MMOOZZGG.com

Why AI does not work

Employee resistance to AI: why staff do not use the AI the company bought

5 min read2 sources

A company buys AI, and a month later it turns out staff do not use it and work the old way. Employee resistance to AI is rarely about the technology: BCG estimates that about 70% of AI adoption difficulties come from people and processes, 20% from technology and 10% from algorithms. Below are three reasons people avoid a new tool, and what works in practice.

On this page · 4 sections

Fear of replacement

The first thing an employee hears about new AI is “it will replace people”. Using the tool then means helping to end your own job. People find reasons not to start: it is awkward, it makes mistakes, there is no time.

Silence from management makes it worse. If nobody talks about AI plans, staff fill the gap themselves, usually with the worst version.

Fear goes away through an honest talk with numbers, not through promises: which tasks go to AI, which stay with people and how the assessment of work will change.

Usually AI takes the flow of routine requests, and people keep deals, hard cases and customers with unusual terms. Where staff are scarce, an agent more often fills a vacancy that has been open for months than takes anyone’s place.

A simple rule works well: in the first weeks, the agent takes nobody’s whole job. It removes the routine nobody likes: repeats, retyping, answering the same thing.

How to talk about this with the team and how to split responsibility between people and agents is covered in the piece on managing a team with AI.

No training on their own tasks

A neural network course teaches you to write chat prompts on someone else’s examples. A manager comes back to work and does not see how it relates to their requests, contracts and CRM.

Train people on their own processes: what the agent does, what it hands to a person, how to accept its work and which number shows that everything works.

Department heads learn first. If a head does not understand what the agent does and how to accept its work, they will not explain it to the team and will not demand results.

The trainer should be someone who knows your processes and answers for the result. That is how team training by an AI director works: people work through their own tasks, not someone else’s.

Without training, there are two extremes. An employee redoes the agent’s work from scratch or trusts it without checking, and in both cases the effect is lost.

The training check is simple: after a week, the employee shows at their own desk which task they gave to AI and how much time they saved. If they cannot show it, the training did not work.

No time for something new

The employee has the same plan, the same requests and the same reports. A new tool takes time to learn, and nobody set that time aside. AI becomes a task for “when there is time”.

Microsoft Work Trend Index 2026: organizational factors affect the value of AI twice as strongly as personal ones, 67% against 32%. The organization, not the employee’s wish, sets the time to learn, the tasks and the assessment of work.

The fix is organizational: during the launch, take part of the load off people and say plainly how much time a day goes to reviewing the agent’s work. In our pilot, the responsible employee spends about half an hour a day on this.

Time to learn pays back only if it was planned. The department head decides in advance who reviews the agent’s work and when, and which tasks are taken off for those days.

Another trap: AI is added on top of the old process instead of built into it. The employee does the work the old way and also checks the AI, so there is more work, not less.

What works in practice

1. One process, not everything at once. The team sees the result in one area, and the next launch goes more easily.

2. An owner for every AI tool. This person sorts out disputed cases, edits the rules and answers for the number.

3. A metric everyone can see: how many requests AI closed, how many it passed to people, and what people did with them.

4. Training on your own tasks in the first days of the launch, not a general course a month before it.

5. A new assessment of work. If a manager is rated by the number of requests handled and the agent took the requests, the manager will resist. Move the assessment to deals and hard cases.

6. An example from the top. If the CEO uses AI and asks departments for numbers, staff see AI as part of the job, not a passing idea.

Employee resistance to AI more often says something about how the rollout is set up than about the people. You remove it through organization: a goal, an owner, time and training.

When people see less routine and a fair assessment, they use AI every day. That is why training staff to work with agents is part of building an AI department, not a separate line after launch.

Sources and fact-checking

Checked by the editors on 13 October 2026.

  1. BCG: 74% of companies get no tangible value from AI, 24 October 2024 (bcg.com)
  2. Microsoft Work Trend Index 2026, 5 May 2026 (blogs.microsoft.com)

Found a mistake? Tell us

Publications

Mike Mozg

Agentic AI use cases at AT&T: nearly 500 processes now run on agents

AT&T no longer uses AI only to help staff work faster: it is rebuilding whole business processes around agents. SiliconANGLE reported on 14 November 2026 that nearly 500 agentic AI use cases at AT&T are already in production. Here is how the company rebuilds them, where a human still checks the agent, and what other large companies can take from it.

News4 min
Contact us