Middle managers cannot wait this one out

In a company of 50 or 60 people, change usually gets stuck in the middle.

The owner sees the shift early. The frontline people are often willing to try a new tool if it saves them two hours. Then the idea reaches the managers, and suddenly everything becomes “risk,” “legacy process,” “not the right timing,” or “let’s wait for leadership to decide.”

None of those phrases sound lazy. That is why they are dangerous. They sound responsible. They sound careful. But if every decision has to float back to the founder, the company is not being careful. It is running on a hidden brake.

AI makes that brake more expensive.

The middle-management job has changed

For years, many middle managers were senior operators with a title. The boss set direction, managers divided work, checked progress, collected problems, and reported upward. In a stable business, that can work. Experience still matters. Process still matters. The old playbook does not fail overnight.

AI changes the rhythm. Stanford HAI’s 2025 AI Index says 78% of organizations reported using AI in 2024, up from 55% a year earlier. Microsoft’s 2025 Work Trend Index, based on 31,000 workers across 31 countries, found that 81% of leaders expect agents to be moderately or extensively integrated into their company’s AI strategy in the next 12 to 18 months.

That is not a software upgrade. It changes how work is scoped, checked, handed off, and priced. It also changes what a manager is paid for.

Old value: keep people busy and keep mistakes low.

New value: get the result back faster, with a team that learns as it works.

Do not treat every slow manager the same

When a company starts pushing AI into daily work, slow managers usually fall into three groups. Mixing them together is a mistake.

The first group has the right attitude but weak methods. They are not blocking on purpose. They simply do not know how to use AI in sales, support, finance, content, recruiting, or product work. Their old judgment is not useless, but it is no longer enough. These people need examples, a short training loop, and clear deadlines.

The second group has enough ability but protects itself by default. They always bring risks upward and leave the choice to someone else. They say “this may go wrong” but never add “here are two ways to handle it.” These managers need tighter rules. Every risk memo should include options, cost, owner, and a latest decision time.

The third group should probably not remain in management. They do not learn, do not test, do not take responsibility, and quietly punish the people under them who want to move faster. Give them one clear correction window if you must. Do not give them endless extensions. One stuck manager can slow down 10 good employees.

Stop rewarding caution without ownership

Many managers are not naturally conservative. The company trained them that way.

If the safest person is the one who never decides, people learn not to decide. If the person who lists every possible problem gets praised for being thorough, people learn to list problems. If the boss always steps in when a decision feels uncomfortable, people learn to wait.

In an AI-era company, that rule set has to change.

A manager can raise a risk, but the risk must come with a proposed move. A manager can run an experiment, but the experiment needs a boundary. A manager can say the team lacks resources, but they must say exactly what is missing and what result the resource buys. A manager can ask the founder to choose direction, but not to make every operating call.

Use a simple reporting rule. Any “we cannot do this” update must contain four lines:

• What is the problem?
• What happens if we do nothing?
• What do I recommend we try first?
• What decision do I need from leadership?

This rule exposes the difference between a manager who is thinking and a manager who is hiding.

AI is not the manager’s replacement. It is the manager’s mirror

Managers are right to feel some pressure. If your job is forwarding messages, chasing updates, cleaning spreadsheets, and turning other people’s work into weekly reports, AI will make the job look thin. It can draft the memo. It can summarize the call. It can scan customer tickets. It can produce a first version of the plan.

But strong managers become more valuable, not less.

Microsoft uses the term “agent boss” in its report. Strip away the language and the idea is simple: managers will increasingly manage a mix of people, AI tools, and automated workflows. A sales lead will not only ask whether a customer was contacted. They will have AI pull the customer history, competitor signals, objections, and next-message draft, then ask the salesperson for judgment. A content lead will not wait for a blank-page draft. They will use AI to build topic options, test headlines, check facts, and free editors to make better calls.

AI widens the gap. Good managers remove low-value coordination. Weak managers create more dashboards, more group chats, and more fake activity.

Managers do not need to become AI experts first

A small company does not need every department head to study model architecture. That can come later, if it ever matters. The practical work starts with five habits.

Use the tools yourself

Do not tell your team to use AI while you stay above it. Pick one real task every day: customer analysis, competitor notes, meeting minutes, recruiting copy, a postmortem draft, a finance anomaly, a code review checklist. Real work only. Demo tasks create demo confidence.

Define the result, not just the activity

“Improve this campaign plan” is a weak instruction for both humans and AI. A better manager names the audience, budget, format, quality bar, deadline, and boundaries. If the manager cannot describe the output, the team will waste time guessing.

Run small pilots

Do not announce a transformation program. Pick one painful workflow and run a two-week test. Customer support can test AI ticket classification and draft replies. Marketing can test first-pass scripts and headline variants. Product can test requirement splitting and edge-case checks. Measure three things: time saved, quality loss if any, and whether people feel less buried.

Keep reviews short

A useful AI review does not need a 20-page document. Ask four questions once a week: where did we save time, where did the tool fail, was the failure caused by people, tool limits, or unclear process, and what rule changes next week?

Redesign work when repetitive tasks disappear

If AI takes over a chunk of repetitive work, do not leave people sitting in the same old boxes. Move them toward judgment work, customer work, quality control, process design, or tool ownership. A manager who cannot redesign work will defend yesterday’s org chart until it becomes a cost problem.

Founders have to stop sending mixed signals

Middle managers often avoid decisions because the founder taught them to. The founder says “try things,” but when something goes wrong, asks, “Who approved this?” After a few rounds, everyone understands the real rule: experimentation is a slogan, blame is real.

The fix is not to accept chaos. The fix is to set boundaries before the experiment starts.

What budget can a manager spend without approval? What customer-facing changes require review? What data cannot be put into external tools? What kinds of mistakes are acceptable inside a pilot? What must be escalated immediately?

Once the boundaries are clear, accountability becomes fair. Failing to test can be judged. Reckless testing can be judged. The worst behavior is pretending to be careful while the opportunity passes.

A rule set a small company can use tomorrow

• Every department runs one real AI workflow experiment each week. Small is fine. Fake is not.
• Every problem report includes a proposed solution. A risk without an option is only half the work.
• Any decision stuck for more than 48 hours must list the blocker, owner, and final decision time.
• Every manager creates a monthly “automation list”: repetitive work AI can take, and human time that should move to higher-value work.
• Every pilot is judged by facts: time saved, rework reduced, quality held, customer or internal recipient not harmed.

None of this is sophisticated. That is the point. Small companies do not fail because they lack slogans. They fail because slogans never become operating rules.

Judge managers by how their work changes

Do not judge a manager by how many AI articles they forward in the company chat. Do not judge them by whether they attended the training. Do not even judge them by whether they sound excited.

Look at the work.

Is the team producing small results faster? When a problem appears, does the manager bring options or just trouble? Have people been pulled out of repetitive work and moved toward judgment, customers, quality, or process improvement?

If the answer is yes, that manager can grow in the AI era even without deep technical knowledge. If the answer is no, seniority will not save them. They are guarding an old process, not managing a new business.

The habit middle managers need to drop is not pride. It is the habit of waiting for the founder to absorb every hard call. Leadership can give direction, boundaries, and resources. It cannot walk every step for the middle layer. In the end, a manager is paid for outcomes, not explanations.

Sources

• Stanford HAI, The 2025 AI Index Report
• Microsoft WorkLab, 2025 Work Trend Index: The year the Frontier Firm is born
• Gallup, State of the Global Workplace
• World Economic Forum, The Future of Jobs Report 2025