← Back to Portfolio

Where AI Actually Helps a ₹10Cr Indian SME
(and Where It Doesn’t)
— A 5-Decision Framework

Every Indian SME founder is getting the same advice in 2026: “You must adopt AI or get left behind.” What nobody tells you is where AI actually helps a ₹10Cr business — and where it will waste your time and money. Here is the honest framework I use with Indian SME founders.

Decision 1: Is your problem a repetition problem or a judgment problem?

AI is extraordinary at repetition. It is mediocre at judgment.

If your team is doing the same task more than 20 times a week — answering the same customer queries, generating the same reports, copying data from one sheet to another, writing the same type of email — that is a repetition problem. AI helps here. Measurably, quickly, cheaply.

If your problem is: which vendor do I trust, should I extend credit to this customer, is this employee performing or coasting — that is a judgment problem. AI can give you data to inform that judgment. It cannot make it for you.

Rule: Before spending a rupee on AI, write down your top 5 operational pain points. Mark each R (repetition) or J (judgment). Start AI only on the R’s.

Decision 2: Do you have the documentation for AI to work from?

This is the question no one asks — and it is the single biggest reason AI pilots fail in Indian SMEs.

AI tools work by reading your existing documentation and generating outputs based on it. If your SOPs live in three people’s heads and your process guides are last updated in 2019, the AI has nothing to work from. It will hallucinate, give generic answers, and your team will abandon it in two weeks.

The companies getting real value from AI in 2026 are the ones that documented their operations first — not because they planned for AI, but because good documentation is just good operations.

Rule: Run a documentation audit before any AI pilot. If less than 60% of your core processes are written down in current, accurate form — start there. The AI investment comes after.

Decision 3: Is your founder a bottleneck in the process you want to automate?

This sounds obvious. It almost never gets asked.

A common pattern in ₹5–50Cr Indian businesses: the founder approves every purchase order above ₹5,000. Every vendor payment. Every customer discount. Every hire shortlist. The team has learned to wait. No one moves without sign-off.

You cannot AI-automate a process where the founder is the critical node — because the AI cannot replicate the founder’s judgment, and the founder hasn’t documented their decision criteria anywhere. Adding an AI dashboard makes it look modern but doesn’t solve the bottleneck.

Rule: Map who approves what in your top 5 processes. If the answer is “the founder” more than twice — fix the delegation structure first. Then add AI.

Decision 4: Which of these 5 use-cases actually fits your business right now?

These are the five use-cases that consistently deliver ROI in the ₹5–50Cr segment in 2026:

  1. Internal knowledge search — your team asks an AI assistant “how do we handle a Tier-2 vendor dispute?” and gets your actual SOP as the answer. Prerequisite: documented SOPs.
  2. Customer query handling (first response layer) — AI drafts the first response to standard queries; your team reviews and sends. Frees 2–4 hrs/day per customer-facing team member.
  3. Inventory/demand signal summarisation — AI reads your sales data weekly and produces plain-English output: “Product X is trending 18% above last month. Current stock runs out in 11 days.” Founders act on this faster than a spreadsheet.
  4. Meeting notes and action tracking — AI transcribes your weekly ops meeting, extracts action items, sends each owner a reminder. Measurable impact on execution rates in teams of 10–50.
  5. Proposal / quotation first draft — AI takes your specs, pricing, and customer requirements and generates a first draft. Your team edits and sends. Cuts proposal time from 3 hours to 45 minutes.
Rule: Pick one. Pilot for 60 days. Measure before/after in hours saved or error rate. Only then pick the second.

Decision 5: What does “success” look like in 90 days — and who owns it?

Every successful AI implementation has two things in common: a clear metric defined upfront, and one named human owner who is not the founder.

“We will reduce proposal preparation time from 3 hours to 1 hour by 1 August, and Priya owns it.”

That is a success definition. “We will use AI more” is not.

Rule: Write one sentence: “By [date], [metric] will change from [X] to [Y], and [name] owns it.” If you can write it, you are ready. If you cannot, go back to Decision 1.

The 5-Decision Filter — run every AI initiative through this before committing

  1. Is this a repetition problem (not a judgment problem)?
  2. Do we have the documentation for AI to work from?
  3. Is the founder out of the critical path for this process?
  4. Is this one of the five use-cases with proven SME ROI?
  5. Can we name a metric and an owner right now?

YES to all five → proceed. NO to any of the first three → fix that first. The AI will still be there.

A note on what this framework is not

It is not a technology evaluation. It doesn’t tell you which AI tool to buy. Those are the easy decisions — every vendor will demo their tool for free.

The hard decisions are the five above. They require honest self-assessment about your operations, your documentation, and your own role as a founder in your processes.

Most Indian SME founders I’ve spoken to already know the answers. They just haven’t written them down yet. That is, in my experience, where the work actually begins.

Chandra Shekhar is an IIMB EGMP alumnus (Grade A in Strategy, Information Systems, Corporate Finance, and Business Law), Anthropic AI Fluency certified, and Lean Six Sigma Expert. He helps Indian SMEs identify the 3–5 places AI actually helps their operations and build the documentation that makes it work. Based in Bangalore. Portfolio  |  LinkedIn