AI is apparently a bubble. At least that’s what the headlines keep shouting. The next dot-com. The hype train has run out of track. Fair enough - valuations look frothy, VCs are tripping over each other, and some pilots have gone nowhere. But let’s not confuse Wall Street drama with whether the tools actually work. Markets blow hot and cold. Workflows don’t care.
Strip out the noise and the story is simpler: a lot of AI tools are already saving people time, money, and grief. That 14% productivity bump in call centres? Real. The part where junior agents suddenly look like seasoned pros? Also real. The writing, summarising, and coding tasks that used to take hours? Now trimmed to minutes. Not perfectly, not everywhere - but enough to matter. The OECD, IMF, NBER, pick your acronym - they’ve all clocked the same thing: this tech does something. It nudges work away from drudge tasks and closer to the stuff humans are actually paid to do.
The Bubble Narrative: Useful context, not the main story
Let’s start with the background noise. Stock prices of AI platform vendors and chipmakers have rocketed. Venture capital has been flinging cash at anything with “AI” in the deck. Analysts warn we’re in dot-com territory again. They might be right about valuations. They might be right that too many startups are chasing thin margins. They might be right that some firms will flame out spectacularly.
And here’s the part that matters less to you as a leader of an SME or portfolio company: the tools won’t disappear. A bubble is about asset prices detaching from reality. Even if prices crash, the actual software doesn’t suddenly stop working. If anything, market corrections often shake out the noise and leave the stronger, more useful tools cheaper and easier to access.
So yes, bubble talk is fair game. But obsessing over it misses the point. For leaders, the real question is whether these tools make work faster, smoother, or more profitable. Spoiler: they do.
The Evidence: AI works (when used properly)
The sceptics want to know: where’s the proof? Fortunately, there’s plenty.
Take the “Generative AI at Work” study. Thousands of call-centre agents, a staggered rollout of AI support, and a clear result: a 14% productivity gain. For junior staff, the boost was more like 34%. The tool didn’t just speed them up - it transferred tacit knowledge, standardised best practice, and even improved customer satisfaction. That’s not hype. That’s measured, peer-reviewed evidence.
Or look at the OECD’s reviews. Across multiple experiments, generative AI consistently cut time on writing, summarising, translation, and basic coding tasks. The productivity gains weren’t small. We’re talking 5% on the low end, 25%+ on the high end. The strongest results came in structured, language-heavy work. Exactly the kind of grind that fills SME inboxes every day.
Then there’s the Dillon et al. field experiment. Knowledge workers with integrated AI tools shifted how they spent their time: less drafting and revision, more synthesis, review, and decision-making. That’s not just speed - it’s upgrading the type of work people are able to do.
And in sector reviews - FinTech, Retail, Manufacturing - AI adoption is linked to measurable productivity lifts through automation, predictive analytics, and personalisation. Not uniform, not magical, but statistically real.
The evidence is stubborn. Whatever the stock market does or doesn't do, AI tools are delivering practical gains.
Why SMEs and PE-Backed firms should care more than most
At this point, the corporate giants are experimenting too. But they have room to waste money on doomed pilots. SMEs don’t. That’s exactly why the gains matter more for smaller and PE-backed firms.
Every saved minute counts harder. A 10% productivity gain in a 30-person business could be the difference between staying ahead or running on fumes. Lean teams feel the impact of automation immediately.
Less legacy drag. A giant with decades of entrenched systems can’t shift quickly. An SME can roll out AI tools and redesign processes without wrestling with ten layers of old IT. The same goes for portfolio companies that PE owners are reshaping - they’re already in flux, so change is easier to bake in.
PE discipline fits perfectly. Investors want measurable improvements in cost, cycle times, retention. AI tools are strong levers for all three. Trim reporting cycles with automated drafting. Cut SG&A costs with AI-assisted admin. Improve retention with faster customer response. These are not shiny demos - they are line-item improvements.
