Situational Awareness for Small Businesses: Why AI Progress Won't Wait Until You're Ready

Situational awareness, applied to AI, means making business decisions based on where AI capability is heading rather than where it sits today. Most small businesses do the opposite. They judge AI by what last year's tools could do, find them underwhelming, and conclude there's time. I don't think there is, and the numbers back that up. AI capability has been improving on a strangely consistent trendline for over a decade now, and the gap between businesses that adapt early and the ones that wait is already showing up in survey data.

The idea comes from one of the most argued-about documents in AI, Leopold Aschenbrenner's Situational Awareness: The Decade Ahead. He wrote it for policymakers and investors. This post is an attempt to translate it for people who run accounting practices, agencies, online shops and manufacturing firms instead.

Who is Leopold Aschenbrenner, and why does his essay matter?

In June 2024, Aschenbrenner (previously a researcher on OpenAI's Superalignment team) published a 165-page essay series arguing that almost everyone was underestimating how fast AI would improve. He wasn't subtle about it either. "Few have the faintest glimmer of what is about to hit them" is from the opening pages.

The method behind the claim is simple enough to fit on a napkin: ignore both the hype and the scepticism, and trust the trendlines. AI capability, he argued, grows along three curves that multiply together. More compute, because the processing power thrown at training keeps scaling up. Better algorithms, because researchers keep squeezing more intelligence out of each unit of that compute. And what he called "unhobbling", which is the odd one, and the one that turned out to matter most for ordinary businesses. Unhobbling means models becoming dramatically more useful without becoming any smarter, simply by being given tools, memory, and permission to act. Browsing. Writing and running code. Using software. Finishing a multi-step job on their own.

Stack those three on top of each other and the leap from 2024 to the late 2020s, he reckoned, would rival the leap from the toy chatbots of 2019 to GPT-4. Roughly: preschooler to smart secondary schooler, a jump almost nobody priced in until it had already happened.

Two years on, the essay has aged uncomfortably well. The unhobbling prediction in particular describes exactly what we've all watched happen. The biggest change in AI usefulness since 2024 hasn't been smarter chatbots. It's been AI that does things: agents that research, draft, reconcile, schedule, and grind through work that used to need a person at a keyboard.

You don't have to buy his boldest claims about superintelligence (plenty of serious people don't). The conservative version is enough: AI capability will keep compounding for the foreseeable future, so planning around today's tools is planning for a world that's already gone.

What does situational awareness mean for a small business?

You'd think trendline-watching matters mostly to governments and AI labs. I'd argue it matters more to a 12-person business, for an unglamorous reason: big companies can afford to be late. They have innovation teams, vendor relationships, and enough budget to buy their way out of a slow start. A small business has none of that. What it has instead is speed. An SMB can go from "let's try this" to a changed daily workflow in a few weeks. But that advantage only pays out if you move while it still counts as early.

And the adoption gap has stopped being hypothetical. In the UK, 54% of SMEs are now actively adopting AI according to the British Chambers of Commerce, up from 35% in 2025 and 25% the year before. Adoption roughly doubled in two years. Salesforce surveyed over 3,350 SMB leaders and found 83% of growing businesses have adopted AI against 55% of declining ones. Correlation isn't causation, sure, but that's a gap you'd rather be on the right side of. And when UK businesses are asked what's holding them back, the answers aren't cost or regulation. In government research covering 3,500 businesses, the most common barrier was simply not having identified a need for AI (71%), followed by limited skills and expertise (60%).

Put those together and the picture is fairly stark. The tools are compounding, your competitors are adopting, and the two things stalling everyone else are nobody having pointed out where AI fits their business, and nobody knowing how to use it once it's there. Both of those are fixable. Neither fixes itself.

So the small-business version of situational awareness has nothing to do with predicting AGI timelines. It's one honest question: if AI tools keep improving at the rate of the last five years, what does my industry look like in three, and where do I want to be standing when it does?

The three planning mistakes the trendline exposes

The first is evaluating AI by your worst experience with it. A lot of owners tried a chatbot in 2023, got something mediocre, and filed the whole category under "overhyped". That's judging the internet by a 1996 dial-up session. The question was never "did this tool impress me", it's whether the category is improving and how fast.

The second is buying tools instead of building capability, and this one is everywhere. The UK data shows most SME adoption is still shallow: ChatGPT subscriptions and Copilot licences, no structured workflows underneath. Tools churn. The skill of working with AI compounds. A team that learns to delegate real work to AI this year absorbs every future upgrade for free; a team that got a licence and a one-hour demo does not.

The third is waiting for things to settle down. They won't, which is the entire point of the trendline argument. "We'll look at AI properly once it stabilises" quietly assumes the curve flattens. That assumption has been wrong every year so far, and each year of waiting costs more than the last because the gains keep compounding for whoever didn't wait.

What should a small business actually do? Three moves for this quarter

Awareness without action is just a new thing to feel anxious about, so here's the practical response. It fits in a quarter.

First, run an honest AI readiness audit. Not a 40-page strategy deck. An afternoon spent listing where the business burns repetitive hours: writing, summarising, researching, data entry, customer responses, reporting. Those are the workflows current AI already handles well. Estimate the hours, rank by value.

Second, ship two or three quick wins inside 30 days. Take the narrowest, highest-frequency tasks from that list and change how they're done. Personalised sales outreach. Meeting notes and action points. First drafts of customer replies. Small visible wins do two jobs at once: they save real hours, and they convince your team that AI is a daily tool rather than another corporate initiative that fizzles.

Third, train for the trendline, not the tool. Teach people how to work with AI on their actual workflows: how to delegate, how to verify, how to build prompts around the job in front of them. Vendor-specific button knowledge depreciates in months; the working habit compounds. This is also the right moment for guardrails, meaning clarity on what data can go into which tools (GDPR and POPIA both apply) and which decisions stay human.

If you'd rather have a partner for that process than a to-do list, that's literally the job at Vyrion. Book a free consultation and we'll tell you what's worth doing in your business, and just as importantly what isn't. No slideware, no six-month roadmap.

Frequently asked questions

Is Aschenbrenner's essay credible, or is it hype?

Genuinely contested. Critics say he extrapolates trendlines too confidently and underweights bottlenecks in energy, chips and data. Supporters point out that his specific near-term predictions, especially agentic unhobbling, largely came true. For business planning the dispute barely matters, because even the conservative reading (steady compounding improvement, no intelligence explosion) is enough to make wait-and-see the riskiest option on the table.

My business is small and local. Does any of this apply to me?

US Census Bureau survey data shows around 82% of businesses with fewer than five employees don't see AI as relevant to them, and the UK picture rhymes: "no identified need" is the most common reason British businesses give for not adopting. In our experience most of them are wrong. Not because they need anything cutting-edge, but because they spend hours every week on writing, admin, research and customer communication that current tools already handle well. The use cases exist. Nobody has pointed them out yet.

Do I need to spend a lot to get started?

No. Most meaningful first steps run on tools costing somewhere between nothing and £30 per person per month. The real investment is attention: a few focused hours finding the right workflows, then the deliberate effort of building the habit across a team. Which fits what UK businesses themselves report: the barriers they name are not seeing the need and lacking the skills, not the price of the tools.

What's the single biggest mistake to avoid?

Buying tools before identifying problems. Start from the workflow ("we spend six hours a week on proposals") and work backwards to the tool, never the other way round. That ordering is the difference between adoption that sticks and a stack of unused licences.

---

Vyrion Tech provides practical AI consulting and team training for small and medium businesses in the UK, South Africa, and globally. If you'd rather work the trendline than watch it, start with a free consultation.