Most AI strategy content is written for a company that already has a data team, a cloud budget, and someone with "transformation" in their job title. An MSME owner reads one of these decks, sees a slide about roadmaps and centres of excellence, and quietly decides AI is for later, once the business is bigger. That is the wrong conclusion, and it comes from reading the wrong kind of strategy document. A 30-person manufacturer or a 60-person trading firm does not need a transformation roadmap. It needs four questions answered honestly.
Start with a workflow, not a wishlist
The enterprise version of this exercise usually starts with "where across the business could AI help", and the answer is always everywhere, which is useless. The MSME version should start narrower: which single task, done repeatedly, is quietly eating the most tired hours of your best people. Not the most interesting task. The most repeated one. Invoice reconciliation, follow-up emails, first-draft quotations, compliance paperwork, inventory re-ordering decisions made the same way every week. That task is where you start, because it is the one place a small win compounds every single week instead of once.
The four questions that replace the strategy deck
Once you have that task, run it through four questions before you spend a rupee on a tool or a consultant:
- How often does it happen? Daily or weekly tasks are worth automating. Something you do twice a year is not, however painful it is.
- How structured is the input? A task that starts from the same kind of document or form every time is far easier to hand to AI than one that starts from a phone call or a walk-through.
- What happens if it is wrong? A wrong first-draft email costs you a rewrite. A wrong number on a client invoice costs you a client. Match the amount of human checking to the size of that gap.
- Who owns checking it? If the honest answer is "no one has time", you have not found a use case, you have found a future problem.
Score every candidate task against these four, and the deck writes itself. You will usually end up with one or two places to start, not the twelve a consultant's slide would give you.
Where MSMEs are actually winning with this right now
Across the businesses I have looked at closely, the wins cluster in three places, and none of them require a data scientist:
- Customer-facing drafting: first-pass quotations, follow-up emails, WhatsApp replies to routine questions, all reviewed before they go out.
- Reading patterns in what you already track: demand and stock patterns from the billing software you already use, not a new forecasting platform.
- Compressing paperwork: summarising long compliance documents, contracts, or vendor terms into the three lines that actually matter for a decision.
None of these need a model you train. They need a model you prompt well, wired into a workflow someone actually owns.
The part nobody budgets for
The tool is rarely the expensive part any more. The part that determines whether this sticks six months later is whether one person in the business is actually responsible for it. Not a vendor, not "the team" in general, one named person who checks the output, notices when it drifts, and owns the decision to expand or drop it. Every MSME AI effort I have seen die, died from that gap, not from the model being too weak.
You do not need a transformation roadmap to start using AI well. You need one task, four honest questions, and one person who owns the answer. Everything after that is detail.