The Question Nobody's Asking Properly
Every business owner I talk to right now is getting the same pressure from all directions: "What's your AI strategy?" Investors want to hear it. Clients are starting to ask. Your competitors are dropping it into every LinkedIn post.
But when you actually sit down and try to answer the question, you hit a wall. You don't have a data science team. You don't have a six-figure innovation budget. And you definitely don't have eighteen months to wait for results.
So the question isn't really "should we use AI?" — it's "how do we use AI without betting the business on it?"
Option 1: Build It Yourself
The appeal is obvious. Build something proprietary, train it on your data, own it completely. Nobody else has it. Competitive moat. Sounds brilliant in a pitch deck.
Here's what it actually looks like for an SME:
- Hiring ML engineers at £70-100k each, with a 3-5 month recruitment timeline
- Cloud compute running £1-5k per month just at pilot scale
- 6-18 months before you see anything in production
- Ongoing maintenance eating 15-25% of the initial build cost every year
I've seen this play out. A business spends nine months and a small fortune building something that works in a demo but breaks in production. The developer who built it leaves. Nobody else understands it. The project quietly dies.
For most SMEs, building from scratch is solving the wrong problem at the wrong scale.
Option 2: Buy Off-the-Shelf
This is where most businesses end up. Sign up for a SaaS product, configure it, go live in a few weeks. Predictable pricing. Vendor handles updates.
The problem comes three months later:
- The tool doesn't quite fit how your team actually works
- You're bending your processes to match the software, not the other way around
- Each new tool creates another data silo
- Per-seat licences start adding up — and you're paying for features you'll never touch
Buying off-the-shelf software is like buying someone else's suit and hoping it fits. It might look fine from a distance, but you'll always feel the places where it doesn't.
Option 3: Augment What You've Already Got
This is the option most people don't consider, and it's usually the right one for growing businesses.
Instead of building from scratch or replacing everything with SaaS, you keep your core systems and layer AI onto the specific points where it makes the biggest difference.
Here's what that looks like in practice:
- Map your actual processes — find the workflows that are leaking time and money
- Build targeted automations — AI that connects to your existing tools via APIs, no rip-and-replace
- Start with human oversight — your team approves and overrides until confidence is proven
- Expand once you've seen the ROI — not before
The timeline is weeks, not months. The cost is thousands, not hundreds of thousands. And you're not locked into a vendor's roadmap — you own the system.
The Five Questions That Actually Matter
Before you commit to any path, ask your team these:
- Is this mission-critical or nice-to-have? Start with the processes that are genuinely costing you money.
- How sensitive is the data? Compliance requirements might rule out certain SaaS tools.
- Do you have technical talent in-house? Be honest. If the answer is no, building isn't realistic.
- What's your budget horizon? Big capital outlay or phased spending?
- How quickly do you need results? If the answer is "yesterday," building isn't an option.
Most SMEs I work with score highest on augment. Start there, prove the value, then decide whether deeper investment makes sense.
The Real Answer
The AI implementation debate is usually framed as a technology decision. It isn't. It's a business decision.
The right question isn't "which AI should we use?" It's "which process is costing us the most, and what's the fastest, lowest-risk way to fix it?"
Start there. The technology follows.