AI for small business: what actually helps vs what is hype

The category of "AI for small business" has expanded faster than the genuine use cases have. Here is a clear-eyed breakdown of what AI actually helps with for a small or medium business, and what you can safely ignore.

What AI genuinely helps with

Drafting. Writing first drafts of things, emails, job descriptions, product descriptions, proposals, is genuinely faster with AI. The output needs editing. The blank page problem is solved. For businesses that produce a lot of written content, this is real time saving.

Summarising. Turning a long document into a short one, extracting the key points from a meeting recording, summarising a supplier contract. AI is good at this. It is not perfect. You need to check the output. But it is meaningfully faster than doing it yourself.

Data patterns. Finding patterns in data that a human would miss or take hours to find. Sales trends, customer cohorts, anomalies in a P&L. AI is good at this if the data is clean and the question is well-formed.

Scenario modelling. Running a what-if. If I raise prices by 8%, what happens to volume? If I extend opening hours on Friday, what does that do to margin? This is where AI for business has genuine leverage that most software does not offer.

What AI does not help with

Making the decision for you. AI can give you a recommendation. It cannot carry the accountability for the outcome. For consequential decisions, the human needs to understand both sides and make the call. AI that presents itself as the decision-maker is AI that is hiding the accountability problem.

Replacing relationships. AI-driven customer service works for transactional queries. It does not work for the client relationships where human judgment and trust are the product. Automating the wrong customer interactions costs more than the saving is worth.

Compensating for bad data. If your data is wrong, inconsistent, or incomplete, AI will produce confident-sounding analysis of the wrong data. The output looks authoritative. The input is garbage. Cleaning your data is not glamorous. It is the precondition for AI being useful.

The AI products worth paying for

For a small business, the AI products that are worth the subscription fee share a few characteristics. They are specific to a task, not general-purpose. They integrate with data you already have. They produce outputs that are auditable, so you can check the reasoning. And they save you time on something you actually have to do regularly.

The AI products not worth paying for are the ones that produce plausible-sounding outputs about general business topics that are not anchored to your specific numbers and context. Generic business advice is freely available. It is not what you need.

The question to ask before any AI purchase

One question cuts through most of the noise. What specific decision or task does this software help me do better, and how do I know if it worked?

If the vendor cannot answer that question with a specific before and after, with metrics, you are buying hype. If they can answer it concretely, you are buying a tool.

Most small businesses have four or five decisions that recur monthly, are consequential to the P&L, and are currently made on experience and instinct alone. Pricing. Stock levels. Staffing. Promotional investment. Supplier relationships. AI that is specifically designed to help with those decisions, with your data, with a record of whether it was right, is worth paying for. General AI assistants for business are a tool for generating text. Do not confuse the two.