AI Automation Agency in Lucknow, Working Across India
Codingclave, run by founder Ashish Sharma from Lucknow since 2017, builds WhatsApp and website chatbots and AI agents for Indian businesses, from ₹35,000 for a rule-based WhatsApp bot. Start with a fixed-price pilot on one task in 2–4 weeks.
200+ projects since 2017. Scope and price in writing before you pay. We reply within 2 hours (Mon–Sat 10:00–19:00 IST); free quote in 24 hours.
Our Build Prices at a Glance
Build prices are one-time and fixed once the scope is agreed. Running costs come on top: hosting from about ₹600 a month, AI model usage, and Meta's WhatsApp fees.
Prices checked on 2026-10-18. The full breakdown, with SaaS plans and AI model fees, is in our chatbot price guide.
Six Workflows We Build for Indian Businesses
These are the kinds of jobs we build, not client stories. Each card shows what starts it, what it does, who signs off, what you get and what it costs.
WhatsApp chatbot development
- Starts when
- A customer messages your WhatsApp Business number, day or night.
- What it does
- Answers the questions you get every day (fees, timings, location, documents needed), asks for name, requirement and budget, and hands the chat to a staff member with the full transcript when it can’t help or the customer asks for a person.
- Who approves
- You approve the answer list before launch. Your staff take over any chat the bot passes on.
- What you get
- Answered enquiries, and a lead with the chat attached.
The number has to be on Meta’s WhatsApp Business Platform. With PayPerWA it cannot also stay active on the regular WhatsApp or WhatsApp Business app.
Website AI chatbot that answers from your documents
- Starts when
- A visitor types a question into the chat box on your website.
- What it does
- Finds the answer in the documents you give it (FAQ, prospectus, price list, policies), replies in plain words, and says it does not know, offering a call-back, when those documents do not cover the question.
- Who approves
- You choose which documents it may use. Questions outside them go to your staff.
- What you get
- An answer, the document it came from, and a list of unanswered questions you can add to your FAQ.
Lead capture and follow-up
- Starts when
- Someone taps your Click-to-WhatsApp ad or fills the form on your website.
- What it does
- Asks your qualifying questions (course or product, budget, area, when they want to start), saves the lead in your CRM or a Google Sheet, alerts the right salesperson, and sends an approved follow-up message after the number of days you set if the lead goes quiet.
- Who approves
- You approve every follow-up template, and Meta reviews each template before it can be sent.
- What you get
- A lead in your CRM with the chat attached, and a reminder to the salesperson who owns it.
AI agent development for invoices and documents
- Starts when
- A supplier bill arrives as a PDF by email, or as a photo on WhatsApp.
- What it does
- Reads the supplier name, GSTIN, invoice number and date, HSN or SAC codes, taxable value, GST and total; checks that the GSTIN is in the right format and that the sums add up; flags anything that does not match.
- Who approves
- Your accountant checks each draft before anything goes into your accounts.
- What you get
- A draft row in Excel, or an import file for your accounting software. Nothing is posted or paid on its own.
Appointment reminders on WhatsApp
- Starts when
- A booking is made in your calendar, clinic software or booking sheet.
- What it does
- Sends a WhatsApp reminder before the appointment with buttons to confirm or reschedule, updates the booking when the customer taps, and follows up with people who did not turn up.
- Who approves
- You approve the reminder wording and timing. Staff see every change.
- What you get
- Confirmed or moved bookings, and a daily no-show list.
CRM updates from every chat
- Starts when
- A chat on WhatsApp or your website ends, whether the bot or your staff handled it.
- What it does
- Creates or updates the contact, moves the lead to the right stage, adds a short AI summary as a note in Zoho CRM, HubSpot or your own CRM, and sends your sales head a daily list of new and stuck leads.
- Who approves
- Your sales head sets the stage rules. Staff can correct any AI note.
- What you get
- A CRM that stays up to date without anyone retyping chats.
We suggest starting with one of these and adding the next once the first is running. Your written quote lists exactly what is in and out. Prices checked on 2026-10-18.
Our Fixed Build Prices and Timelines
One fixed price for the build, agreed in writing before we start. Running costs are listed separately, because you pay most of them to Meta, the AI provider and your cloud host.
