Most organisations don’t arrive at AI with a clear plan. It’s because someone on the board asked about it, or a competitor announced they were doing something with it, or a team member came back from a conference fired up about the possibilities. The pressure to act comes first and the thinking about how to act tends to follow.
That’s not a criticism, it’s just where most organisations are right now. And if that’s where you are, the good news is that getting started doesn’t require a big strategy or a significant investment. It requires working out a few fundamentals before you do anything else.
Sort out the platform question first
Before you pick a use case, you need to understand what you’re actually signing up for when you use an AI tool.
A lot of the free and low-cost options available right now operate on a simple trade. You get access to the tool, and in return, you give permission for the model to learn from what you put into it. That’s fine if you’re asking it to help you write a birthday message. It’s a problem if you’re putting in client data, internal documents or anything your organisation would consider confidential.
In most workplace settings, using a personal AI account for work purposes means you’re almost certainly outside the bounds of your employer’s data policies, even if nobody has told you that explicitly yet. It’s worth checking before you’re in a situation where it matters.
The fix is straightforward. Look for a business account with an enterprise licence that explicitly commits to not training on your data. Check what the vendor is and isn’t committing to. That’s the starting point. Everything else builds from there.
Start with work you can actually judge
Once you’ve got the right account, the instinct is often to find the biggest, most impressive use case and go straight there. Resist that.
The best place to begin is with one-off, low-stakes tasks where you can read the output carefully and make a real call on whether it’s good enough to use. Drafting an email to a difficult client. Reshaping a paragraph that isn’t quite landing. Testing a few different ways to frame a proposal.
These feel small, but they’re doing something important. They’re helping you develop a feel for what the model does well, where it tends to go wrong, and whether the output actually sounds like you, or like someone trying to sound professional without saying anything in particular.
That last test matters more than people realise. AI systems are designed to produce responses that feel plausible and helpful. They’re not designed to be correct. If the model doesn’t know the answer, it will make one up, and it will do so with the same confident tone it uses when it does know. Getting a sense of that pattern early, before you’re relying on it for anything critical, is worth the time.
It’s also worth keeping your ego in check. If the tool is telling you your idea is brilliant and your strategy is watertight, that’s the system doing what it was built to do, not an honest appraisal of your work. It’s a sign that the system is designed to make you feel that way. A more useful prompt is to ask what’s wrong with your thinking, not whether your thinking is right.
Give it context, not just questions
One of the most common early frustrations with AI tools is that the outputs feel generic. That’s usually a prompting problem rather than a capability problem.
Most people start by using these tools the way they’d use a search engine. They type a short question and expect a useful answer. What actually works is closer to briefing a smart but very junior team member who knows nothing about your organisation, your clients or your specific situation. The more context you give, the more useful the output.
That might mean telling the tool who you are, who your audience is, what you’re trying to achieve and what you’ve already tried. It might mean giving it your current KPIs alongside a brainstorming prompt, or sharing a list of existing campaigns before asking it to suggest new ideas. The technical term for this is context engineering, but the practical reality is simpler. The more relevant information you put in, the better the output you get back.
AI is particularly useful for brainstorming, critique and idea development. Ask it to poke holes in a plan. Ask what you might have missed. Ask it to play devil’s advocate on a decision you’re already leaning toward. Used that way, it can surface things you hadn’t thought of. Not because it’s smarter than you, but because it approaches the problem without your existing assumptions baked in.
Keep people in the review seat
The thing worth holding onto through all of this is that these systems don’t actually think. They predict. They generate responses that are statistically likely to be useful, based on patterns in the data they were trained on. That’s a powerful capability. But it’s not the same as understanding.
Which means the judgement call still sits with you. Not as a formality, but as the actual point of the exercise. The organisations getting real value from AI right now aren’t the ones who’ve handed over the most responsibility to the tools. They’re the ones who’ve worked out how to use AI to do more of the groundwork, while keeping experienced people firmly in the seat where decisions get made.
Start small. Give it proper context. Review everything. That’s not a cautious approach to AI. It’s just a smart one.
If you’re ready to move from AI curiosity to something that actually works in your business, we can help. Let’s talk.