And there’s the levelling effect. AI lets a small firm produce pitch decks, client reports, or customer service responses at a standard once reserved for large enterprises with more staff. That kind of leapfrogging changes competitive dynamics.
The Pitfalls (and how not to fall into them)
If the evidence is so positive, why are there still failed pilots and frustrated leaders? Because AI is often misapplied or badly managed.
Not every task is a good fit. The so-called jagged frontier is real. Use AI to draft, summarise, triage, or scan - clear, bounded tasks. Don’t expect it to crack strategy or make judgement calls that require deep domain nuance. In some consulting experiments, consultants did better on writing tasks with AI, but performance dropped on strategic synthesis. Wrong tool, wrong job.
Management practices matter. The ONS found UK firms with strong planning and accountability were far more likely to adopt AI successfully. AI doesn’t fix poor management. It amplifies what’s already there. A sloppy firm gets sloppier. A disciplined firm gets sharper.
The people factor matters even more. Employees worry about job threats. They don’t trust the outputs. Or they just don’t know how to use the tools effectively. Left unchecked, resistance kills adoption. The cure is training, guardrails, communication. Let people test small wins. Build confidence slowly. Fear ebbs when staff see the tool making their day easier rather than threatening their role.
And finally: avoid vanity pilots. Too many firms run projects chosen for novelty, not impact. Then they die quietly. Pick use-cases that actually matter - like speeding up reporting cycles or cutting customer service response times. Define success in numbers. Kill weak pilots early. Scale the winners.
What if the bubble really does burst?
So let’s imagine the bubble does pop. Nvidia’s stock halves. Startups implode. Journalists move on to the next tech fad. What happens then?
Vendors consolidate. Some fringe outfits vanish. But the core capability - language, code, analysis - sticks around. It’s already embedded in mainstream platforms.
Costs probably fall. Market corrections often commoditise technology. That’s good news for SMEs who need lower barriers to entry.
The maturity gap widens. Early adopters with proven use-cases get stronger, while late adopters find themselves scrambling, forced to buy tools on less favourable terms. Advantage accrues to those who already built AI into their processes.
And the productivity gains? They don’t evaporate just because valuations crash. Hours saved remain saved. Customer response times stay improved. Workflows don’t backslide because the FTSE took a hit.
In fact, for SMEs, a burst bubble might be a net positive: cheaper tools, less noise, more space to focus on utility.
A Checklist for Leaders: How to treat AI like a lever, not a lottery ticket
Here’s the pragmatic playbook:
- Start narrow. Pick one or two workflows where the pain is obvious - customer emails, proposal drafting, finance reconciliation.
- Define success in numbers. Hours saved, cost per transaction, turnaround times. No fuzzy metrics.
- Run lean pilots. Small teams, quick iterations. Don’t overcommit.
- Scale fast when it works. Embed it into the daily process. Don’t let momentum die.
- Invest in training and trust. Staff need to feel safe, not sidelined. Build adoption step by step.
- Stay vendor-agnostic. Don’t weld yourself to one platform. Better options will keep emerging.
- Measure adoption as a KPI. Track usage, error rates, value. Treat AI like any operational tool.
- Cull vanity projects. Kill what doesn’t deliver. Quickly.
- Leave room for flexibility. Keep some budget and headspace for new tools. The landscape won’t stand still.
After all is said and done: Ignore the Bubble Noise, follow the Utility
So yes, maybe AI is in a bubble. Maybe valuations are ridiculous. Maybe we’ll get a wave of ugly headlines when the hype train derails. But underneath, the tools keep doing their job. Drafting faster. Summarising quicker. Cutting hours out of admin. Lifting junior staff up the curve. Making SMEs more competitive and portfolios more profitable.
Bubbles burst. Utility persists. For leaders, the choice is simple: chase the headlines, or chase the workflows. The firms that do the latter will be the ones still standing when the noise has moved on to the next big thing.