There are two conversations happening about AI inside most organizations. One is in the boardroom, where the language is transformation and efficiency. The other is in the break room, where people think: how are we going to keep up?
We’ve had a front-row seat to both. avertra has been working with AI and automation in the utility world long before any of this was fashionable, and we’ve run plenty of our own experiments, made our own mistakes, and watched partners stumble into the same traps we did. Here’s the honest take from our experience, including the parts most companies would rather not say.
Start by framing AI as subtraction and you poison everything downstream. The most damaging move a leader can make is to introduce it as a way to do the same work with fewer people. The moment the question becomes “how many roles does this replace,” you’ve already lost the room. People read the tool as a threat and act on it. Some hide how much they lean on it, so you lose all visibility into what’s actually working. Others quietly refuse to get good at it, because who wants to sharpen the knife meant to cut them? The organizations that move ahead ask a warmer and more useful question. What can our people finally attempt that used to be out of reach? That framing tends to produce energy where the other one produces fear.
Optimizing for output is its own trap. AI makes it trivial to produce ten times more of everything, so companies do, and their audiences respond by tuning out the sameness. When every team reaches for the same tools and the same default prompts, the work converges into one flat tone, polished on the surface and hollow underneath. There’s a speed illusion tangled up in this too. A draft appears in thirty seconds and everyone celebrates, while nobody counts the hours later spent fact-checking it, fixing the tone, and cleaning up the parts that were confidently wrong. For anything that reaches a customer or feeds a real decision, the speed of that first draft barely matters. What matters is how long it takes to reach something you would actually stand behind.
The deepest risk is the one almost nobody says out loud. You can quietly hollow out your own people. When juniors never do the reps because the tool does them, they never build the judgment they will need later to tell good work from work that merely looks plausible, right as the volume of the latter explodes. The damage stays invisible quarter to quarter, until it’s too expensive to ignore. In our world that cost is not abstract. A wrong answer can reach a field crew, a regulator, or a household that trusts the lights to come on and the bill to be right. Utilities carry a promise to the communities and regulators they serve, and generic AI pulled off the web has no idea that promise exists. There’s a quieter version of this risk too. When the free tool off the internet is the only option, sensitive customer data and hard-won institutional knowledge get pasted into someone else’s system, on someone else’s servers, under someone else’s terms.
Here’s the part that should give leaders hope. None of this is inevitable. The same technology that flattens a passive company can lift a thoughtful one, and the difference comes down to what you point it at.
When we stopped aiming AI at raw output and pointed it at research, planning, and validation instead, the whole picture changed for us. Our juniors grew more confident to produce and experiment. Seniors started driving solutions and writing code at a level we would normally expect from their managers. And the managers, freed from the weeds, put their time into validating the work and multiplying it across the team. Work that would have been shelved as too hard, or written off as impossible, started getting done in a fraction of the time. The goal was always to extend the judgment already in the room and give good people more range.
That experience is what gave us the confidence to build something of our own, shaped by two decades in this industry and every lesson we gathered along the way. We have never believed in one-size-fits-all. So we built a strong, principled base and then shaped the system, the language, the tooling, and the workspace around the realities of the people who would use it. It moves more deliberately by design, keeps people firmly in charge of the decisions, and treats careful research and validation as more valuable than raw speed.
Now we are extending that work to a major utility partner and tuning it to fit their world specifically. They understood that generic AI off the web was never going to survive contact with the complexity of their operations, protect the integrity of what they owe regulators and the public, or keep their data and their intellectual property inside their own walls where it belongs. And we are approaching it the right way, with a careful trial first, then a small group, then a wider one, and with training and change management treated as the main event rather than an afterthought.
So here is the question worth carrying into Monday. Are you using AI to amplify the people and the trust that make your organization worth choosing, or are you slowly trading them away for a volume of work you will come to regret? The companies that come out ahead over the next few years will be the ones that stayed unmistakably human while everyone else blurred together. We intend to help the utility world be among them.
