Leadership and Social Influence: The Skill Rising Fastest in the Age of Machines
Somewhere in your company this week, a machine made a recommendation, and a room full of people decided what to do about it. Perhaps the model proposed cutting safety stock at three warehouses. Perhaps it flagged a supplier as a risk, or suggested a route no dispatcher would have chosen. The recommendation arrived in seconds. What happened next took days, and it had nothing to do with computation. Someone had to persuade someone else that the number could be trusted, or that it could not. Someone had to convince a skeptical planner to change a habit, a cautious finance director to accept a new kind of risk, a proud expert to admit that the machine had seen something he had missed. The technology did its work in milliseconds. The humans did theirs in conversations.
That gap, between what the machine can recommend and what the organization will actually do, is where a quiet shift in the labor market becomes visible. In January 2025, the World Economic Forum published its Future of Jobs Report, built on a survey of more than a thousand employers representing fourteen million workers. The headline findings were about technology: AI, big data, cybersecurity. But buried in the skills outlook was a stranger result. The skill that had risen fastest since the previous edition of the report was not a technical one. It was leadership and social influence, up twenty-two percentage points in the share of employers calling it a core skill, and now ranked third among all skills required of workers today, behind only analytical thinking and the trio of resilience, flexibility, and agility. Employers expect it to remain among the defining skills through 2030.
Sit with the oddity of that for a moment. At the precise point in history when machines are absorbing more of the work than ever before, the skill appreciating fastest is the most stubbornly human one we have: the ability to move other people without forcing them. This article is about why that is happening, and about the machinery underneath influence itself, which turns out to be something psychologists call agency: the belief, held or lost, that your actions can change outcomes. It is also about the questions that belief raises for anyone leading through an AI transformation. Can agency be damaged? Can there be too much of it? Should some things simply be taken for granted? And how do you decide, on an ordinary Tuesday, whether to push for change or accept the world as it is?
Why automation makes influence more valuable, not less
The intuitive story runs the other way. If machines optimize routes, forecast demand, and draft the emails, surely the human arts of persuasion matter less. The system decides; people execute. Yet every executive who has lived through an AI deployment knows the intuitive story is false, and the reason can be stated in one line: machines optimize, but only humans commit.
An algorithm can tell you the best route. It cannot make a driver take it. A forecasting model can produce a number with a narrow confidence interval. It cannot make a sales director plan against it, or stop a plant manager from keeping a private spreadsheet of the numbers he actually believes. Adoption, the unglamorous last mile of every transformation, is not a technical process. It is a social one. It happens or fails to happen in the space between people, and the force that operates in that space is influence.
Microsoft’s 2026 Work Trend Index makes the same point from the other direction. As organizations fill with AI agents that can execute work on request, the differentiating question stops being what the technology can do and becomes what humans choose to direct it to do, and whether the people around them follow. The report’s authors reach for a telling phrase: human agency. When execution is cheap, judgment and commitment become the scarce inputs, and commitment is manufactured through influence, not instruction.
There is a structural reason as well. The organizations deploying AI are also flattening. Work runs through matrices, partnerships, and cross-functional teams in which almost nobody can simply order almost anybody. The World Economic Forum’s analysts point to geoeconomic fragmentation and rapid technological change as forces pushing employers toward leaders who can manage complex social dynamics rather than administer stable hierarchies. Strip away the report language and the observation is simple: the modern leader spends most of the day influencing people over whom she has no authority at all, including, increasingly, the people who decide whether the machine’s recommendation lives or dies.
So the paradox resolves. Influence is rising in value for the same reason gold rises when currencies wobble. It is the asset that still works when the formal system, the org chart, the mandate, the steering committee, is not enough.
Three words we keep confusing
Before going further, it is worth separating three things that get collapsed into one another in everyday management talk, because confusing them is expensive.
Authority moves bodies. It is the power of the role: approve, assign, escalate, veto. Authority is real and necessary, and nothing in this article argues against it. But authority produces compliance, and compliance is thin. People do what the role-holder says while the role-holder is watching, to the letter of the instruction and not a centimeter past it. Authority is also borrowed. It belongs to the position, not the person, and it is repossessed the day the title changes.
Influence moves minds. It produces commitment rather than compliance: people who keep doing the thing when nobody is watching, who improve it without being asked, who defend it to their own teams. Influence travels through trust and credibility rather than through reporting lines, which means it works sideways and upward, not just down. And unlike authority, it is owned rather than borrowed. It follows the person across roles, companies, and careers.
