“Don’t be seduced by the sycophantic nature of these LLMs,” Lisa Moretti says. It’s a cracking line, but it also sums up her whole approach to AI: curious, practical and just sceptical enough to stop people doing something stupid.

Moretti is no anti-tech doom merchant. Far from it. She loves what technology can do. But she’s also spent enough time across journalism, research, policy and product to know that the AI gold rush has left plenty of businesses dazzled by the sales pitch and a bit dizzy on the reality.

While AI can absolutely save time, sharpen ideas and help one person do the work of three, it can also churn out nonsense, flatten your brand voice and send sensitive data who-knows-where if you’re not careful. And that, in a nutshell, is Moretti’s message: use AI, by all means. Just don’t hand over the keys and hope for the best.

From journalist to AI expert

Moretti’s interest in AI didn’t begin in a lab. It started in journalism, right when the industry was being turned upside down.

“I started out as a journalist, and when I entered the world of work, I was part of that early generation of journalists who were experiencing transformation of the publishing industry in their first year of practice,” she says.

Back then, publishing was shifting under everyone’s feet. Blogs were booming, Web 2.0 had arrived and the old gatekeepers were losing their grip. After a few years in journalism and content marketing, Moretti went back to uni and completed a master’s in digital sociology. That’s where the AI obsession kicked in.

“It was through that that I learned about artificial intelligence, machine learning, the datafication of society. And that’s where that interest started and has only intensified and grown since then.”

These days, Moretti is a digital sociologist. Her work is about helping organisations understand how technology is changing society, and how the digital breadcrumbs we all leave behind can be used to make sense of the world. It’s a pretty handy perch from which to watch the current AI circus.

The Ai shift

Ask Moretti what’s really happening in AI right now, and she doesn’t go straight to chatbots or flashy demos. She points to something bigger: concentration of power.

“One of the biggest shifts we’ve seen is the increasing monopolisation of AI,” she says.

In the past, AI was largely tucked away in research labs, universities and government projects. Now it’s a full-blown commercial beast, dominated by a handful of giant private operators.

“I don’t think that that’s necessarily a good thing,” she says. “It means that such a powerful technology is concentrated in the hands of a very few number of players.”

Her concern is that these tools are increasingly shaped by a narrow worldview that doesn’t necessarily reflect the needs, values or realities of the rest of the world, including Australia. That might sound a bit lofty when you’re just trying to use AI to write a product description or sort out next month’s marketing plan, but it matters. The tools small businesses rely on are being built with assumptions baked in, and those assumptions aren’t always a neat fit for local businesses, local customers or local values.

Is AI really replacing workers?

There’s been no shortage of headlines about AI replacing humans, but Moretti says the picture on the ground is messier than that. Some of the workforce cuts pinned on AI are really part of a broader cycle. Tech companies hired heavily during COVID, when demand for digital tools exploded, and plenty have been trimming back since.

But there’s another wrinkle too: some businesses jumped headfirst into replacing people with AI, only to realise they’d made a shocker of a call.

Moretti points to examples like Klarna, which made customer service and marketing roles redundant and then had what she calls “huge AI regret”.

“They suddenly realised that, oh my goodness, this is not beneficial to our customers. They want to speak to people.”

The same goes for quality. AI can mimic, summarise and generate, but it doesn’t really understand context, nuance or what good actually looks like. That’s why Moretti believes some businesses have been sold a fantasy: that AI can be swapped in for a person, like changing a toner cartridge on a printer.

“I think hype is definitely driving some of the redundancies,” she says. “People thinking that AI can like for like replace a human, and then when that happens, realising, oh, actually we’ve made a mistake.”

Why AI can be brilliant for small business

Still, Moretti believes AI can be a seriously useful tool for small business owners.

“For a solo entrepreneur, my goodness, AI has been a huge, huge tool, a brand new kind of Swiss Army knife really.”

That’s where things get interesting for Business Builders readers. A one-person operation or lean startup often doesn’t have the luxury of a full design team, data analyst, strategy consultant and copywriter on standby. AI can help fill some of those gaps, at least to a point.

Moretti says it’s especially helpful for turning ideas into something visible and workable.

“If you are somebody who’s very visual but you don’t have strong expertise or strong skills in data analysis, you can suddenly work with an LLM like Claude to help you crunch through loads of data,” she says.

Equally, if you’re a strong strategist but not a designer, AI can help you get an idea “out of your head onto a page” so you can have better conversations and brief specialists more effectively.

Moretti suggests that’s the sweet spot: using AI to get moving faster, test ideas and build momentum. Not pretending it’s a magical replacement for expertise you don’t actually have.

Avoiding rookie errors

The biggest trap for small business owners is treating AI output like it’s gospel. Moretti says hallucinations remain a major issue, with reported rates varying wildly, but the bottom line is that they happen, and they’re not going away.

“What I read from that data is that hallucinations are happening. We can’t solve hallucinations. They’re a feature of LLMs rather than a bug that can be erased.”

Her suggestion: If you’re using AI to help with content, analysis, code, design or planning, you need to check the work. Properly.

That’s easier said than done when you’re using AI to support a skillset you don’t have in the first place. If you’re not an expert in code, data or design, how do you know whether the output is any good? That’s why Moretti urges business owners to stay sceptical and resist the ego-stroking confidence of AI tools.

