Research
What 100 SMEs Told Us They Actually Want from AI — And Why Most of Them Still Don't Have It
We spent four weeks interviewing 100 small and medium-sized business owners across Australia, New Zealand, and Southeast Asia. We asked them one question to start: "What would AI need to do for your business to justify the hype?"
Their answers didn't sound like the AI discourse you read on LinkedIn. Nobody asked for "a copilot." Nobody mentioned "prompt engineering." What they described, over and over, was something much simpler, and much harder to deliver.
They want AI that actually changes how work gets done. Not a tool to play with. A system that delivers.
The gap between buying AI tools and getting results from them
Every founder we spoke to had tried AI. Every single one. ChatGPT, Claude, Gemini, most had bounced between all three. And almost universally, they described the same arc:
Week one: excitement. They draft emails faster, summarise documents, brainstorm marketing angles.
Week four: plateau. The novelty wears off. They're still copying and pasting between tabs. The AI doesn't know their business. It can answer questions, but it hasn't changed a single workflow.
Month three: the tools sit unused. The subscription auto-renews. Nobody can explain what ROI looks like.
"We bought three different AI tools last year. Used all of them for about a month. Now they're just line items on the credit card statement."
"AI is like hiring a brilliant consultant who gives you a 50-page report and then disappears. Great insights. Zero implementation."
Anthropic's recent study of 81,000 Claude users found that 81% said AI had already taken a step toward their vision. But when we drilled into what SME owners specifically wanted, not what they'd settle for, the picture was more complicated.
They aren't asking for better AI tools. They're asking for someone to make AI actually work inside their business.
What 100 SMEs actually want
We classified each founder's primary desire into categories. The pattern that emerged mirrors some of what Anthropic found at massive scale, but with a distinctly operational flavour unique to people running businesses with 2–50 employees.
1. AI that targets the right problems first — 28%
The single largest group wasn't asking for more AI capabilities. They wanted clarity on where to start. They'd seen the demos, read the case studies, attended the webinars, but couldn't bridge the gap between "AI can do amazing things" and "here's specifically what it should do in my business."
"Everyone tells me AI will transform my business. Nobody tells me which part of my business to point it at."
"I've got 30 employees and about 200 workflows. I don't need someone to sell me on AI. I need someone to walk through my operations and tell me which five things are worth automating."
This desire for identification, figuring out what's actually worth building before anything gets built, was the dominant theme. Most founders felt overwhelmed by possibility and paralysed by choice. They didn't need more options. They needed a filter.
2. Solutions built for their specific workflows — 24%
The second most common frustration was that every AI tool they'd tried was generic. Off-the-shelf chatbots. One-size-fits-all automation templates. Nothing that reflected how their specific business actually operates.
"I've been in business 14 years. My workflows are specific. My client communications have a tone. My approval processes have steps. No off-the-shelf AI product knows any of that."
"I don't want to adapt my business to fit an AI tool. I want AI that adapts to how we already work."
This echoes what Anthropic found under "cognitive partnership" and "professional excellence", but for SMEs, the need is concrete and operational. They need AI that's configured around their existing systems, integrated into their existing tools, and built to handle their actual workflows, not a demo version of someone else's.
3. Team adoption, not just deployment — 19%
Nearly one in five founders said their biggest fear wasn't that AI wouldn't work, it was that their team wouldn't use it. They'd seen it before with CRMs, project management tools, and other software implementations. The technology was fine. The adoption failed.
"I've rolled out three different ‘game-changing’ platforms in the last two years. My team used each one for about six weeks before going back to spreadsheets and WhatsApp."
"If my team doesn't understand it, trust it, and actually use it every day, it doesn't matter how good the AI is."
They wanted hands-on training. Someone to sit with their team, show them how the new systems fit into their daily routines, and stay long enough to make sure the change actually stuck. Not a handover. Not a user manual. Real adoption support.
4. Security, governance, and trust — 16%
Nearly one in six founders raised data security unprompted. They'd heard about hallucinations. They worried about confidential client data. Several had actively avoided AI adoption because they didn't understand where their data was going or who could see it.
"My clients trust me with their financials. I can't just pipe that through tools I don't fully understand."
"I need to know that whatever we build has proper access controls, that it's auditable, and that it won't hallucinate something into a client-facing document."
5. Measurable ROI before they scale — 13%
The remaining group was blunt: show them proof it works before asking them to invest further. They wanted pilots, not promises. A working proof of concept they could test in real workflows before committing to a full build.
"I don't want a pitch deck about what AI could do. Build me a pilot for one workflow. Let me see the results. Then we'll talk about scaling."
The light and shade: what excites and what worries them
Borrowing from Anthropic's framework, we found the same entanglement of hope and fear in our sample. The founders most excited about AI were also the most anxious about it. These tensions weren't between different people, they were within the same person.
Efficiency vs. dependency. The founders most eager to automate also worried about losing the institutional knowledge that made their businesses work.
Speed vs. trust. They wanted AI to move fast, but they'd been burned by fast implementations before.
Cost savings vs. team impact. Even founders eager to automate felt uncomfortable about what it meant for their people.
Transformation vs. disruption. They wanted change, but manageable change. Rolled out in phases. Tested before scaling. With their people trained and confident before anything went live.
This mirrors Anthropic's finding that the people most engaged with AI's upside tend to also be the most engaged with its downside. It's not optimists versus pessimists. It's leaders grappling with real complexity.
Why most AI initiatives fail SMEs
The pattern in our interviews was clear. It wasn't the technology failing. It was the implementation.
Wrong starting point. Founders picked the AI use case that seemed most exciting rather than the one that would deliver the most impact.
No integration. AI tools sat alongside existing workflows rather than being embedded into them.
No follow-through. The AI was deployed but never adopted. No training. No workflow redesign. No ongoing optimisation.
"We didn't fail at AI. We failed at change management. The AI was fine. Everything around it wasn't."
What we're building at Lumis, and why
This is why Lumis exists.
We don't sell AI tools. We don't push a specific platform. We help businesses identify the AI opportunities that will actually move the needle, build solutions tailored to their workflows, and train their teams to make it stick.
Identify
Decide what's actually worth building. We walk through your operations, interview your team, and narrow the work down to what will create measurable impact.
Develop
Build it right so it works from day one. We integrate into your existing systems and design for reliability, security, and real-world use.
Adopt
Make AI part of how work actually gets done. We train your team, refine the systems with real feedback, and stay long enough for the change to stick.
The bottom line
Anthropic's study of 81,000 people found that the underlying desire behind most AI visions isn't about working faster, it's about living better. Our 100 SME interviews found the same thing, expressed through the lens of business: they want their time back, their headspace back, and the confidence that nothing important is slipping through the cracks.
The technology to deliver that already exists. What's been missing, for most businesses, is someone to identify the right opportunities, build the right solutions, and make sure the team actually adopts them.
That's Lumis. We don't just talk AI. We deliver it.
AI is here. Most businesses will react. The ones with a plan will lead. We build for those ones.
Want to see where AI fits in your business? Talk to us at lumis.ai — we'll walk through your workflows and show you exactly what's worth building, before you commit to anything.