A new client engagement kicks off tomorrow. Tonight I sat down to prep the opening of the first session, and I realized I write some version of these same notes every time. Different client, different industry, different stack of tools they've already bought. The opening doesn't really change.
This piece is that opening. We've been running engagements like this at Gadoci for a while now, and three observations keep showing up at the start of every one. They're the things I want a client team to hear before we get into the actual work, because the engagement goes better when the room shares the framing.
If you're about to kick off your own AI adoption or transformation, these are the notes I'd give you too.
The work is mostly about people
Most of the conversations I have at the start of a new engagement assume the hard part is the technology. It isn't. The technology piece is real, but it's the smaller share. BCG has been publishing the same number for a few years now: roughly seventy percent of the value of an AI program comes from the people change, not the algorithm. That number lines up with what we see in practice.
Everyone has access to the same frontier models. ChatGPT, Claude, Copilot, Gemini, Perplexity. The licensing piece is rarely the bottleneck once a real budget is on the table. The advantage is not the model. The advantage is the ecosystem you build around the model, and the team you build to use it.
The pilots that don't reach production almost never fail for technology reasons. They fail because nobody owned the workflow redesign, or the change management, or the new way the team had to coordinate around the tool. The gap from impressive demo to "this runs every Tuesday morning" is human work. It's where most of the budget should go, and where most of the budget rarely does.
Business leaders own the program. Not IT. We've watched enough transformations succeed and fail to know the pattern. The ones that work put the VP, the COO, or the CFO in the driver's seat, with technology in support. The ones that stall put IT in the lead and treat the rest of the business as a customer of the project. AI lives in the work itself. It belongs to the people doing the work.
Fear and old habits slow adoption more than any tooling gap. Job-security worry, "this is how we've always done it," and the friction of changing a routine that already produces an acceptable outcome. None of that gets solved by a better model. It gets solved by giving people room to try the tool on real work, in their own way, with someone next to them they trust.
The unlock, when it comes, is your people. Not a smarter model. The work gets redesigned by the people who already know it, and that's where the real productivity shows up.
The second note is about pent-up innovation
For most of the last twenty years, the great ideas inside a company came from somewhere that wasn't really designed to produce them. The Post-it Note happened because Art Fry was singing in a church choir and wanted a bookmark that wouldn't fall out of his hymnal. The Frappuccino happened because two Starbucks store managers in Los Angeles built it themselves, on a hot afternoon, against the wishes of corporate. Gmail happened because Paul Buchheit had twenty percent of his time to do whatever he wanted with it.
These stories get told as proof that great ideas come from the edges of the organization. They are. The more honest version is that for decades, ideas from the edges were the exception. Most people inside companies had useful ideas. Most of those ideas died on the vine. There were too few resources, too little attention from leadership, too high a bar to get the time to build anything, and no real way for a frontline operator to do the work themselves. The bottleneck wasn't ideas. The bottleneck was the cost and friction of acting on them.
AI changed that. What used to take a small team a quarter now takes one fluent operator an evening. The tools are good enough that someone with a real understanding of the work can build the thing themselves, without filing a request, joining a queue, or convincing a roadmap. They can prototype, test, refine, and ship.
The implication is bigger than any single tool. We're about to live through a release of pent-up innovation that has been building inside companies for years. Every operator who has ever said "I wish we could just" is about to find out they actually can. The leaders who recognize this early, and design their organizations to channel that energy into the actual work, will pull away from the ones who don't.
The third note is about how the work itself is changing
The longest-running observation in our practice is that the people who get value from AI work differently at the keyboard than the people who don't. Two years in, a thousand sessions in, that pattern hasn't moved. The line between "AI works for me" and "AI doesn't work for me" is almost never about the model. It's about the behavior.
The people who get value treat the tool like a capable colleague. They open thin, share context, react to what comes back, redirect when the model drifts, and stay in the conversation until the answer converges. The people who don't get value type a half-sentence, scan the output, and decide AI isn't there yet. They're using a search-engine reflex that twenty-plus years of Google trained into all of us, and that reflex doesn't translate. The modern AI tools are not search engines wearing a costume.
We've written about this at length in Become a Conductor of Knowledge and Context. The short version is that the skill is no longer the prompt. The skill is orchestration. A conductor plays no instrument. She brings the score, decides what gets played, watches the musicians, and reads the room. The model is the orchestra. Your people are the conductors. The output sounds like one piece of music because someone in front of the orchestra was making sure the right inputs arrived at the right moment.
The behavioral piece runs underneath everything else we do in an engagement. New tooling helps. New skills, custom GPTs, projects, MCP connectors, all of it helps. The line that actually separates the value-getters from the rest is whether someone has made the move from prompter to conductor, and whether they're willing to keep practicing it.
Why we're here
The reason a Gadoci team shows up at the start of an engagement is not because we know your business better than you do. We don't. You are the expert on your workflows, your customers, your data, and your priorities. Nobody is going to overtake you on that.
We sit on the other side of that. We spend our days inside the world of AI. The models, the tools, the features, the news, the patterns we keep seeing across clients. Our job is to translate between the two, and to stay close to the people doing the work until the AI piece becomes part of how the work runs. The engagement is most useful when the room treats us that way. Not as a vendor selling you software. Not as the people who will replace your judgment. As the translation layer that helps the people in your building turn what's possible with AI into how the work actually gets done.
A standard for the kickoff
A few notes, then. The work is about people. The unlock is the pent-up innovation already inside your organization. The new shape of knowledge work is conducting, not typing. We're here to translate, not to take over.
That's what I want every room to hear at the start of an engagement, including the one we're starting tomorrow. We've been doing this for a while now. None of it is theoretical. All of it shows up in the first sixty days of every engagement, in every industry, with every stack of tools the client has already bought.
If you're about to kick off your own program, take these as the standard. The technology will keep moving. The room of people who actually have to make it work won't. Get the framing right at the start and the rest of the engagement gets easier.