“You must maintain a degree of scepticism about these outputs,” she says. “Don’t be seduced by the sycophantic nature of these LLMs.”

She also warns against blindly publishing AI-generated work. She recalls the excitement around AI website-building tools in Figma, only for some outputs to prove inaccessible to people using screen readers and assistive technology.

It’s a useful reminder that “looks fine to me” is not the same thing as actually fit for purpose.

Where to start?

If you’re a small business owner who hasn’t really dipped a toe in yet, Moretti’s advice is to start with the problem, not the tool.

“Think about what your needs are,” she says. “Do you need to create content? Do you need to generate images? Do you need to generate proposals? Do you need to do data analysis?”

That sounds obvious, but plenty of businesses are doing the reverse: grabbing the shiniest AI tool first, then trying to wedge it into every part of the business. Moretti recommends trying one or two tools rather than locking yourself into a yearly subscription straight away.

“Try Claude and ChatGPT and see which one performs better for different tasks,” she says. “Don’t just go all in.”

And if you’re using free versions, tread very carefully with business information.

“I would really caution you not to put in any IP or sensitive data into those tools.”

Instead, she suggests using broader descriptors. Rather than naming your business and giving away specifics, describe the business in general terms and keep private data out of the prompt. Once you’ve tested a tool properly and know it suits your needs, then consider paying for it. But even then, Moretti says it’s worth maintaining a healthy dose of caution.

Be responsible

For Moretti, responsible AI isn’t just a buzz phrase for corporates with legal teams and policy decks. It’s something even the smallest business should be thinking about. One big question is disclosure: when should you tell people AI was used?

Her view is that if AI is doing a big chunk of the work, businesses should be upfront about it.

“If you have written an article and that article is more than 50 per cent generated by AI, I think you do have a responsibility to disclose that,” she says.

More broadly, she worries that over-reliance on AI can slowly erode the very skills that make a business stand out.

“Make sure that you sit down and you write out the kind of key points that you’re looking to include,” she says. “Make sure that you’re doing some of the strategic work and critical thinking upfront.”

That’s because if everyone uses the same tools in the same lazy way, everything starts sounding, looking and feeling identical.

“There is a flattening effect on creativity that starts to happen, and you don’t want that,” she says.

When you run a small business, your edge is often your voice, your perspective, your personality and the odd brilliant way you solve problems. If AI irons all that out, you’re left sounding like everyone else on the internet.

Introduce some guardrails

Moretti says businesses should give their AI policy the same level of thought they give their brand values, mission and vision.

“You should be spending the same amount of time when you’re thinking about your AI policy for your organisation.”

That means being clear about how AI fits with your values, where you want to use it, and where you absolutely don’t. She recommends creating AI principles for the business, using Australia’s national AI ethical principles as a starting point if needed, and putting together a use-case policy so staff know what information they can and can’t put into tools.

“If you are not clear about that, people don’t know,” she says.

“That’s how businesses end up with staff pasting company information into free personal accounts and creating shadow AI use behind the scenes.”

Instead, she says, set the rules early, train people properly and make it clear which tools are approved. Otherwise you’re basically leaving the fridge open and hoping the milk stays cold.

Problem first, tech second

If there’s one idea Moretti wants people to take away, it’s this: stop being tech-centric and start being problem-centric.

“We should be putting problems in the centre and saying, what is this problem? How do I define this problem? What am I looking to do? And then choosing the technology that we feel would best support us in solving that problem.”

That’s a useful circuit-breaker in a market overflowing with hype, pressure and FOMO. You do not, in fact, need to use ChatGPT just because someone on LinkedIn said you’d be left behind by lunch if you didn’t.

Sometimes AI will help. Sometimes it won’t. Sometimes the problem is actually a process issue, a people issue or a plain old strategy issue dressed up as a tech problem. Moretti says businesses are starting to wake up to that. In the next 12 months, she expects many organisations to move into what she calls the “trough of disillusionment”, where the gap between what AI promised and what it actually delivered becomes impossible to ignore.

“We see AI everywhere except in the numbers,” she says, paraphrasing what she’s hearing from clients.

That means more scrutiny, tighter budgets, better measurement and, ideally, a bit less blind faith.

Australia’s not as behind as we think

Despite the hand-wringing, Moretti doesn’t buy the idea that Australia is hopelessly behind the rest of the world on AI.

“Every time I speak to someone, I get told a lot about how Australia is behind. But I don’t think Australia actually is as far behind as they think they are.”

From her perspective, plenty of organisations in the UK are still grappling with low AI literacy, patchy policies and no clear strategy. The bigger gap, she says, is in the local responsible AI ecosystem, which feels more established overseas than it does here. Still, that may change as more businesses realise that charging ahead without policies, principles or proper training is a recipe for chaos.

Really, that’s the heart of Moretti’s argument. AI is useful. Exciting, even. But it’s not magic, it’s not neutral, and it’s definitely not a substitute for judgment.

Lisa Moretti was a speaker at Southstart. The author was a guest at the event.

Want more? Get our newsletter delivered straight to your inbox! Follow Business Builders on Facebook, Twitter, Instagram, and LinkedIn.