I get asked about AI more than anything else at the moment, and almost never in a useful way. The question is usually some version of "should we be doing something with AI", which is a bit like asking whether you should be doing something with electricity. Doing what, exactly, and to which part of the business?
I build with AI tooling every working day. I also tell people to stop, fairly regularly. So here is what I actually think a small business in Hull or East Yorkshire should do about it, having had this conversation a few dozen times now with people running real companies with real margins.
Start with a job, not a tool
Every business I have seen get value out of this started from a specific, boring, repeated task. Not from a subscription.
The tasks that keep coming up round here look like this. Turning a set of scrappy site notes into a written quote. Drafting the first reply to enquiries that all ask roughly the same four questions. Writing product descriptions for a catalogue that has sat half finished for two years. Pulling the useful bits out of a long supplier document. Chasing overdue invoices in a way that sounds like a person.
None of those are exciting. All of them are hours a week that someone senior is currently spending. That is the whole opportunity, and it is a lot less glamorous than the conference talks suggest.
Where it genuinely earns its keep
Three areas, in my experience.
- Writing something that already exists in your head. You know what the quote should say. Getting it onto the page is the slow bit. A first draft you then correct is much faster than a blank document.
- Reading piles of text. Summarising, comparing, pulling out the clause that matters. Machines are patient in a way people stop being at four in the afternoon.
- Repetitive admin with a clear rule. Categorising, tidying, reformatting, moving information between two systems that were never going to talk to each other.
Where it does not
Anything where being wrong is expensive and nobody is checking. Anything that amounts to regulated advice, whether that is legal, medical or financial. And anything that is really the relationship rather than the task.
That last one matters more here than it would in London. In a city this size your reputation travels through a fairly small number of people. Automating the part of your business that is actually you talking to a customer you have known for eleven years is a bad trade, however many minutes it saves.
The local part that actually matters
There is no Hull version of the technology. What is different locally is the shape of the businesses.
Most of the firms I work with round here have small teams, no IT department, and an owner who is still doing three jobs. The binding constraint is not which model is cleverest this month. It is whether anyone has ninety spare minutes to set the thing up and write down how it works. Any advice that ignores that is written for companies with a transformation budget, and you do not have one.
The flip side is the opportunity. Very few of your competitors in East Yorkshire are doing any of this properly yet. You do not need to be ahead of the industry. You need to be ahead of the four other firms bidding for the same work, and right now that bar is low.
What a sensible first three months looks like
- Month one. Pick one task. Do it the normal way but with an assistant open next to you, and note how long it took before and after. One task, one person, no announcements.
- Month two. Write down the process, including the exact instructions that worked. If it only lives in one person's head it is a hobby, not a business improvement.
- Month three. Hand it to someone else and see if it survives. Then, and only then, look at whether any of it is worth automating properly.
If after three months you cannot point at the hours saved, stop. That is a perfectly respectable outcome and it has cost you almost nothing.
The mistakes I keep seeing
- Buying the platform first. Someone sells a licence, it gets rolled out, and six months later nobody can name a task it does. Find the job, then find the tool.
- Letting it write the website. I have written about tanking my own rankings doing exactly this. Volume without judgement is worse than nothing.
- No human check. Every workflow needs a named person who reads the output before it reaches a customer. If you cannot name them, the workflow is not finished.
- Pasting client data into whatever is free. Covered below, and it is the one that can genuinely land you in trouble.
Data, briefly
Consumer tiers and business tiers are not the same product. Business and team plans generally commit to not training on your inputs, and free consumer tiers often do not. Read the terms of the specific plan you are on rather than the marketing page, because this varies by provider and changes.
The practical rule I give people: nothing goes into a chatbot that you would not be relaxed about a competitor reading. Customer records, staff details, anything commercially sensitive, keep it out until you have checked the plan. There is a fuller version of this in my note on what AI governance means for a small business, and it is shorter and less frightening than the phrase suggests.
What it costs
Less than people expect for the tools, more than people expect for the setup. The subscription is rarely the real number. The real number is the time to work out what you are actually automating and to write it down properly. I broke the figures down in how much AI costs for a small business, and there is a companion piece on which tools are worth paying for.
Where I would start if I were you
Write down the three tasks in your week that you resent most. Not the hardest ones, the most repetitive ones. Then ask whether each is mostly writing, mostly reading, or mostly moving information about. If it is one of those three, there is probably something here for you. If it is a judgement call or a relationship, there probably is not, and anyone telling you otherwise is selling something.
If you want a second opinion on your particular list, that is a conversation I am happy to have. Have a look at how I actually build with this stuff, then get in touch and bring the three tasks with you.
AI for a local business, answered
Common questions from Hull and East Yorkshire business owners about where to start with AI.
Start with a specific, boring, repeated job, not a subscription. The ones that keep coming up locally are turning site notes into quotes, drafting first replies to similar enquiries, writing product descriptions, and summarising long supplier documents. Those are hours a week someone senior is spending now. That is the opportunity, far more than any conference talk.
It earns its keep writing something that already exists in your head, reading and summarising piles of text, and handling repetitive drafting. It is weak where being wrong is expensive, where judgement matters, or where it would be touching money or legal detail unchecked. I tell people to stop using it there fairly regularly.
Not for the everyday jobs. A good subscription and someone shown how to use it covers most of what a small business needs. You start needing help when you want AI wired into your own systems, your website or your data, which is where I come in. For drafting quotes and emails, you do not need me at all.
Yes, if you aim it at the repeated jobs eating senior time rather than at something impressive but rare. The return comes from giving back hours every week, not from a flashy one-off. On tight margins that time saving is the point. The waste comes from paying for tools nobody has been shown how to use.
Pick one weekly job, try it with one person for a month, and judge it on whether it actually saved time. Do not roll it out to everyone on day one, and do not buy five tools. One job, one person, one month. If it helps, widen it. If you want a steer on which job to pick, I am happy to talk it through.