Generative AI makes a thousand promises, while concrete gains are still slow to materialize. Yet some players are getting more out of it than others. Anthony Carange, CEO of Carange Solutions, says he’s helped 120 municipalities adopt AI “without major resistance.” He shared 5 lessons he’s drawn from these rollouts.
1. Business Knowledge
Before hiring a team of AI professionals, a generative AI project needs to be led by people who have “business knowledge.”
“I know artificial intelligence, but I don’t know the ins and outs of municipal work in great detail,” he explains.
Those who have that business knowledge know better than anyone which processes need optimizing. So that’s the concrete starting point.
2. Documentation
Even with experts on the team, AI can only work effectively if the company has content and context to feed it. The organization needs to have at least minimally documented its policies and processes before launching a project.
Anthony Carange gives the example of a municipal manager who used an agent to prepare their three-year capital plan. They submitted an Excel file with all their data, their policy, and their equipment replacement requirements. They set a maximum investment amount and asked the AI to indicate, after its evaluation, which vehicles and equipment should be prioritized.
“The manager was surprised: the AI landed on exactly the same figures he had come to in his manual analysis,” he says.
3. Connectivity
Using AI without connectivity between an organization’s tools is like hiring an intern you have to show everything to and constantly tell what to do. “Write me a quote. Go find a price list. Write me an email. Analyze this document…”
“If there’s no connection across the platform, it gets tiring to redo that mental — or computer — wiring every single time just to guide the AI. Sure, you get a little dopamine hit when an AI agent answers your questions. But the next level is moving toward agentic AI. And that requires connection between platforms.”
4. “Champions” Who Reveal Themselves Along the Way
From experience running several AI projects, Anthony Carange notes that the internal “champions” of AI aren’t always who you’d expect.
“At the start of every training session, we ask people what their fears about AI are. And we’ve noticed that the people who are most resistant at first often end up becoming the AI champions in an organization. It’s happened so often we can’t call it a coincidence anymore,” he says.
5. ROI That’s Hard to Quantify
Despite all the enthusiasm Anthony Carange brings to his AI projects, he acknowledges one thing: the return on investment is hard to put a number on.
“You can’t measure the time spent on every task before and after. That would require way too much investment just to take that kind of measurement. Especially since AI sometimes brings a qualitative improvement to a product or service rather than a time saving. So instead of saving time, you’re improving quality.”
training.isarta.com 
