custom software cost guide derives both rates from Kochi salary data and Clutch's India rate band: about ₹1 to 2 lakh per person-month for a lean Kerala team (twice the salary of the people doing the work) and ₹3.8 to 7.5 lakh at Clutch's India band of $25 to $49 an hour. Applied to the five scopes:

Our planning assumptions, checked on 4 October 2026, excluding GST, usage and upkeep. Not a quote and not our fees; confirm with written quotes.

The published prices above sit inside or close to the lean-team column almost row for row, and the Clutch column shows what the same scope costs from a firm billing $25 to $49 an hour. The example in the first row is a real one: the admissions page we built for DiYES International sends every enquiry to the school's CRM tagged with the grade and the campaign it came from, a rules-based flow with no AI in it.

A worked example makes the effort concrete. Take a rule written in plain words: "When a new Meta lead arrives for the Kochi branch, give it to the least-loaded counsellor there, set their follow-up for tomorrow at 10 am and send our welcome WhatsApp template." Built for a business whose CRM accepts new leads through an API, that rule needs, by our estimate:

• the Meta lead connection: about a day;
• least-loaded assignment and the follow-up, read from and written to the CRM: about two days;
• the WhatsApp template and its send, once Meta approves the template: about a day;
• retries, a run log and a test mode: two to three days;
• tests against 50 past leads: about a day.

That is 7 to 8 working days, plus a day or two of slack for the template and the CRM's quirks: about 0.4 to 0.5 person-months, or ₹40,000 to ₹1 lakh from a lean team. Add an AI step that summarises each first call and proposes the next follow-up for the counsellor to approve, with a test set of past calls and the approval screen, and you add another 5 to 10 working days. Inside a CRM that already has these parts, the same rule is a description and a save in its builder, which is why the system you already run moves the price more than anything else.

What does a production automation layer contain?#

The demo of an automation is the happy path: the lead arrives, the message goes. Most of the build cost is everything else, and each line below is work someone has to build and test:

• A job queue, so an automation never blocks the person saving the record
• Retry rules per action, and a worker server that runs apart from the app
• A scheduler, and rules for what must never be retried
• A run log people can actually read, and screens to search it
• A dry-run mode for every action
• Guards that stop one rule setting off another forever
• Per-lead state, so a welcome message goes once and never at 2 am
• The AI kept outside the running path, with a person approving what it writes

An AI assistant whose answers follow what the asking user is allowed to see, each showing the filters it used, is work of the same kind, which is why the two assistant scopes in the estimate cost more than a chatbot demo.

Use the list as a checklist. A quote that covers the trigger and the actions alone leaves these lines out, and they are what keeps an automation running after the person who built it has moved on.

What does an AI automation cost to run each month?#

Usage and upkeep depend on volume, so here are the reference points, each worked out in a guide of its own:

The model bill moves a hundredfold with the model you choose, so test whether a cheaper one is good enough before defaulting to the biggest. And price review time too: if a person spends four minutes checking a draft that took five to write by hand, the AI step saved one minute, as our AI agents guide puts it.

What makes an AI automation cost more or less?#

The first two rows decide most of it. Anything a customer sees needs the approval, consent and hand-off work that internal tools can skip, and an AI that acts on its own needs limits and logs that an AI that drafts does not.

How do you keep an AI automation's cost down?#

• Automate the rule before the AI. The cheapest scope is the one with no model in it, and most of the early savings come from there. Our AI agents guide has a matrix for deciding which tasks need AI at all.
• Start with one workflow and a 30-day pilot. Measure the time it takes today, run the automation with approval on everything, and compare.
• Try the AI your CRM already ships. If you run Zoho, HubSpot or Salesforce, our guide to AI in CRM shows what each includes and charges before you pay for a build.
• Put the reliability rows in the quote. Ask, in writing, which of the production-layer rows above are included.
• Keep the accounts in your name. The WhatsApp Business account, the model API keys and the automation platform should belong to your business, so you can change who maintains them.
• Cap the usage. Set a monthly spending limit with each model provider, and an alert before it.

If what you need is mostly lead capture and routing inside a CRM, that is usually CRM work rather than a separate automation project, and it may already be a setting in the CRM you have.

Frequently asked questions#

How much does it cost to build an AI chatbot in India?#

Published one-time prices for a single bot with no integrations run from about ₹50,000 to ₹1.5 lakh in Neogen Media's guide, and Matrics Consulting lists a support agent at $499, about ₹47,900, including set-up and one month of maintenance (both read on 4 October 2026). A bot that answers from your own live data, inside each user's permissions, costs more to build. Usage comes on top: model calls, messages and hosting; what a WhatsApp chatbot costs to run each month is worked out in our guide to WhatsApp AI chatbots.

What does WhatsApp automation cost in India?#

Two bills: the build and Meta's messages. A rules-based WhatsApp flow sits at the simple-workflow end of the published prices, from about ₹20,000. Since 1 October 2026, Meta charges ₹0.115 plus GST for each service reply to an Indian number, after 1,000 free a month per business number, and marketing templates cost about a rupee each with GST.

How much does it cost to build an AI agent?#

Innovatrix Infotech's guide puts a production AI agent at ₹3 to 10 lakh one-time, and Matrics Consulting prices single-purpose agents at $499 to $999 with set-up and a month of maintenance. By a person-month model, an agent across several systems, with approvals, limits and logs, takes three to six person-months: about ₹3 to 12 lakh from a lean Kerala team, before usage.

Does an AI automation need a monthly retainer?#

Something has to cover upkeep, because APIs, templates and models change under any automation. Published retainers run from ₹20,000 to ₹1.5 lakh a month, depending on how many workflows are covered and how fast fixes must happen. A team with an in-house developer can do the upkeep itself, if that person has the access and the run logs.

What to do next#

List the three tasks your team repeats most and put each through the five scopes above. If a task is a rule, it belongs at the cheap end; if it needs AI, price the approval step and the review time with it. For a second opinion on a quote or a scope, tell us what the automation has to do, and we will say which scope it is and what a fair build looks like, including when the answer is a setting in software you already pay for. Or talk it through with Althameem Ali, our co-founder, in a one-hour consultation.

Changelog#

• 2026-10-09: our own starting prices are now public on phacetwo.com/pricing, so this line gives them.