Every CEO I meet this year wants to talk about "the algorithm."
Which model. Which vendor. Which copilot they are rolling out to the salesforce, the contact center, the clinic, the store. The slides are immaculate. The use cases are exciting. And almost none of them are talking about the actual algorithm that will decide whether any of it works.
Here is that algorithm, and it is not made of silicon. It is the human being who runs the Tuesday huddle. The shift supervisor. The team lead. The middle manager standing between your beautiful AI roadmap and the people who have to live inside it. That manager is the operating system your strategy runs on. Brilliant code on a corrupted OS still crashes. And right now, in most organizations, we are shipping the most disruptive workplace technology in a generation onto an operating system we have neither patched nor enabled.
I call this the manager-as-algorithm problem, and it is the quiet variable that separates the AI rollouts that lift engagement from the ones that hollow it out. Because whatever your tool does, your people will experience it through their manager. The manager is the translation layer. The manager is the meaning-maker. The manager is the algorithm.
The conversation that wasn't
Let me give you the data, because this is not a hunch. It is one of the better-documented findings in workplace research, and it has nothing to do with the slogans on your lobby wall.
Gallup studied employees who voluntarily quit their jobs and asked a deceptively simple question: could anything have kept you? The answer should unsettle every leader betting big on AI. A striking 42% of voluntary leavers said their departure was preventable — that their manager or organization could have done something to keep them, and didn't.
Now sit with the reason. In the three months before they walked, 45% of those leavers reported that neither a manager nor any leader proactively talked with them about their job satisfaction, their performance, or their future. Almost half left in a silence nobody broke. The exit interview is too late; the conversation that mattered never happened.
This is the most expensive thing managers are not doing — and it is about to get more expensive. Drop AI anxiety into a workforce where half of departures are already preventable and already silent, and you are pouring accelerant on a fire you can't see. People don't quit a technology. They quit the feeling that no one is steering them through it.
The good news hides in the same research. Gallup found that when a manager has just one meaningful conversation a week with each direct report, employees are four times as likely to be highly engaged — frontline, hybrid, or remote, it didn't matter. Four times. From a 15-to-30-minute conversation. That is the highest-leverage, lowest-cost engagement driver in the building, and it runs entirely on enabled managers.
Why enablement, not announcement
Most AI rollouts are an announcement. A tool appears, an email explains it, a training video is assigned, and leadership moves on, puzzled when adoption stalls and morale dips. Announcement is not enablement. Announcement informs the manager. Enablement equips the manager to translate.
The difference is everything, because managers carry a cost most strategy decks ignore. Gallup estimates that replacing a leader or manager runs about 200% of their salary. These are not interchangeable parts. They are your translation layer, and when they break — or quietly disengage themselves — the whole rollout breaks downstream of them, customer by customer.
So what does enabling the algorithm require? Three things the announcement always skips.
Context, before the tool. Managers cannot translate what they don't understand. Before the copilot lands on Monday, the manager needs the why: what business problem this solves, what it will and won't do, and — critically — the honest answer to the question every employee is silently asking, what does this mean for me? A manager who can answer that with calm candor protects more engagement than any town hall.
Permission to coach, not just deploy. We tend to hand managers a tool and a target and call it a day. But a manager's real job in an AI transition is human, not technical: notice who's anxious, name the fear, hold the weekly conversation, connect the work to something that matters. That is coaching, and coaching only happens when managers are given the time and the explicit permission to do it instead of being buried in the very dashboards AI was supposed to simplify.
Air cover to be honest. Enabled managers can say hard true things — "yes, this changes your role; here is how we'll grow you into the new one." Managers who fear for their own standing default to corporate fog, and employees smell fog instantly. Enablement includes telling your managers it is safe to be human in front of their teams.
The store that proved it
You don't have to look far for proof of what an enabled manager unlocks. Years ago, while I was inside Starbucks writing about their culture, the line I heard repeated was deceptively simple: leaders were taught to put "partners" first, trusting that engaged partners would, in turn, take care of customers. Not a poster. A sequence. Take care of the people who take care of the people.
The mechanism behind that sequence was always the store manager. Corporate could set values, design the apron, write the playbook — but whether a barista felt seen on a brutal Monday rush came down entirely to the person running that store. The store manager was the algorithm that turned a national brand promise into a felt experience at 7:14 a.m. Enable that manager and the promise comes alive. Strand that manager and the same playbook produces a tired, transactional shrug.
That is precisely the lesson to carry into AI. Your model is the brand promise. Your manager is the store. No promise survives a strand.
Three moves to enable the algorithm
1. Brief managers first — and differently. Give your managers a head start on every AI rollout, with context the rank and file won't get for weeks: the strategy, the honest talking points, the answer to "what about my job?" A manager hearing the news at the same moment as their team has nothing to translate. They become a bystander to their own rollout.
2. Protect the weekly conversation like revenue. If a 15-minute weekly check-in quadruples the odds of high engagement, treat it as the operational ritual it is. Put it on the calendar, defend it from the meeting tide, and coach your managers on what makes it land — goals, recognition, a real question about how the person is doing with all this change.
3. Measure the manager, not just the tool. Most AI dashboards track adoption and deflection. Add the human metric: are managers having the conversations? Do their people feel equipped or surveilled? The tool's usage stats will lie to you about engagement. Your managers' conversations won't.
The reframe
Here is the shift I urge every leadership team to make. Stop asking, which algorithm should we deploy? Start asking, have we enabled the algorithm we already have? — the human one, standing in front of the team, deciding in real time whether your AI feels like a gift or a threat.
The tools will keep getting better. They will not, on their own, hold a weekly conversation, name a fear, or make a frightened employee feel seen on the morning everything changed. That work belongs to managers. And in the AI era, the organizations that win engagement won't be the ones with the best model. They'll be the ones that remembered to enable the human running it.
So the question worth carrying into your next leadership meeting: you've invested fortunes in the algorithm made of code. What have you invested in the algorithm made of people?
If you're not sure your managers are equipped to translate AI into engagement instead of anxiety, that's exactly the work I do with leadership teams. Come find me at josephmichelli.com/contact.