- Build
- ₹35,000–₹1,25,000
- Monthly
- ₹600–₹2,000
- Timeline
- 2–3 weeks
- Build
- ₹1,25,000–₹3,00,000
- Monthly
- ₹2,000–₹8,000
- Timeline
- 4–6 weeks
- Build
- ₹2,00,000–₹5,50,000
- Monthly
- ₹5,000–₹25,000, plus Meta and PayPerWA message fees
- Timeline
- 5–8 weeks
- Build
- ₹2,50,000–₹6,00,000
- Monthly
- ₹8,000–₹40,000
- Timeline
- 6–10 weeks
- Build
- ₹1,00,000–₹3,00,000
- Monthly
- Depends on volume; worked out in your quote
- Timeline
- 3–5 weeks
- Build
- ₹3,00,000–₹7,00,000
- Monthly
- Depends on volume; worked out in your quote
- Timeline
- 6–10 weeks
Prices checked on 2026-10-18, from our chatbot price guide and AI agents guide. As in the price guide, the monthly column covers hosting, AI model fees and a maintenance retainer where one applies; Meta and PayPerWA message fees are extra, and the made-up example further down shows a bill without a retainer. Timelines run from agreed scope to a working version; Meta's verification of a new WhatsApp number can add time. Large, multi-system assistants for banks or insurers are quoted separately.
What moves the price
- How clever it has to be: fixed menus, recognising what people mean, or AI answers written from your documents
- Each system it writes to (CRM, calendar, payments, your own software): ₹15,000–₹80,000 each
- Channels: one channel is cheapest; the same bot on your website, WhatsApp and Instagram can double the build
- How much it has to read: 20 FAQs is simple; thousands of pages add ₹50,000–₹3,00,000
- How many languages, and how many answers you want checked before launch
Our terms, in writing
- Fixed price once the scope is agreed
- 40% at start / 40% at an agreed mid-project milestone / 20% at launch
- Source code and IP yours on final payment
- Free bug fixes for 6 months after launch
What other Indian agencies publish
So you can compare. These are prices printed on each agency's own page, checked on 2026-10-18. We have not checked what each price includes, so compare the written scope, not just the number. Some are lower than ours.
| Source | Published prices |
|---|---|
| Edesy (WhatsApp chatbot page) | Quick chatbot ₹50,000 one-time plus ₹3,000 a month maintenance; advanced ₹1,25,000 plus ₹6,000 a month maintenance; enterprise ₹2,00,000+ plus ₹10,000+ a month maintenance |
| PS Digitise (AI chatbot page) | Ongoing support from ₹8,000 a month; its enquiry form offers budgets from under ₹35,000 to above ₹1,50,000 |
| NovaX AI (Lucknow page) | Packages of ₹6,000–₹20,000, each bundled with a website; the AI chatbot and WhatsApp automation sit in the ₹15,000–₹20,000 package |
Start With a Fixed-Price Pilot (2–4 Weeks)
A small paid build on one task, measured against your own data, so you see it working before you commit to a full project.
What a pilot covers
- One workflow, one channel, one language
- Built and tested on your own sample messages or documents; names and numbers can be removed first where the task allows
- A written scope that says what the pilot must handle and what is out
- A fixed price once the scope is agreed, paid 40% at the start, 40% at an agreed midpoint milestone and 20% at go-live
- After the pilot you extend it or stop there. The code is yours on final payment
Rule-based bots and single-task agents are small enough that the pilot is the first full version: from ₹35,000 for a rule-based WhatsApp or website bot, from ₹1,00,000 for a single-task AI agent. A full AI chatbot takes 5–10 weeks, so its pilot is a narrow first version, such as answering admission questions only, and the written quote gives that pilot its own fixed price. Prices checked on 2026-10-18.
Signs of a good first task
- It happens every day, or many times a week
- It takes your staff more than a few minutes each time
- It follows rules a new joiner could learn in a day
- A person can quickly check the result before it goes out
How a pilot runs
- 1Send the task. A short screen recording or a few sample chats: what starts the job, which apps it touches and how often it happens.
- 2Free quote in 24 hours. One fixed figure and a written scope that says what is in and out, or a note that a tool you already pay for can do it.
- 3Accounts you can own. The WhatsApp number, AI model account and server can be opened in your business name, with a spend cap you set, so the bills and logins stay with you.
- 4Build on your sample data. We build against your sample messages or documents. The midpoint milestone for the second 40% is written into the quote.
