Here are five things from our perspective that are worth getting on with sooner rather than later, none of them especially exciting, but all will pay dividends down the road.
The unglamorous one: Your data
Whatever AI tool you end up adopting, whether that’s a chatbot, a copilot inside your accounting software, or something more bespoke, it can only ever work with what it can see. If your customer records sit across three different systems, your stock count lives partly on a clipboard, and the only person who really knows what your product margins look like is you, no tool is going to rescue you from that. Sort the data first. Make sure it lives somewhere a machine can read. Everything downstream of that gets easier, including a lot of things that have nothing to do with AI.
Pick one problem
Skip the “we need an AI strategy” conversation. It’s the corporate version of “we should do something”. Better to pick one pinch point in the business that’s actually costing time or money and see whether a tool can help. Quoting. Reconciling invoices. Answering the same five customer questions on repeat. Drafting the first version of a proposal. Pick one and have a crack at it for a fortnight. You’ll likely learn more in that time than you would from six months of strategy work.
Give the team time
A lot of business owners assume the next step is buying software. More often it’s giving the team a bit of breathing room. Most of the platforms you already pay for, Microsoft 365, Xero, your CRM, have AI features baked in that nobody has tried. Block out a couple of hours, let people muck around (ideally with dummy data), then ask them what they found useful and what was a waste of time. The people using the tools daily are the ones who’ll spot the genuine opportunities.
Set the rules before you need them
Before someone on your team pastes a client contract into a free chatbot to summarise it, have the conversation. What’s okay to put into an external tool. What isn’t. Who signs off on a new one before it gets used on real work. A single page of guidance is plenty for most small to medium businesses and it’s a much easier document to write before something has gone sideways than afterwards.
Measure, or don’t bother
The most common mistake we see is rolling something out and never going back to check whether it worked. Write down what you expect the tool to do, how much time or money you think it’ll save, and a date to look at it again. A surprising number of AI pilots feel productive in the moment but cost more in subscriptions, setup and training than they ever save in output. You only know which side of that line you’re on if you’ve measured.
None of this is exciting, but it’s what working with AI in a SME business actually looks like before the interesting stuff can happen. The businesses we’ve seen getting real value from any of it got the basics straight first.
If you’re ready to figure out where AI fits in your business, get in touch with the team at Icehouse.
