AI for MSMEs in India: Where to Start (a 90-Day Plan for Owners)

Short answer: Don't start with a tool. Start with one bottleneck that costs you money every week, like slow quotes, missed follow-ups, unanswered WhatsApp enquiries or manual invoices. Fix it with a low-cost AI tool your team already understands, measure it for 30 days, then add the next one. Most MSMEs can run their first useful AI pilot within a month.

I'm Vinay Pasricha. I run GoodSpace AI, a hiring company in Noida that has served about 1,200 corporate customers over seven years, and I wrote AI for Business Leaders. Most of the companies we work with are in manufacturing and auto, retail and FMCG, and healthcare and pharma, and many of them are MSMEs or supply to them. This guide is what I'd tell an owner who asks me, "Where do I actually begin?"

Why MSMEs should start now

MSMEs aren't a side story in India's economy. According to the Economic Survey 2025-26, they account for about 31.1% of GDP, 35.4% of manufacturing output and 48.58% of exports. As of 22 September 2026, the Ministry of MSME's Udyam dashboard showed more than 9.67 crore enterprises registered on the Udyam and Udyam Assist platforms. (If you're not sure you're an MSME: from 1 April 2025, a micro enterprise has up to ₹2.5 crore in plant and machinery and up to ₹10 crore turnover. The limits are ₹25 crore and ₹100 crore for small, and ₹125 crore and ₹500 crore for medium.)

Most owners I meet already know AI matters. They just don't know where it fits. A 2025 World Economic Forum playbook on AI for India's SMEs found the same thing in its interviews: most SMEs are "aware that AI deployment could benefit them but they are not sure how."

The risk isn't AI replacing you. In one of my short videos I tell the story of Rajiv (based on a real case, with the name and details changed), who runs a ₹180 crore company. A loyal client tells him his quality is good, but the newer suppliers are faster. Their quotes come the same day, and his take a week. Rajiv wasn't losing to AI. He was losing to a smaller rival that used AI better.

Waiting feels safe, but it isn't. While you debate, competitors are learning, and every month they learn compounds.

First, see your company as a brain

The idea at the centre of AI for Business Leaders is simple. Every company already runs on four capabilities: memory, reasoning, action and feedback. Most MSMEs run all four by accident:

• Memory: customer history, prices and supplier terms live in one person's head or phone.
• Reasoning: decisions go to whoever is loudest or most senior.
• Action: follow-ups depend on someone remembering.
• Feedback: lessons from a lost order or a bad hire vanish by next quarter.

AI doesn't fix chaos. It scales it. So the job isn't to "add AI". It's to pick one weak capability and let AI strengthen it: structured memory, faster reasoning, reliable action, closed feedback loops.

Where to start: six low-cost use cases

Pick one of these first, not all six.

1. Sales follow-up and quotes

The problem: enquiries arrive on calls, WhatsApp and email, and quotes take days.

What AI does: drafts quotes and follow-up messages from your price list, summarises call notes, and reminds salespeople who to chase today. A simple CRM with built-in AI, or even a shared sheet plus ChatGPT or Gemini, is enough to start.

Keep human: final pricing, discounts, key-account calls.

Measure: hours from enquiry to quote, and follow-ups missed per week.

At GoodSpace, we took this all the way. Every day our AI scans the internet for signs that a company may need to hire, like fresh funding, a new factory, rapid expansion or a surge of job posts on Naukri and LinkedIn. It researches each company in depth: its leadership, how fast it moves, its reputation and what its employees say about it. Then it writes a personal email to the right person, based on that research. We never send a generic email. It answers follow-up questions, nudges on WhatsApp or by phone, judges when a client is ready, and even handles the contract until the mandate is signed and handed to our delivery team. Our sales team now spends its time only on the people who drop off along the way.

You don't need all of that on day one. Start by letting AI draft a personal follow-up for every enquiry, based on what you know about that customer, and let your salesperson review it and hit send.

