Every week I speak to a small or mid-size business owner who has just come back from a conference, a webinar, or a conversation with a vendor, convinced that artificial intelligence is the next thing they need to implement. Sometimes they are right. More often, they are not — and the honest thing to do is to tell them so.
This is not a contrarian take. I work in AI consulting. My business depends on companies implementing AI. But I have seen too many projects fail not because the technology was wrong, but because the foundation was not there.
The real question nobody asks
Before asking "how do we implement AI?", the right question is: "what problem are we actually trying to solve?"
The more precisely you can define the outcome, the easier everything else becomes. "We want to use AI to improve customer service" is not a specific outcome. "We want to reduce the average time to first response on customer enquiries from 4 hours to under 30 minutes" is specific. One of these can be evaluated. The other cannot.
In most cases, when we dig into the answer, we find one of three situations:
The process is broken. There is a manual, repetitive, error-prone process that wastes hours every week. The instinct is to put AI on top of it. The right move is to fix the process first — or automate it with something far simpler than AI. A well-configured workflow tool or a simple script can eliminate 80% of the problem in a fraction of the time and cost.
The data does not exist or is not clean. AI learns from data. If your customer records are scattered across three systems, your sales data lives in someone's inbox, and nobody agrees on how to define a "conversion" — AI will not save you. It will give you fast, confident, wrong answers.
The team is not ready. A tool that nobody uses is not a solution. I have seen companies spend six figures on AI platforms that were abandoned within three months because nobody trained the team properly and the tool did not fit how people actually worked.
What most businesses actually need first
If you run a small or medium business and you are thinking about AI, here is a more honest starting point:
- Document your processes. Write down how things actually get done — not how they are supposed to get done. You will find the real bottlenecks.
- Identify the repetitive work. What tasks take the most time and require the least judgment? Those are your first automation targets.
- Start with simple automation. Rule-based automation solves the majority of business inefficiencies at a fraction of what AI costs.
- Build your data foundation. Start collecting and organising the data you will need before you need it.
Only once these foundations are in place does AI investment start to make sense.
When AI is actually the right answer
There are situations where AI genuinely creates value that simpler tools cannot:
- When you need to process unstructured data at scale — documents, emails, images, audio
- When patterns in your data are too complex for rule-based logic
- When you need to personalise at a level that humans cannot maintain manually
- When prediction is the core of what you are trying to do
The difference between these cases and the others is not always obvious from the outside. That is exactly what a good AI assessment should tell you — before you spend anything.
The uncomfortable truth about AI vendors
Most AI vendors are not incentivised to tell you that you do not need their product yet. They will show you the most impressive demos, cite the biggest ROI numbers, and create urgency around not being left behind. Some of what they say is true. Much of it is optimised for closing a sale.
The question to ask any AI vendor — or any consultant, including me — is: "what would you recommend if AI was not an option?" If they cannot answer that question coherently, they are not giving you advice. They are giving you a pitch.