- 5Go-live and handover. You get the code and the logins on final payment. Bug fixes are free for 6 months after launch.
Running Costs You Pay to Meta, Your AI Provider and Your Host
The build is one bill. These are the others, most of them paid straight to the provider, from accounts that can be in your name.
AI model usage
Per 1,000 typical conversations of about 12,000 tokens in and 1,500 out (the assumption in our price guide), at about ₹95 to the US dollar. Rule-based bots use no AI model, so this line is zero for them.
Hosting
About ₹600–₹1,000 a month for a rule-based bot on one small server, and ₹1,500–₹3,000 for a small AI chatbot with its document search on the same server. A busier setup with a managed database, backups and two app servers on AWS Mumbai or Azure Central India is about ₹8,000–₹25,000 a month.
The server can sit in your own cloud account, so the bill comes to you and the data stays under your control.
Hosting plans and prices comparedWhatsApp messages
Meta's charge per message in India, before GST:
Inside Meta's free entry point window (up to 7 days), which opens when someone messages you from a Click-to-WhatsApp ad or a Facebook Page button and you reply within 24 hours, Meta charges nothing. On top of Meta, PayPerWA charges ₹0.20 per message, GST included, on every message you send, bot replies included. No monthly plan.
Checked on 2026-10-18: model prices on claude.com/pricing and OpenAI's pricing page (Haiku 5.5 price for prompts up to 100,000 tokens) and the dollar at Wise's mid-market rate of about ₹95. The service-message rule (1,000 free a month per number, then the utility rate, from 1 October 2026) and the free entry point window of up to 7 days are from Meta's non-template pricing page and pricing page as updated on 30 September 2026; the pricing page links Meta's INR rate card. The India rates above are the ones in our WhatsApp API pricing comparison; rates change, so check Meta's card before you budget. Hosting figures are our own estimates.
A made-up example, not a client case study
A WhatsApp AI bot for a coaching centre, one month
A Lucknow coaching centre gets 600 WhatsApp enquiries a month from its website and Google listing, not from ads. The bot sends about 5 replies per enquiry, so 3,000 replies. It also sends 200 demo-class reminders and 400 follow-ups to students who went quiet. It runs on a small, cheap AI model and one small server, and sends through PayPerWA.
This applies Meta's 1,000 free service messages a month for one business number, so 2,000 of the 3,000 replies are charged at the utility rate from 1 October 2026. If enquiries came from Click-to-WhatsApp ads, replies inside the free entry point window would cost nothing from Meta; PayPerWA's ₹0.20 still applies to each message. If every answer went to Claude Sonnet 5.5 instead, model usage would be about ₹2,200. Before any support retainer. Figures as checked on 2026-10-18.
Price my message volume on PayPerWAYour Data: DPDP Act Basics, Where It Is Stored, Who Sees Prompts
An AI bot reads your customers' messages, so you should know where they go before you sign anything, with us or anyone else.
DPDP Act basics
- The Digital Personal Data Protection Act was enacted on 11 August 2023, and its Rules were notified on 14 November 2025, with an 18-month phase-in.
- Consent-manager rules apply after 12 months (about November 2026). Most duties, including notices, consent, security safeguards and breach reporting, apply after 18 months (about May 2027).
- You are the Data Fiduciary: you decide why and how customer data is used. An agency building for you works as your Data Processor, and the Act lets you use one only under a valid contract.
- Consent has to be clear and informed, and people can withdraw it at any time. They can also ask to see, correct or erase their data, and you must respond within 90 days.
- A breach must be reported to the Data Protection Board and to the people affected. Failing to keep reasonable security safeguards can cost up to ₹250 crore; failing to report a breach up to ₹200 crore.
Where your data is stored
- The bot’s server and database can sit in a cloud account in your name. For an Indian business we suggest an Indian region, such as AWS Mumbai or Azure Central India.
- WhatsApp messages pass through Meta’s Cloud API. Meta lets a business choose in-country storage of message content per phone number, and its own example uses India; it can only be set while the number is unregistered, and content may still be processed in Meta data centres abroad.
- OpenAI offers India storage at rest on supported endpoints only to approved customers (you apply through its sales team and agree modified retention terms), and its table marks India “No” for regional processing, so the model may read your text outside India.
Who can see prompts and chats
- Your staff who have access to the dashboard or inbox.