Manipulation is the counterfeit. It uses the same psychological pathways as influence, which is what makes it tempting, but it borrows trust without repaying it. The manipulator wins the decision and loses the relationship, and in an organization, where every negotiation is a repeated game, that trade is ruinous. Robert Cialdini, whose research we will lean on later, spent his career cataloguing the levers of persuasion, and he is blunt about the boundary: the same principles used dishonestly destroy the influence they briefly create. A practical test separates the two, and it is worth writing down. Influence survives transparency. If the person you are persuading could see exactly what you are doing and why, and would still say yes, you are influencing. If your method only works while it is hidden, you are manipulating, and you will get to use it approximately once.
Leaders reach for authority because it is fast and for manipulation because it is easy. The skill the labor market is now repricing is the third thing, the slow, compounding one.
The engine underneath: agency
Here is what the how-to articles about influence usually skip. Before the first stakeholder map, before the first carefully framed proposal, influence begins with a belief, and the belief is not about the audience. It is about yourself.
The psychologist Albert Bandura spent decades studying what he called self-efficacy: a person’s belief in their own capacity to produce desired effects through their actions. His conclusion, repeated across hundreds of studies, is that this belief is the central mechanism of human agency. As he put it, unless people believe they can produce desired effects by their actions, they have little incentive to undertake activities or to persevere in the face of difficulties. The belief comes first. Effort, persistence, and recovery from setbacks all flow downstream of it.
This is why two managers with identical evidence and identical stakes behave so differently in the same meeting. One raises the concern, reframes it when the first framing fails, finds an ally by Thursday. The other, who may be more capable, and may even be more right, says nothing, because somewhere beneath articulate reasons, what were the odds anyone would listen, it was not the right meeting, a quieter verdict has already been reached: nothing I do here will change anything. Influence attempts are investments, and nobody invests out of a belief in futility.
Bandura’s research also mapped where the belief comes from, and the sources are usefully concrete. The strongest is mastery experience: succeeding at something, especially something slightly beyond what you thought you could do. Small real wins beat large hypothetical ones. The second is vicarious experience: watching someone recognizably like you succeed. Not the keynote speaker, the colleague two desks away. The third is encouragement from people whose judgment you trust, which works less by cheerleading and more by recalibrating what you consider possible. The fourth is your own physiological and emotional state: exhausted, anxious people systematically underestimate what they can affect.
It is worth pausing on how quietly this belief is formed and deformed, because almost none of it happens in performance reviews. It happens in micro-moments. A manager rewrites a subordinate’s analysis instead of sending it up under her name, and a small piece of her belief that her work can reach decision-makers dies without anyone noticing. A leader asks a junior planner what he would do before revealing what the model says, and a small piece of belief is born. Multiply either moment by a thousand working days and you get two different people, and eventually two different organizations: one that generates influence attempts, ideas, objections, and initiative from every level, and one that waits to be told. No line in the budget records this difference, but it may be the most consequential asset either organization owns.
Read that list again as a leader, because it doubles as a job description. Every item on it is something you can engineer for other people. You can carve a large change into small winnable pieces. You can make early successes visible so that peers watch peers succeed. You can lend your credibility, publicly, to someone whose confidence lags their competence. You can manage load and pace so that your team is not deciding about the future from the bottom of an energy debt. This is the two-way street at the heart of leadership and social influence, and it is why the WEF pairs the two words. Influence is not only the agency you exercise. It is the agency you grant. The leaders who compound influence over decades are rarely the most persuasive people in the room. They are the ones who leave other people more capable of acting than they found them, and who are then owed, trusted, and followed accordingly.
The risks to agency
If agency is the engine, it can also be damaged, and AI-heavy workplaces are discovering that it can be damaged from two opposite directions at once.
The first direction is erosion. When a system speaks with mathematical confidence, silence becomes the path of least resistance. The planner who once argued with the forecast learns that the forecast no longer argues back; it simply gets regenerated. Over months, a subtle lesson is absorbed: my input is not a variable in this equation. Researchers publishing in Frontiers in Psychology in 2026 found that the intensity of AI application in a workplace has a curved relationship with employees’ sense of insecurity, and that two factors reliably soften it: employees’ own self-efficacy, and leadership that actively empowers and recognizes people. The finding matches what anyone who walked a warehouse floor during a rollout already suspects. The technology does not have to threaten anyone’s job to threaten their agency. It only has to make them feel like spectators of their own work. And people who feel like spectators stop volunteering judgment, which is precisely the input the machine cannot supply. In our April article on deep listening, we described employees working around systems they did not trust rather than challenging them openly. That is what eroded agency looks like from the outside: not rebellion, just quiet exit from the decision loop.