2. Customer support on WhatsApp

The problem: your customers are on WhatsApp, and so is your team, but replies depend on who is free.

What AI does: the free WhatsApp Business app already gives you a business profile, a catalogue, greeting and away messages, and quick replies. When volume grows, the paid WhatsApp Business Platform, usually set up through a provider, can answer common questions, share order status and hand off to a person.

Keep human: complaints, refunds, anything emotional.

Measure: first-response time and enquiries left unanswered.

3. Invoices, payments and accounts

The problem: manual data entry, invoice mistakes, and chasing payments.

What AI does: reads supplier bills, suggests ledger entries, flags mismatches, and drafts polite payment reminders. Your accounting software (TallyPrime, Zoho Books, Vyapar and others) is the natural place to start. Ask your provider which AI or automation features you already pay for. If your aggregate turnover has crossed ₹5 crore in any year since 2017-18, GST e-invoicing is already mandatory for you, so your invoice data is already digital, which makes this easier.

Keep human: approvals, tax judgement, bank transfers.

Measure: days sales outstanding and hours spent on data entry.

4. Hiring

The problem: hundreds of CVs for one role, candidates who accept and then vanish, and feedback like "candidate was good."

This is the area I know best, because it happened to us at GoodSpace. We didn't need AI everywhere. We needed it exactly where hiring kept breaking: screening, follow-up and judging people fairly. Now an AI interviewer meets every candidate, every offer is tracked until joining day, and managers decide on evidence. It was the same team, with less chaos and better hires.

What AI does for an MSME: writes clear job descriptions, screens applications against real requirements, schedules interviews, and follows up with selected candidates until they join.

Keep human: the final decision and the offer conversation.

Measure: days to fill a role and offer-to-joining rate.

For the full picture, including where AI fails in hiring and a 30/60/90-day plan, read my guide to AI in recruitment in India.

(Disclosure: I founded GoodSpace AI, which sells AI hiring tools.)

5. Inventory and demand

The problem: too much stock of slow items, stock-outs of fast ones.

What AI does: takes your past sales from Tally, your ERP or Excel and helps spot patterns: seasonality, festival spikes, slow movers. Start by asking an AI assistant inside Excel or Google Sheets to analyse your last 24 months of sales. Only then look at dedicated inventory software.

Keep human: purchase decisions and supplier negotiation.

Measure: stock-outs per month and value of dead stock.

6. Content and marketing

The problem: no time for product descriptions, brochures, social posts or Hindi and regional versions.

What AI does: drafts product copy, catalogue text, social posts and translations. Canva's AI features and any general assistant can do this today.

Keep human: brand voice, claims about quality or certifications, and a final read before anything goes out.

Measure: posts or catalogue pages shipped per week, and enquiries they bring in.

How to pick your first use case

Score each candidate on two questions: How much money or time will this save? and How easy is it for my current team to use? The WEF playbook recommends the same approach: start with use cases that have high impact and are relatively easy to implement. If a use case needs clean data you don't have, park it.

The 90-day plan

Days 1–30: Choose and prepare

• Week 1: List your top five weekly bottlenecks with the people who live them. Pick one, using the two questions above.
• Week 2: Write down how you'll measure it and record today's baseline (for example, "average quote time: 4 days").
• Week 2–3: Choose one tool. Prefer something already inside software you use. Set up a free trial or basic plan.
• Week 3–4: Write a one-page AI red lines note. It covers which customer and employee data never goes into public AI tools, and what always needs a human sign-off. Train two or three "champions" on the team.

Days 31–60: Pilot

• Run the pilot on real work with a small group (one salesperson, one branch, one product line).
• Hold a 20-minute weekly check: What did AI get wrong? Where did people bypass it? What did it save?
• Fix the process, not just the prompt. If the data is messy, cleaning it up is part of the work.