- Our team, during the build and while we support it. If the accounts are in your name, you can remove our access at handover.
- The AI provider’s systems, under its API terms. Neither OpenAI nor Anthropic trains on API data by default; how long each keeps it is in the FAQ below.
- Meta, for WhatsApp messages, and the WhatsApp platform you send through (PayPerWA or another provider), which keeps the inbox and message logs.
Guardrails we can build in
- A person approves before money moves, a discount is offered or a marketing message goes out.
- Aadhaar and PAN numbers masked before text goes to an AI model, wherever the task does not need them.
- A spend cap on the AI model account, set in your own account.
- A log of every chat and draft, and an off switch your staff can use.
General information, not legal advice. Checked on 2026-10-18: the Press Information Bureau's DPDP Rules backgrounder (17 November 2025), OpenAI's data controls guide, Anthropic's articles on model training and data retention, and Meta's Cloud API local storage page. Our DPDP Act compliance guide covers the Act in more depth.
When You Don't Need a Custom Build
A build only makes sense when it pays back. In these three cases it usually does not.
Your bot only needs fixed menus and buttons
If it shows menus, takes bookings or tells people their order status, and you have fewer than about 1,500 conversations a month, a monthly flow-bot tool is cheaper than a build. AiSensy’s chatbot add-on, for example, was ₹2,500 a month for 5 bots, on top of an AiSensy plan, on its pricing page on 18 October 2026. PayPerWA, our own platform, has a drag-and-drop chatbot builder with no monthly plan.
Compare WATI, Interakt and AiSensyThe real problem is missing software
If staff retype data because there is no proper system, fix that first. Our ready-made clinic software is ₹29,999 one-time, the real estate CRM ₹39,999 and GST billing ₹14,999. For anything else, custom software comes before automation.
Custom software, CRM and ERPThe task happens a few times a week
If it is rare, quick, or needs judgement on every case, the build will not pay back. Leave it manual, or use a feature inside an app you already pay for. We will tell you if that is the answer.
When SaaS beats a custom chatbotPrices checked on 2026-10-18.
If a ready-made monthly plan fits your size, Ashish will tell you so on the first call instead of selling you a build.
A Founder-Led AI Automation Agency in Lucknow
You talk to the person who scopes and signs off your build, not a sales team.
Codingclave Development LLP
- Run by founder Ashish Sharma since 2017, with 200+ projects for clients in India and abroad.
- Office: 280/10A, Vrindavan Colony, Lucknow 226029. We work with businesses across India over WhatsApp and video calls.
- 4.9 on Google from 92 reviews, Top Rated on Upwork with 100% Job Success and 5.0 on Freelancer from 58 reviews. Read them on our reviews page.
Our work page shows websites and software we have built, not AI case studies, and we would rather say so than dress it up. More about us on the about page, and our wider Lucknow work on web development in Lucknow.
Choosing an AI chatbot or AI agent development company in India: four checks
- 1Is the price in writing, with running costs (hosting, AI model, WhatsApp fees) clearly in or out?
- 2Whose name are the WhatsApp number, AI model account and server in?
- 3Do you get the code and a handover document, and when?
- 4Who approves payments and customer messages, and what happens when the bot is unsure?
Ask these of every agency you talk to, us included. A big number on a website tells you little unless you can check who the work was for.
AI Automation Questions
What Indian business owners ask before automating their first task.
How much does AI automation cost in India?
At Codingclave, a rule-based WhatsApp or website FAQ bot costs ₹35,000 to ₹1,25,000 to build; a WhatsApp AI chatbot ₹2,00,000 to ₹5,50,000; an AI chatbot that answers from your own documents ₹2,50,000 to ₹6,00,000; a single-task AI agent, such as one that reads supplier invoices, ₹1,00,000 to ₹3,00,000; and a multi-step agent ₹3,00,000 to ₹7,00,000. Running costs are separate: hosting from about ₹600 a month, AI model usage (a few hundred rupees per 1,000 conversations on small models) and Meta’s WhatsApp charges. Prices checked on 18 October 2026.
What does an AI automation agency actually do?
It takes a task your team repeats, such as answering the same WhatsApp questions, typing invoice details into a sheet or chasing leads, and builds software that does the repeated part, calling an AI model only where text needs reading or writing. It connects that to the tools you already use, decides with you who approves what, and should tell you when a ready-made tool is cheaper than a build.
