The owner of a fourteen-person professional services firm sat down to a coffee with three different advisers in the same fortnight. One told her to build a custom AI system because her business was unique. One told her to buy off-the-shelf software because that was the safe, sensible choice. One told her to hire a consultant for three months because she did not have time to figure this out herself. Each adviser was certain. None had asked her what her actual situation looked like before recommending. She came away more confused than when she started.
That moment is the reason this post exists. There are three real paths to getting AI working in a small business, and many owners pick the one that suits their temperament rather than the one that fits their situation. The good news is that the matching criteria are clearer than they look. The four variables that actually decide it are easy to score honestly in under an hour, and the result often surprises the owner about which path is right.
What are the three paths to getting AI working in a small business?
The three paths are build, buy and hire. Build means custom development on top of vendor APIs, typically wrapping OpenAI, Anthropic or Google models in business logic specific to your firm. Buy means activating off-the-shelf software like Microsoft 365 Copilot or HubSpot Breeze on a monthly subscription. Hire means bringing in a consultant, agency or fractional CTO to do the work or guide the decision.
Each path carries different ownership. Building leaves you holding the system, the maintenance and the model behaviour. Buying hands all three to the vendor at the cost of configurability and vendor lock-in. Hiring outsources execution but typically leaves your firm with whatever the engagement produced, plus a knowledge gap when the consultant walks out the door. MIT Sloan adds a useful fourth option called boosting, which is buying a vendor’s model and enriching it with your proprietary data through retrieval-augmented generation or fine-tuning. For many owner-led firms, boosting is a variation on buying with a bit of hiring on the side, and the same matching rules apply.
Why does the path you pick matter so much for your business?
It matters because Forrester puts the failure rate of wrong build-versus-buy software decisions at sixty-seven per cent, and a firm with twenty staff has no safety net for that. The same money committed to the wrong path produces shelfware, technical debt, or an expensive consulting engagement that delivered a deck rather than a working system. The path is also hard to reverse once the contract is signed.
The deeper reason is that each path has a different shape of risk that bites at a different time. Buying risks vendor lock-in and shelfware, which surface over the next twelve to twenty-four months as the subscription quietly bleeds cash for a tool no one is using. Building risks maintenance debt, which compounds over years as the system needs patching, updating and re-skilling each time a developer leaves. Hiring risks knowledge silos, where the consultant’s work is correct but no one in the firm understands it well enough to evolve it. Picking the wrong path does not feel wrong on day one. It feels wrong on month nine or month twenty-six.
What four variables decide which path actually fits?
Four variables decide it. Technical capacity, do you have a founder or staff member with coding ability and meaningful time to dedicate. Urgency, does this need to work in six weeks or six months. Capital shape, do you prefer a predictable monthly opex line or a larger one-off capex investment. Use-case unusualness, is the work commoditised across your sector or genuinely differentiated. Score honestly on each.
A firm with no technical capacity, a six-week urgency, a tight cash position and a commodity use case is a clear buy. A firm with a technical founder, a twelve-month timeline, available capital and a use case that reflects unusual proprietary data is a candidate to build. A firm with no technical capacity, a moderate urgency and a use case that needs vendor selection and configuration is a clear hire. The hard cases are the ones where two variables pull in different directions. Forty-six per cent of small business owners in the FSB’s 2024 research said they lack the skills to use AI confidently, which is the single most common variable in practice and pushes the default towards buy plus a short hire for setup.
When do owners default to the wrong path, and what causes it?
Owners default to the wrong path in three predictable ways, each tied to who they are rather than what the situation calls for. Technical owners over-build because they enjoy the work and the maintenance burden feels manageable from the inside. Non-technical owners over-buy because buying feels safer and pushes the decision-making to a vendor’s reputation. Time-poor owners over-hire because hiring feels faster.
Joel Spolsky’s defence of not-invented-here syndrome captured the build trap precisely. A founder who believes their invoices, their clients or their workflow is unique builds a bespoke system that delivers eighty per cent of what a vendor would have delivered for twenty per cent of the cost, and inherits years of maintenance for the privilege. The buy trap is shelfware, which Flexera’s 2024 IT asset research puts at twenty to thirty per cent of organisational software spend. The hire trap is what PathOpt calls hot-potato ownership, where the consultant delivers something that technically works but no one inside the firm owns the outcome, and the work quietly drifts. Knowing your own default is half the discipline. Picking against it when the variables call for it is the other half.
What does the hybrid pattern look like once a firm gets it right?
The pattern many small businesses settle into after twelve months is buy the obvious, hire for the specialist, build only the unique edge. Buy covers the commodity functions where a vendor has already solved the problem and a subscription gives you a working tool inside two weeks, productivity copilots, customer support chatbots, document drafting, scheduling, basic forecasting. The McKinsey, MIT Sloan and CIO.com frameworks all converge on this as the default starting point.
Hire covers the bit that buying alone fails on, selecting the right vendor, configuring the tool to your workflows, and training the team. A four-to-eight-week implementation engagement at five to fifteen thousand pounds is the standard SME shape, and it usually returns more than it costs in shelfware avoided. A fractional CTO at three to ten thousand pounds a month is the upgrade for firms that want ongoing strategic guidance without a full-time hire. Build is reserved for the narrow edge where your data, process or customer insight is genuinely unusual and the work creates lasting competitive advantage. For the typical owner-led firm that is one function out of ten, not five out of ten. The portfolio review is quarterly, not one-off. As the firm grows, as your team builds capability and as vendors release new products, the right allocation shifts, and the matching question gets asked again.
If you are weighing three advisers giving you three different answers, or you want a peer to sense-check the path before you sign anything, book a conversation.
