You have a task you think AI could help with. Should you buy a ready-made tool, have something built for you, or bring in someone to work it out with you? The honest answer is that you can often start without spending anything new, and that custom building is usually the last step rather than the first. This guide explains how to decide, and when outside help earns its fee.
Key takeaways
- Start with AI already built into software you pay for. Check it before buying anything new.
- Buy a ready-made product when your problem is a common one that others have already solved.
- Build something custom only when the task is specific to your business and worth the upkeep.
- Bring in help when you can name the problem but lack the time or skills to choose or set up the answer.
What are the routes, in plain terms?
There are four, and you may well use more than one.
| Route | What it means | Fits when | Watch out for |
|---|---|---|---|
| Use what you have | Switch on AI features in software you already pay for, such as your email, accounts or customer system | You want to learn cheaply and the task is routine | Features that are switched off by default or that nobody has tried |
| Buy a product | Subscribe to a ready-made tool built for the task | Lots of businesses have the same problem | Paying for a product that repeats what you already have |
| Build something | Have a tool made or configured for your own process | The task is specific to you and matters to your business | Ongoing upkeep: someone has to look after it |
| Bring in help | Hire a consultant, agency or freelancer to scope, set up or build | You know the problem but not the answer, or have no time | Becoming dependent on them after they leave |
The last route is not really separate. Outside help can steer you to any of the other three. A good adviser may well tell you to use what you already have.
Why start with what I already own?
Because it is the cheapest way to learn what AI can do for you. Some common business tools now include AI features for drafting, summarising, sorting or answering routine questions. Some are switched off by default. Some are switched on and simply unused.
Before you spend on anything new, list the software you pay for and check what AI features each one has. Try the ones that match your task on a small sample of real work. You may find the job is already half done. If not, you will at least know exactly what is missing, which makes the next conversation with a supplier or adviser much sharper.
When does buying a ready-made product make sense?
When your problem is a common one. Reading invoices, sorting enquiries, booking appointments, drafting quotes from past examples: other businesses have the same needs, so products exist for them. A product that many customers use has been tested on far more cases than anything built just for you could be.
Look for products designed for your kind of work rather than general tools. A general AI assistant is fine for drafting and research. When the output feeds something that matters, such as a price, a payment or a promise to a customer, a tool built for that task is usually a safer bet.
Prefer pricing that lets you start small and grow, and check the running costs at the volume you expect, not just the starting price.
When is building something custom worth it?
When the task is genuinely specific to your business, matters to how you win work, and no product does it well. That combination is rarer than it sounds.
Custom work has a cost that does not show up in the quote: someone has to look after it. AI tools need checking, adjusting and sometimes rebuilding as the underlying technology changes. If nobody in your business can do that, you are signing up for an ongoing arrangement with whoever built it.
Building also rarely means starting from nothing. Custom AI work can usually adapt an existing model to your own documents and data. Be cautious of any proposal to build a model from scratch for a small business. Ask why adapting an existing one would not do.
When does outside help pay for itself?
Outside help earns its fee when you can name the problem but cannot easily choose or set up the answer yourself. Typical signs:
- You have several possible uses for AI and cannot tell which is worth doing first.
- You have tried tools on your own and are not sure whether the results are good enough.
- The task touches several systems, and getting them to work together is beyond your team.
- Nobody has the time to run a trial properly, however capable they are.
A sensible way to use help is in two stages. First, ask an adviser to check your priorities, cost them and put them in order. Second, deliver one small first project together, with teaching your team written into the agreement. That way you finish with a working result and with people who understand it.
Whoever you hire, agree before the work starts which changes your team should be able to make without them afterwards. If every small change needs a call to the supplier, the tool will fall behind your business quickly.
Which kind of help: consultant, agency or freelancer?
None is better in general. What matters more is whether they fit the size of the job, will start by understanding your problem, and will leave you able to run the result. A single adviser may suit scoping and a small first project; a larger team may suit work across several systems; and you can mix them, for example one person to scope and another to build. Our guide to consultants, agencies and freelancers covers the differences in detail.
How do I make the decision?
Work through these in order and stop at the first that fits.
- Write down the task, where it costs you time or money, and what would count as better.
- Ask whether it needs AI at all, or whether a simpler fix or plain automation would do.
- Check the AI features in software you already pay for, and try them on real work.
- Look for a product made for this kind of task, and test it on a sample of your own data.
- Consider a custom build only if the task is specific to you and someone can look after it.
- Bring in help at any step where you lack the time or skills, with handover agreed up front.
What if we have already built the wrong thing?
It happens, and pride is the expensive part. If a custom tool is struggling, compare it honestly with the products available now: the market moves fast, and something that did not exist when you started may now do the job. Keep the data and the lessons; they are useful whatever you choose next.
When you are ready to look for help, start with how to choose an AI consultant and how to write an AI project brief.