AI needs to prove itself where it matters most
People don’t step back from AI because they’re fearful of the technology itself. They step back when it doesn’t align with how their business actually works, writes Sage CEO Steve Hare.
A few weeks ago, I spoke with a small business owner who had heard AI could help with her accounts.
She wasn’t sure how. She just knew she was spending hours in her evenings and weekends on invoices and spreadsheets when she’d rather be growing her business.
That conversation stuck with me because it captures exactly where we are with AI in 2026.
The real challenge for AI this year is not whether it’s powerful or even whether it’s impressive. It’s whether people can trust it, understand it, and confidently use it in the moments that actually matter to them.
People aren’t asking for more intelligence in the abstract.
They’re asking for fewer late nights, fewer errors, and more certainty that the numbers they rely on are right. They’re looking at how to save time and make better use of what time they have.
I’ve had dozens of conversations like this with small business owners throughout the course of the year.
They’re not sitting up late debating multimodal models or hallucination rates. They’re thinking about cash flow, payroll, sales, growth, late payments, compliance, and whether they can afford to hire that extra person.
If AI is going to matter to these people, it has to solve those problems and pain points – reliably, predictably, and without adding new complexity or risk, not in theory, but in practice.
This is where much of today’s AI hype begins to fall apart. The focus should be real solutions for real people, not demonstrations designed to look impressive but fail under day-to-day pressure.
Trust is the real measure of AI success
We’re seeing more “AI theatre”: big promises, confident demos, and sweeping claims that don’t stand up once real data, workflows, and accountability are involved.
Over-promising might generate headlines, but when tools don’t work as expected, people lose time, confidence, and trust. And in areas like finance and payroll, that loss of trust is hard to win back.
For me, the most important question for 2026 is not how intelligent the model is. It’s whether AI can consistently deliver outcomes people can depend on, week in and week out.
How do we put tools into the hands of entrepreneurs that save them hours every week, reduce errors, and give them confidence in the results, so they can focus on growing their business?
Put simply, it’s not about “How advanced is the model?” it’s “How did this make someone’s job easier, quicker, better?” Or even more simply: “Did this help someone get home on time?”
AI has genuine potential to transform how businesses work. But that potential only matters when it proves itself in the everyday realities of running a business and when people feel confident using it.
In the world of finance and payroll, AI tools need to be accurate and predictable. They need to automate the right tasks, in the right way, while keeping humans in control. That’s how people build trust over time and adopt new technology at a pace that feels right for them.
Adoption isn’t a single moment; it’s a journey built on confidence. It won’t happen everywhere at once, and that’s exactly as it should be.
Businesses don’t embrace change because someone tells them to. They move when the new way of working delivers clear results, when errors go down rather than up, and when people feel skilled, supported, and in control.
Trust isn’t trained into existence, it’s earned through consistent performance.
So, what should business leaders watch out for in 2026?
First, AI hype and AI theatre will backfire.
Announcing bold plans that never materialise, or launching tools that look clever but don’t fit real workflows, damages credibility.
Over-promising creates hidden costs: wasted time, manual workarounds, and scepticism that slows future adoption. People notice, even if they don’t always say it openly.
Second, general-purpose AI will increasingly fall behind more focused solutions.
Most businesses don’t need an AI system that can do everything. They need one that does a small number of important things extremely well and does them the same way every time.
The AI that succeeds will be bespoke, tailored to specific workflows, and trusted with sensitive data. It will reduce errors, behave predictably, and sit quietly inside everyday systems, proving its value through results rather than promises.
What does this mean in practice?
It means we need to be more disciplined in how we talk about and deploy AI.
For businesses, AI should be treated less as a headline topic and more as part of the underlying infrastructure technology that supports everyday work, reliably and over time, rather than something designed to impress in isolation.
Over-promising doesn’t just risk disappointment; it can slow adoption by undermining confidence and creating unnecessary scepticism.
People don’t step back from AI because they’re fearful of the technology itself. They step back when it doesn’t align with how their business actually works.