The second direction is less discussed, and leaders who have absorbed years of challenge-everything advice may resist it: agency can also run to excess. Call it hyper-agency, the reflex to contest everything, re-litigate settled decisions, and treat every system output as an invitation to a debate. A team of twelve people who each feel entitled and obliged to push on every recommendation does not produce twelve times the judgment. It produces change fatigue, decision queues, and the slow death of the automation dividend, because a system that must win an argument before every execution is slower than no system at all. Hyper-agency also bankrupts the individual. The colleague who objects to everything is soon discounted on everything; his pushes stop carrying information. Influence, as we will see, is a budget, and hyper-agency is the habit of spending it on everything, which is indistinguishable from having none.
So the leadership problem is not maximizing agency, and it is not protecting people from the machine. It is allocation: keeping alive the belief that pushing is possible, while building the judgment about where pushing is worth it. Which raises a question most leadership writing avoids asking directly.
Should some things just be taken for granted?
Yes. Deliberately, and more than the current fashion admits.
Trust is an efficiency technology, arguably the original one. Organizations function because most things are not re-litigated: the payroll run, the safety protocol, the accounting standard, the thousand routines that would swallow every hour of every day if each had to be argued from first principles each morning. Taking things for granted is not a failure of critical thinking. It is what critical thinking is for: deciding, consciously, where scrutiny gets spent, because scrutiny is the scarcest resource a leader has. A useful way to see an AI system is as a new entrant asking to join that category of settled things, and the honest answer is that it should join, eventually, for most of its outputs, most of the time. That is the entire point of having it.
The Stoic philosopher Epictetus, writing from considerably less comfortable circumstances than a supply chain review, put the underlying discipline in one sentence: some things are up to us, and some things are not. He was not preaching passivity. He was preaching aim. Anguish and wasted effort come from pushing on what you cannot move and neglecting what you can, and the craft of a serene, effective life is learning to tell the difference. Transplanted into a modern operating review, his sentence becomes a portfolio rule. Your agency is finite. Spend it inside the overlap of two circles: what genuinely matters, and what you can actually affect. Everything outside that overlap should be accepted, and accepted well, without the low-grade resentment that pretends to be engagement.
One discipline keeps deliberate acceptance from decaying into blind faith, and it costs a sentence. When you decide to take something for granted, name what would make you look again. We accept the routing engine’s plans, and we will revisit if override rates cross five percent or two hubs flag the same pattern in a month. Acceptance with a tripwire is a decision. Acceptance without one is merely a nap.
Push or accept: three questions
Which still leaves the ordinary Tuesday problem. A specific thing in front of you seems wrong: a policy, a plan, a number from the model. Push, or let it go? Three questions, asked in order, do most of the work.
Does it matter enough? Not is it annoying, not is it imperfect, but does it move something you would defend in a year: safety, customers, money, trust, the health of the team. Most irritations fail this test, and letting them go is not weakness. It is what funds the pushes that pass.
Can I actually affect it? This is Epictetus’s question wearing work clothes, and it is asked about standing as much as about the issue. Do you have the evidence, the relationships, the credibility with these particular people on this particular topic? If the cause passes the first question but you honestly fail this one, the answer is not surrender, and it is not a doomed frontal assault that spends credibility for nothing. It is to change the answer: gather the data, run the small pilot, recruit the ally who does have standing. Agency includes the patient work of building the position you will eventually push from.
Is now the time? Timing is the question of the credibility budget. Every push draws on an account; wins refill it, and losses drain it faster. Cialdini’s principle of commitment and consistency has a sharp edge here: people build stable pictures of one another, and the picture attached to a serial objector is he always says that, at which point his objections stop carrying information at all. The person who pushes selectively is not more timid than the person who pushes constantly. She is legible. When she says this one is different, the room believes her, precisely because of all the battles she declined.
