Literacy: Your Best AI Hire Might Already Be on Your Team
When I talk to businesses about AI literacy, the conversation almost always starts or ends up in the same place, with them asking whether they should hire AI specialists.
It’s an understandable instinct. AI is technical, unfamiliar, fast moving, and also alien to most, so the natural response is to bring in people who already understand it. A few businesses will benefit from that, but most won’t, or at least, not as their first move.
The best question that any business should ask themselves is, “who do we already have that knows the business, and could become dangerously good with AI?”
As an Automation and AI business, we’ve got a little more insight than most into what’s actually required, because if there’s one thing we’ve learnt, it’s that domain knowledge plus AI fluency beats AI fluency without domain knowledge, every time. When we started doing this, there were more people who’d walked on the moon, than had experience of deploying domain-specific AI tooling in any given sector. An AI specialist who doesn’t have commercial awareness, business acumen, or a customer-first mindset, will lack sufficient context and will always be working with one hand tied behind their back. The domain expert who becomes AI literate has both hands free.
We’ve already run this experiment once
This isn’t a new phenomenon. It’s the same one the automation industry and business world learned over the last decade, and it’s played out long before generative AI arrived.
The RPA programs that actually delivered value weren’t the ones stacked with technical automation specialists working in isolation. They were the ones resourced with people who understood the processes of the business being automated, with just enough technical capability to shape the automation themselves, with support where needed most. The programs that failed or fizzled out were usually the ones where automation was treated as a purely technical project, handed to a team that understood the tooling but not the work.
AI has the same fact pattern, just much faster but with higher stakes. The accountant who understands exactly where clients get nervous, where the judgement calls really sit, and where numbers looks right or not, is worth far more once they’re AI literate than an AI specialist will ever be without that context. A recent McKinsey research made this same case too, using the example of a metallurgist, saying that: “it’s easier to teach a metallurgist to use AI than it is to teach an AI specialist metallurgy”. If you swap metallurgist for senior manager in an accountancy practice for example, the point still stands.
Literacy isn’t a training day
In reality, most businesses aren’t really approaching literacy at all, and are relying solely on the ease of use of the popular AI platform they’re using plus an instructional video or two. Then there are the ones that go a step further with some one-off training sessions, maybe an afternoon workshop or a boot camp, but then they assume that’ll be enough for everyone to be “AI literate” from that point forward.
That’s really not literacy. That’s barely an introduction. That wouldn’t even be enough for someone new to your accounting platform to be proficient or to be self-reliant, so it certainly won’t be enough for AI.
Real AI literacy is layered, and it looks different depending on the role. A partner needs to understand enough to make judgement calls about when AI output can be trusted and when it needs real scrutiny. A manager needs enough hands-on fluency to actually use the tools productively in daily work. Someone building or configuring agents needs a meaningfully deeper technical grounding again. Treating all three as the same “AI training” need is exactly how businesses end up with a room full of people who’ve seen a demo, and almost nobody who can actually be trusted to use AI well unsupervised.
This is also where the earlier governance conversation and literacy start to overlap. An AI-literate senior manager is one of your best governance controls. They’re the person who notices an output doesn’t look right, who knows to ask what data an agent actually touched, who understands enough to push back rather than accept confident-sounding nonsense at face value. Literacy isn’t just a productivity lever. It’s a risk control that happens to also make people better at their jobs.
What “dangerously good” actually means
I use the phrase “dangerously good” deliberately. It’s not competence or comfortable, but dangerously good, in the sense that a domain expert with real AI fluency can move faster and see further than most businesses currently think is even possible, precisely because they’re able to combine two things that rarely sit together, and that’s deep contextual judgement with genuine technical capability.
That requires more than attending a session on ChatGPT or CoPilot. It means structured skill progression. It means knowing what dangerously good actually looks like at each stage, and having a way to track whether someone’s getting there or not. Most businesses have no way of knowing how AI literate their team really is, beyond maybe a gut feeling or initial outcomes.
Drowning in manual, repetitive work? Tell us the task and we’ll show you what to automate.
And that’s not really good enough for the serious phase of AI. If the exploring phase was about knowing what your AI agents can do, literacy is about knowing what your people can do with AI Agents, and right now, most businesses genuinely can’t answer that.
Practical starting points
In each part of this series, we’ll leave you with some practical next steps and some links to helpful tools to get you started, so here’s your AI Literacy imperatives.
Stop starting with “who should we hire?” Start with “who do we already have who should become AI literate first?” Look for the people who already understand the process, the client, or the risk better than anyone else in the room.
Build tiered literacy, not one-size-fits-all training. Partners, managers, and agent builders need different depths of understanding. A single generic session serving all three levels well is a myth.
Treat literacy as a skill progression, not an event. Define what “literate” actually looks like at each level, and track progress against it the same way you’d track any other professional development.
Use your AI-literate people as an informal governance layer. The people who understand AI well enough to question its output are one of your best early-warning systems for the risks.
Get a real baseline before you invest in training. Guessing at your team’s current AI literacy wastes budget on training that either repeats what people already know or skips what they actually need.
Know where you stand
Most businesses can tell you who’s “good with AI” based on outcomes alone. Far fewer can tell you, with any real structure, what each person has actually learned, what they still need to, and where the gaps sit across the wider team.
Our AI Literacy Assessment gives you exactly that: a structured framework for each individual, covering the skills they’re working toward and the ones they’ve already completed, building an actual training and literacy record rather than a vague sense of who seems confident with the tools.
Contact us to take the AI Literacy Assessment
Next in this series: Part 4, AI Tokenomics: The True Cost Lands Only After the Pilot.
Daniel Lawrence is the CEO and co-founder of bots for that, creators of the Automation Operating System (AOS) and Agent Console. He has spent more than a decade deploying enterprise automation and AI in regulated industries including accounting and professional services. AI: The Serious Phase, is a six-part series exploring what AI-native operations look like for mid-tier and large UK accounting firms.