Days 61–90: Decide and scale

• Compare against your baseline. If the metric moved, roll it out to the rest of the team and write the new process down.
• If it didn't, stop, note why, and pick the next use case. That still counts as progress.
• Choose use case number two. The second one is always faster, because your team now knows how to do this.

This is the short version. AI for Business Leaders has the full 90-day plan, step by step.

Costs and tools

You don't need a big budget to start. Most MSMEs can begin with tools that have free tiers or low monthly plans. Prices change often, so check current pricing on each provider's site before you decide.

The bigger cost is your team's time in the first 60 days. Budget for it.

A note on data. India's Digital Personal Data Protection Act, 2023 now has its Rules, notified in November 2025. Most business obligations, such as notice, consent and security safeguards, apply from May 2027 under the 18-month phase-in. Build the habit now: don't paste customers' or employees' personal details into public AI tools, and know what data your vendors store. (This is general guidance, not legal advice.)

Common mistakes

1. Buying tools before naming the problem. A subscription nobody uses isn't a strategy.
2. Trying five things at once. Each gets a fifth of the attention it needs. One bottleneck at a time.
3. No baseline. If you didn't measure before, you can't prove anything after.
4. No red lines. Staff will paste sensitive data into free tools unless you tell them not to.
5. Leaving it to IT or a vendor. The owner has to know where AI fits. The vendor only knows their product.
6. Waiting for certainty. The companies that learn fastest win, not the ones that plan longest.
7. Going all-in too early. Experiment early, but invest at scale only when the technology is ready.

We made that last mistake ourselves. About four years ago we bet on AI voice agents at GoodSpace. It was too early for voice, and we lost money. Today much more of what we do runs on voice. It isn't everywhere yet, but it is getting there. So experiment with everything, but remember that everything happens in its own time. Robots are the same today. It isn't time to put them across your factory floor yet, but it is certainly time to experiment with them.

FAQ

How can AI help small businesses in India?

Mainly by saving hours on repetitive work: replying to enquiries, drafting quotes, entering invoices, screening CVs and writing product content. It also helps you spot patterns in your sales data. Start with one of these and measure it.

What are the best AI tools for MSMEs?

The best tool is usually one that sits inside software you already use: your accounting package, your CRM, WhatsApp Business, Excel or Google Sheets, plus one general assistant like ChatGPT, Gemini, Copilot or Claude. Check current pricing and free tiers before buying anything.

How do I implement AI in a small business?

Pick one bottleneck, set a baseline, choose one tool, write simple rules about data, pilot it for 30 days with a small group, then decide whether to scale it. The 90-day plan above lays out the steps.

How do I use AI in my business on WhatsApp?

Start with the free WhatsApp Business app: catalogue, greeting messages, away messages and quick replies. When enquiry volume outgrows one phone, move to the WhatsApp Business Platform through a provider, which can automate common answers and hand off to your team. Meta charges per message there.

Are AI agents useful for an MSME?

They can be, for narrow, repeatable jobs such as following up on quotes or tracking an offer until joining day. Get one simple workflow working reliably first. An agent automating a messy process just makes the mess faster.

Where to go next

If this guide was useful, the full framework is in my book AI for Business Leaders. It covers the company brain (memory, reasoning, action, feedback), where AI belongs and where it doesn't, and a 90-day plan you can actually run.

For more on AI in Indian companies, including short videos and a section by sector, see AI for Indian Business.

For founders and senior leaders who want to work through this for their own company, I run the AI Leadership Course. It's a six-week cohort, capped at 18 operators, and ends with your own AI red lines, handoff protocol and team redesign.

Watch:

• He Wasn't Losing to AI. He Was Losing to a Company That Used It Better (the Rajiv story, 59 seconds)
• How We Figured Out Where AI Belongs in Hiring (the GoodSpace example)

Vinay Pasricha is the founder of GoodSpace AI and the author of six books, including AI for Business Leaders. Watch his videos on YouTube.

Published 26 September 2026 · Last updated 28 September 2026