Watch the filter work on a live example. A planning director believes the new demand model systematically underestimates promotional spikes. Does it matter? Yes: stockouts during promotions cost real revenue and real customer trust. Can she affect it? Partly. She has the data science team’s respect but no mandate over the model. So she does not write the angry email to the steering committee. She runs a quiet four-week comparison on two product families, model versus her adjustment, and recruits the one data scientist who built the promotions module. Is now the time? Not yet; the comparison is the timing. When she finally raises it, she arrives with evidence, an ally from inside the model’s own team, and a track record of not crying wolf. The push takes one meeting. Observers will call it luck, or good timing. It is neither. It is agency, spent like money by someone who knows what things cost.
Three yeses: push, and push properly, with allies and evidence and stamina. Any no: accept for now, set the tripwire, and reinvest the energy. Answering the questions honestly, rather than letting mood or fear answer them for you, is itself an act of agency, and it is the difference between acceptance as a strategy and acceptance as a slow habit of defeat.
TransNord, six months on: Markus and the override that nobody used
Readers of our June article will remember TransNord Logistics: the routing engine that went dark at ten o’clock one night, the seven a.m. crisis meeting, and the cast around chief executive Eva Lund’s table, including Lars, the operations chief, and Markus, the architect who built the system. When we left them, the outage had been repaired and the harder truth was sinking in: the company now knew what it depended on.
Markus took that truth personally, and to his credit, he built something. He designed a human override protocol: a structured way for any planner to challenge the routing engine’s plan, log the reasoning, and feed the outcome back into the model. It answered the outage’s deepest lesson, that the system was starved of exactly the human judgment it was replacing. The executive team approved it in one meeting. Eva sent a company-wide note calling it a priority.
Then Markus did what confident experts do. He rolled it out to everyone at once. All four hundred planners across every hub: webinars, e-learning modules, a polished dashboard, three reminder emails with attendance tracking. He was thorough, he was right, and the protocol was good. Six weeks later, adoption was close to zero. A handful of overrides, most of them test entries. In meetings, everyone agreed the protocol was important. On the floor, nobody touched it. Authority had produced its usual crop: agreement in words, nothing in behavior. And the failure went somewhere deeper than Markus’s project plan. The man who had rebuilt his standing after the outage by building something excellent had discovered, again, that being right moved nobody. People who watched him in those weeks saw the older pattern returning, the defensiveness, the longer and more technical explanations, the quiet conclusion taking root that the planners simply could not be reached. His problem was no longer the protocol. It was his own draining belief that anything he did would matter.
Eva’s intervention was one question, asked without ceremony. You are trying to convince four hundred people you have never met, she said. Who are the five people the others actually watch?
Markus did not know, which was the honest beginning. He went to the hubs to find out, and what he found rearranged his assumptions. The people the planners watched were not the most senior planners and not the loudest. In every hub, the answer pointed the same direction: the shift managers, the people who ran the morning stand-up, assigned the loads, and translated every corporate initiative into we are actually doing this or ignore it, it will pass. The shift managers were something the org chart did not have a word for. They held authority, modest but real, and they held influence, enormous and invisible. Whatever they adopted, their teams adopted.
So Markus started over, five people at a time, and without naming it, he began working through the levers that Cialdini’s research describes. He sat with three shift managers in one hub and asked what the override interface got wrong, then fixed their top complaint before asking them for anything at all. Reciprocity: he gave first, and the gift was labor, not lunch. He asked each manager for something small and public, five overrides a week, mentioned in their own stand-ups, rather than a grand commitment to transformation. Commitment and consistency: small, voluntary, visible promises grow roots. Their teams watched people exactly like themselves use the protocol and benefit, which no webinar could have simulated. Social proof, the genuine kind. The managers’ modest formal authority legitimized the time spent; nobody wondered whether overriding the engine was allowed, because the person who assigned the loads was doing it. And when other hubs asked what was happening, Markus recruited one champion per hub, chosen not for seniority but for being trusted and liked, the person others had coffee with. Liking, and something Cialdini added to his list late in his career: unity, the sense that the message comes from one of us.
Adoption did not jump. It compounded, hub by hub, the way trust does. By the fourth month the overrides were numerous enough, and well-reasoned enough, that the model itself improved, which handed Markus the argument he had tried to make with slides: the system is better because the humans push back. Planners who had spent a year feeling like spectators discovered that their judgment was a design input, and Sofia, watching the people as always, reported the change no dashboard measured: people had stopped talking about the system as something that happened to them.
