Buy the off-the-shelf tool first. Build only once you know exactly which part of your workflow the generic tool cannot handle, and only if that gap is big enough to justify the ongoing cost of maintaining custom software. Most SMEs get this backwards — they commission a bespoke build before they have used a commercial product long enough to know what “custom” would actually need to solve.
This decision is landing on more desks because the tooling has changed. Anthropic recently launched Claude for Small Business, joining a growing field of packaged AI products aimed squarely at businesses too small to have an engineering team. That is a genuinely different offer to what was available even eighteen months ago, when “adopting AI” often meant either a generic chatbot subscription or a costly custom project with no guarantee of fit. The choice founders now face is sharper — and getting it wrong in either direction is expensive.
Why this question is coming up now
A few things have converged. JPMorganChase’s research into small business AI usage shows adoption is broad but shallow — most small firms are using AI for something, but few have moved past experimentation into workflows that actually change output. Business Insider has documented the “snafus” that come with rushed adoption: tools bought without a clear use case, data fed into systems without a policy, staff using consumer AI for tasks with real business risk attached.
At the same time, marketing operations teams are reporting that generic AI tools underperform without proper data foundations — a theme MarketingProfs has covered in detail. Put those two signals together and the pattern is clear: off-the-shelf AI works well for well-defined, high-volume tasks, and works poorly the moment a business tries to force it into a process the tool was never designed for.
That is exactly the fork in the road: buy for the tasks that are common across businesses, build (or customise) for the tasks that are specific to yours.
What buying off-the-shelf actually gets you
Commercial AI products — whether general-purpose assistants, CRM copilots, or accounting-software AI features — are built to solve a common problem for thousands of businesses at once. That has real advantages for an SME:
- Speed. You can be using a tool inside a week, not after a three-to-six month build cycle.
- Lower upfront cost. Subscription pricing, often scaled to seat count, versus a development budget that has to be paid whether or not the project succeeds.
- Maintained and updated by someone else. Model upgrades, security patches, and new features arrive without you commissioning them.
- Lower risk if it doesn’t work. Cancel a subscription. Unwinding a failed custom build is a different order of cost and disruption.
The trade-off is fit. A generic tool solves the 80% of your workflow that looks like everyone else’s — drafting, summarising, basic customer support responses, meeting notes. It will not natively understand your specific pricing logic, your regulatory obligations, or the three legacy systems your operations run on.
What building actually costs you — beyond the invoice
The build option looks more attractive the moment a founder imagines a tool that does exactly what they want. But the real cost of custom AI isn’t the initial development fee — it’s everything that comes after:
- Ongoing maintenance. Models change, APIs deprecate, and someone has to own the upkeep indefinitely, not just at launch.
- Data readiness. Custom AI is only as good as the data it’s trained or grounded on. Most SMEs discover mid-project that their data is inconsistent, siloed, or simply doesn’t exist in usable form — the exact issue MarketingProfs flagged as the reason so many AI marketing projects underdeliver.
- Opportunity cost of time. Every month spent building is a month a commercial alternative could have already been generating value, however imperfectly.
- Talent dependency. Someone in the business — or an external partner on permanent retainer — needs to understand the system well enough to fix it when it breaks.
None of this means custom builds are wrong. It means they should be reserved for problems that are genuinely core to your competitive position, not for generic tasks a subscription already handles.
The decision framework
Five questions help clarify the decision before committing to either path:
| Question | Leans “Buy” | Leans “Build” |
|---|---|---|
| Is this task common across most businesses in your sector? | Yes | No |
| Do you need it working in under a month? | Yes | No |
| Is your process data clean and centralised already? | Doesn’t matter much | Must be yes |
| Does this workflow touch your actual competitive advantage? | No | Yes |
| Can you commit to ongoing technical ownership? | Not required | Required |
Score honestly. Most SMEs answer “buy” on four or five of these the first time they ask the question — which is exactly why we tell founders to start there, and revisit custom development only once the off-the-shelf tool has been in use long enough to show precisely where it falls short.
This is the same logic behind why we recommend a proper diagnostic before any AI spend — something we cover in detail in Do You Need an AI Gap Analysis Before You Spend on AI Tools?. The gap analysis tells you which tasks are common versus specific — the exact input this framework needs to work.
The hybrid path most SMEs actually land on
The businesses that get the most value rarely choose purely one or the other. They buy a commercial platform for the 80% of tasks that are common — drafting, scheduling, first-line customer queries, reporting — and then invest selectively in a thin layer of customisation on top: a workflow automation connecting the AI tool to their CRM, a prompt library trained on their specific tone and pricing rules, an internal wrapper that enforces their data-handling policy before any input reaches the model.
This hybrid approach keeps the maintenance burden low (you are not maintaining a model, just a thin integration layer) while closing the fit gap that generic tools leave open. It is also far easier to unwind if it doesn’t work — you are not writing off a six-figure development project, just retiring an integration script.
A worked example
Consider a Malaysian professional services firm with fifteen staff and a monthly AI budget of roughly RM 15,000. Buying a commercial AI suite for drafting, research, and internal knowledge search might cost RM 3,000–5,000 a month in licences and take two weeks to roll out. That leaves RM 10,000–12,000 a month for the genuinely firm-specific problem — say, an automated client-intake workflow that pulls from their proprietary matter-management system, something no off-the-shelf tool handles natively.
Spend the full RM 15,000 trying to build a custom system to do both, and the firm typically ends up with a slower rollout, a higher failure risk on the general tasks a subscription would have solved in a fortnight, and a maintenance obligation nobody budgeted for past month six.
How OMO helps SMEs make this call
Choosing between buy and build isn’t a technology decision — it’s a resource-allocation decision, and it should be made with the same rigour you’d apply to any capital expenditure. If your team is still deciding where to start, our earlier piece on Which AI Tools Should Your SME Actually Adopt in 2026? walks through the current commercial options in more detail.
Frequently asked questions
How do I know if my data is “clean enough” for a custom AI build?
If your customer, pricing, or operational data lives in more than two disconnected systems, requires manual reconciliation, or has known gaps and duplicates, it isn’t ready. Fixing data hygiene first is almost always cheaper than building AI on top of a messy foundation and discovering the problem later.
Can we start with an off-the-shelf tool and move to custom later?
Yes, and this is the path we recommend most often. Running a commercial tool for three to six months gives you a real record of where it underperforms, which turns “we think we need custom AI” into a specific, defensible business case.
What’s the biggest mistake SMEs make with build vs buy?
Committing to a custom build based on what a tool should do in theory, before testing whether an existing commercial product already does it well enough in practice. The second-biggest mistake is the opposite: buying a generic tool for a task that is genuinely core to the business’s advantage, then wondering why competitors using the same tool get the same generic results.
Does adopting AI actually move revenue for small businesses?
Survey data — including recent findings covered by ColoradoBiz — suggests small businesses using AI are reporting revenue gains, but the effect concentrates in firms that moved past isolated experiments into workflows embedded in daily operations. Isolated tool trials rarely show up in the numbers; embedded use does.
Working out whether to buy, build, or blend is easier with a second set of eyes on your specific workflows and budget. Book a free strategy call with OMO to map the right path for your business.