Driving AI Adoption: how to stay in the driver’s seat

There is a moment that probably sounds familiar: you are about to use AI for something (drafting an email, writing a bit of code, putting together a presentation) and you pause. “Should I be doing this?” or “Is it actually any good?”
Most people have had that feeling more times than they would like to admit. AI is increasingly finding its way into daily workflows, and it’s worth putting into words what actually drives that switch, and at what point it becomes so obviously convenient. The honest answer, on reflection, comes down to something simple: how time and effort are allocated has quietly shifted. Here’s how that shift plays out, and, further down, how to try it without the guesswork.

The 80/20 flip

The clearest sign of that shift is a natural, if astonishing, inversion in the ratio of effort to results. It echoes the Pareto principle, or 80:20 rule, playing out in reverse: where roughly 80% of effort used to go into producing a first version of something, and 20% into deciding whether it was any good, that ratio is turning on its head. Less effort goes into the doing, more into the judging, the framing, the deciding.

Developers feel this directly: it’s now possible to get a considerable result with far less development effort than before, which leaves much more time spent being an architect than a mechanic. It’s not just a feeling, either: GitHub’s own research on Copilot found developers completing coding tasks 55% faster, and with a higher success rate, than those working without it.

Once that ratio flips, a few things become possible that weren’t before:

• Trying more than one option before committing. Instead of picking a direction and hoping it works, you can sketch two or three and compare them before spending real time and money.

• Making unfamiliar territory less intimidating. It can explain a concept in plain terms, break a vague task into smaller questions, or flag what you might be missing, without you needing to already be the expert.

• Freeing up time for the part that actually needs a person. Once the first pass takes minutes instead of hours, there’s more time left for weighing options, catching mistakes, and deciding what’s actually worth pursuing.

This plays out in two pretty different ways, depending on who’s holding the tool. If you already know the subject well, it’s a real boost, like finally having on tap the kind of specialist assistants you always wished you had. If you’re still learning it, whether that’s a new domain in general or a new job at a new company, it does something else: it shortens the learning curve, cuts down on how much hand-holding a newcomer usually needs, and lets you ask a question and work through the answer yourself before interrupting someone busier than you. Neither replaces the years of experience, but both shorten the path to a useful contribution.
None of that comes free, though.

Where AI Can Hurt You

You can only catch a wrong answer if you already knew what a right one looked like. Ask it to summarise a report, and it can invent a figure that was never there, stated with exactly the same flat confidence as the parts it got right; you’d only notice if you already knew that report well. Ask for the quickest way to cut costs without knowing which constraint actually matters most, and you’ll get a fast, confident answer that’s wrong for what you needed, and you won’t find out until it’s too late. Or you’re building something genuinely new, and you never ask it to check its own suggestions against what already exists, past decisions, related work, the sources it could actually be tested against, so it invents something plausible in isolation, disconnected from everything around it that could have grounded it.

If you haven’t started yet

So, should you be doing this? Is it actually any good? Yes, and yes, it turns out. Here are a few triggers and quick suggestions worth keeping in mind to get started:

• Talk about it. Don’t just go down your own rabbit hole, however deep or interesting it gets. Share what you find, encourage others to try it too, and don’t let it become a taboo.

• Push it past the first answer. Go back and forth, refine the prompt, test where it starts to break. That extra round either pays off in the result, or, at worst, sharpens your own sense of what counts as good work.

• Don’t limit it to your job description. Push a little past your usual boundaries and see if it can help with other transversal processes too. We know that clean, well-defined tasks aren’t all we deal with.

This is the kind of practical, no-hype approach we try to bring to AI adoption at COMMpla, in our own work and alongside the teams we support. If you’re weighing up where AI could actually help your team, get in touch, or have a look at what we do at commpla.com.

Elio Colligiani | Full-stack Developer