The conversation around AI for business has a problem: it operates at two extremes that are both unhelpful.
On one side, you have breathless claims that AI will automate everything and make half of all jobs obsolete within five years. On the other, you have legitimate sceptics pointing out that most AI tools are expensive, unreliable, and solving problems that did not need solving.
Both positions are understandable. Neither is particularly useful if you are running a small business and trying to decide what, if anything, to actually do about AI.
Here is what I have found to be true, working with small businesses and building on top of AI systems directly: some AI tools have genuine, measurable value for small teams right now. Most of the hype is not about those tools.
The useful test
Before evaluating any AI tool, ask one question: what specific, recurring task does this replace or accelerate — and how long does that task currently take each week?
If you cannot answer that question concretely, the tool is not ready for your business. Not because AI is bad, but because you have not identified a real use case.
If you can answer it, you have the basis for an evaluation. A tool that saves a two-hour task and costs $50 per month is worth it if that task happens at least twice a month. A tool that costs $200 per month and saves fifteen minutes of something you do occasionally is not.
This sounds obvious. Most AI purchasing decisions skip this step entirely and are driven by demo videos and fear of missing out.
What is actually working for small teams right now
Writing first drafts. Not writing full content — writing first drafts. AI tools are genuinely good at taking a brief, some bullet points, or even a rough voice note and producing a structured first draft. The draft will need editing. It will sometimes be wrong about facts. But getting from blank page to something workable in minutes instead of hours is a real time saving for teams that produce a lot of written content.
Customer-facing FAQ and support. If your business gets a high volume of repetitive enquiries — the same ten questions asked by different customers in slightly different ways — an AI-powered response system can handle a significant portion of that volume. This works well for support email triage, website chat, and initial screening of enquiries. The key word is repetitive: AI handles known territory well and novel situations poorly.
Data summarisation and extraction. Reading through long documents, research, or reports and extracting the relevant information is something AI does efficiently. If your work involves regularly reviewing contracts, reports, proposals, or large volumes of text to find specific information, AI tools can compress that process substantially.
Meeting notes and action items. Tools that transcribe calls and meetings and summarise them into structured notes with action items are now reliable enough to use in professional contexts. For teams that spend significant time in meetings, the time saving on post-meeting documentation adds up quickly.
What is mostly hype right now
Fully automated content pipelines. The idea that you can set up an AI system that continuously generates blog posts, social media content, and email newsletters with no human review, and that this content will actually perform well, is not true for most businesses. AI-generated content without human editing tends to be generic, confident in ways that undermine trust, and optimised for nothing in particular. Use AI to speed up content creation; do not eliminate the human entirely.
AI “agents” that run your business. There is a lot of demo content showing AI agents completing complex, multi-step business tasks autonomously. Some of this is real and impressive in controlled conditions. In production, with real-world variability, edge cases, and accountability requirements, fully autonomous AI agents are not yet reliable enough for most small business operations. Use AI to assist; keep humans in the loop for consequential decisions.
Generic AI consultants selling transformation. If someone is offering to “transform your business with AI” without first understanding your specific operations, what tasks consume the most time, and what your actual bottlenecks are — they are selling hype. AI transformation starts with a boring audit of what you actually do all day.
The practical starting point
If you want to start using AI in a way that has real impact, start with this:
List the five tasks that take the most time in your business each week. For each one, ask whether it involves processing information, producing written output, or answering questions that have known answers. If yes, there is probably an AI tool worth evaluating for that task specifically.
Start with one. Measure the time saved after a month. If it is saving more than it costs — in money and in the overhead of managing it — keep it and add another.
This is less exciting than the AI transformation narrative. It is also actually how AI creates value for small businesses.
Oshan Shrestha is CEO at Triovate Labs, where we build on the Anthropic API and work with businesses on practical AI integration. If you want to figure out where AI actually fits in your business — start here.