What is the best AI agent for automation?
There is no single best one. It depends on the task, the language your customers write in and how many messages you get. A small, cheap model such as GPT-4o mini or Claude Haiku 5.5 handles most FAQ and sorting work; harder reasoning goes to a larger model such as Claude Sonnet 5.5 or GPT-4o, at roughly 15 to 20 times the cost per conversation. The sensible test is to try a small and a large model on a sample of your real messages and choose on results and cost, not brand.
What does a WhatsApp chatbot from Codingclave cost, with Meta and PayPerWA fees?
The build is ₹35,000 to ₹1,25,000 for a rule-based bot, or ₹2,00,000 to ₹5,50,000 for a WhatsApp AI chatbot. Each month you also pay Meta per message (in India ₹0.8631 for marketing and ₹0.115 for utility, before GST; from 1 October 2026 bot replies are charged at the utility rate too, after the first 1,000 service messages a month per number) plus ₹0.20 per message, GST included, if you send through PayPerWA. A bot sending 3,600 messages a month, mostly replies, pays about ₹598 to Meta before GST and ₹720 to PayPerWA, before model and hosting costs.
What is the difference between a chatbot and an AI agent?
A chatbot answers people in a chat. An AI agent finishes a task using tools: it reads a bill, checks it and writes a draft entry, or qualifies a lead and books a call in your calendar. Many projects are both: the chat is the front, and the agent does the work behind it. Either way, we suggest a person approves any step that moves money or messages customers.
Is my customers’ data safe, and does ChatGPT or Claude train on it?
Through the API, not by default. OpenAI says data sent to its API is not used to train its models unless you opt in, and keeps abuse-monitoring logs for up to 30 days. Anthropic says it does not use API inputs or outputs for training by default and deletes them within 30 days, longer if a chat is flagged under its usage policy or the law requires it. The consumer ChatGPT and Claude apps work differently, so staff should not paste customer data into them. The model account can be in your name with a spend cap, and under the DPDP Act you remain the Data Fiduciary.
Does it work in Hindi and Hinglish?
Current GPT and Claude models read and write Hindi and Hinglish, including Hindi typed in English letters. What matters is how they cope with your customers’ spellings, short forms and product names, so test the bot on a sample of your real messages before launch. A rule-based bot can show its menus in Hindi and English, but it will not understand free-typed Hinglish.
How long does a pilot or a full build take?
A pilot takes 2 to 4 weeks: one workflow, one channel, one language. Full builds take 2 to 3 weeks for a rule-based bot, 4 to 6 weeks for booking or CRM flows, 5 to 8 weeks for a WhatsApp AI chatbot, 6 to 10 weeks for an AI chatbot on your documents, and 3 to 10 weeks for an AI agent. Meta’s business verification and template approvals for a new WhatsApp number can add time, so they start on day one.
Can I keep my WhatsApp number?
Usually, yes, but it has to move onto Meta’s WhatsApp Business Platform. With PayPerWA the number cannot also stay active on the regular WhatsApp or WhatsApp Business app. If your team wants to keep using the app on that number, we check whether Meta’s coexistence option (the app and the API on one number) works for you, or suggest a second number for the bot.
What happens when the bot gives a wrong answer?
Three things limit it. The bot answers only from the documents you approve and says it does not know when they do not cover the question. It hands the chat to a person, with the transcript, when it is unsure or the customer asks. And anything that sends money, a discount or a marketing message can be set to wait for staff approval. Every chat is logged, so you can spot a wrong answer and correct the source document; bugs in our work are fixed free for 6 months after launch.
Sources checked on 2026-10-18: Meta's non-template pricing page, OpenAI's data controls guide, Anthropic's training and retention articles, and claude.com/pricing.
Related Pages
Our price guides, the WhatsApp platform we run, and the software you may need instead of a bot.
Show Us the Task Your Team Repeats Most
Send a screen recording or a few sample chats. You get one fixed figure and a written scope that says what is in and out.
Reply Within 2 Hours
We reply within 2 hours (Mon–Sat 10:00–19:00 IST); free quote in 24 hours. No waiting days for a callback or email.
100% Free Consultation
Tell us your idea. We'll give you an honest estimate, tech recommendations, and a roadmap — free.
200+ Projects Shipped
Delivered worldwide since 2017, founder-led by Ashish Sharma.
Upwork JSS
Projects