Last year, a PR director called me. She was frustrated. Her team of three was spending four hours each, every single day, manually scanning news articles to find opportunities for their clients.
She asked the question I hear from SMB leaders every week: "Should we build a chatbot?"
That was the wrong question. The right question isn't about the technology; it's about the process. When you reframe "AI project" as "Automation project," the path to ROI becomes clear.
The Intern Test
Before investing a pound in AI, apply this simple filter to any process you're thinking of automating:
"Could you write down the guidelines for this process clearly enough that you could hand them to an intern with no prior experience, and they could do the job?"
Strong Automation Candidate
Repeatable, rule-based, high ROI potential.
Needs Human Judgment
Requires instinct or years of experience. AI will struggle.
1. LLMs: The Reading Machines
LLMs like ChatGPT or Claude are incredibly powerful at augmenting humans. They are "reading and matching" machines.
Case Study: Pulse AI
We built a system that reads news articles and compares them against brand profiles.
2. Computer Vision: Visual Intelligence
If LLMs read text, Computer Vision reads pixels. It's perfect for any task where someone is looking at a screen or a physical object repeatedly.
Case Study: Angel Protection
Real-time threat detection across 50+ CCTV feeds simultaneously.
- 300ms detection latency
- Edge-first: No data leaves the building (GDPR compliant)
- 24/7 operation without fatigue
3. Algorithms: The Hidden ROI
This is the project type that surprises people most. The highest-impact part often barely uses "AI" in the modern sense. It's about smart engineering.
Case Study: DG Financial
Unified search across 80,000 documents in S3, OneDrive, and SharePoint.
"Fuzzy search and smart indexing gave them hours back every week. The LLM was just the cherry on top for summaries."
The Comparison
ROI does not follow the complexity curve. Often, the simplest technology provides the most foundational value.
| Technology | Setup Cost | Flexibility | ROI Priority |
|---|---|---|---|
Algorithms | Low | Rigid | Highest (Foundational) |
LLMs | Medium | High | High (Augmentation) |
Computer Vision | High | Moderate | High (Specialized) |
ROI Calculation Framework
Before commissioning any work, run this calculation. It takes 15 minutes and tells you whether this is worth exploring.
The "3-Month Payback" Math
3 people × 4 hours/day × £30/hour × 250 working days
Typical SME automation MVP (6–8 weeks delivery)
Assuming 80% efficiency gain (conservative for these projects)
After less than 3 months, every pound saved is pure profit for the business.
Gradient Insight
Ready to find your first AI win?
We help SMEs identify and build high-ROI AI solutions in 6–8 weeks. No fluff, just results.