Two mechanisms in this story are worth pulling into the light. The first is that Markus’s turnaround began with his own agency, not his audience’s. The first win with three managers was small, almost trivially so, but it was real, and it did for him exactly what Bandura’s research says mastery experiences do: it rebuilt the belief that his actions produced effects, and every larger move grew from that. The second is the shift managers themselves. The standard advice says influence the influencers. The sharper version TransNord stumbled into is that the most powerful key users are the people in whom influence and authority are woven together, because they can do the one thing neither pure authority nor pure influence can: make adoption simultaneously legitimate and desirable. Executives command but are not watched at ground level. Peers are watched but cannot legitimize. The first-line manager does both, which is why transformations are won or lost there, and why the twenty-two-point rise in the WEF’s tables is, in practice, a repricing of exactly that layer of leadership.
The toolkit: five practices
Theory earns its keep in behavior, so here is the article compressed into five practices, each small enough to start this week.
Bank credibility before you need it. Influence is withdrawn from an account that must be funded in advance, and the deposits are unglamorous: kept promises, accurate claims, help given before it is asked for, credit passed along. Markus’s repair of the interface before requesting anything was a deposit. Make one this week with the person you will need in six months.
Listen before you frame. Every failed influence attempt contains a diagnosis that was skipped. The objection you prepare for is rarely the one that matters, and the one that matters is usually offered freely to anyone who asks and then stays quiet. This is the craft of deep listening we wrote about in April, redeployed as the first move of persuasion rather than a separate virtue.
Shrink the ask. Grand commitments produce nodding; small public ones produce behavior. Five overrides a week, said aloud in a stand-up, outperforms a signed-off transformation roadmap. Design your next request so that saying yes costs less than debating.
Recruit the watched, not the senior. Before your next rollout, answer Eva’s question with names: who are the five people the others actually watch? If the answer includes first-line managers, the people in whom authority and influence are already woven, start there and let adoption travel outward on its own.
Name the trade-offs yourself. Influence survives transparency, so volunteer the costs before your audience discovers them: what the proposal risks, who bears the transition pain, what would prove it wrong. This is the honest judgment we described in our analytical judgment article, and it doubles as the strongest differentiator from every manipulator your audience has already learned to distrust.
The infinite game, again
The June article ended with James Carse’s distinction between finite games, played to win and end, and infinite games, played to keep playing. Influence belongs entirely to the second kind. There is no meeting after which you have permanently won trust, no quarter in which credibility is banked forever; there is only the long, compounding game of being worth believing, and of leaving the people around you with more agency than they had before they worked with you. That is also why influence is the asset the machines appreciate rather than depreciate. Everything that can be reduced to computation is being repriced downward. What remains scarce is what happens between people: the planner deciding to trust, the manager deciding to push, the team deciding to follow. The World Economic Forum’s tables are simply the market noticing.
The question the numbers cannot answer is the one in front of you on Tuesday: which battle deserves your push, whose agency you will build this quarter, and what you are, deliberately and with a tripwire, willing to take for granted.
This is the work of Session 5: Influence, the next session of the NorthStar Skills Academy Open Program, on 22 July at 2.00pm CET, 8.00pm Singapore time. We will put these ideas under pressure with leaders from across industries, inside a scenario where being right is not enough. Join us at northstarskills.com/sign-up.
References
World Economic Forum (2025). The Future of Jobs Report 2025, Chapter 3: Skills Outlook. weforum.org.
Bandura, A. (1982). Self-Efficacy Mechanism in Human Agency. American Psychologist, 37(2). See also the American Psychological Association’s overview, Self-Efficacy: The Theory at the Heart of Human Agency. apa.org.
Cialdini, R. B. (2021). Influence, New and Expanded: The Psychology of Persuasion. Harper Business.
Microsoft WorkLab (2026). 2026 Work Trend Index: Agents, Human Agency, and Opportunity. microsoft.com/worklab.
Frontiers in Psychology (2026). The Impact of Artificial Intelligence Application on Employees’ Job Insecurity: The Moderating Roles of Self-Efficacy and Transformational Leadership. frontiersin.org.
DDI (2026). Leadership Trends 2026. ddi.com.
Epictetus. Enchiridion.
Carse, J. P. (1986). Finite and Infinite Games. Free Press.
