The **Mid-level** now does the work of three. On the salary of one
TrendsGenerative AIAnalysis

The Mid-level now does the work of three. On the salary of one

AI narrowed the gap between Junior and Mid-level exactly where the Mid-level had its edge: executing fast and well. What is happening to the middle level and what has to change to avoid getting stuck in the middle.

Macareno6 min read

Junior → Senior in the Age of AI series · Chapter 3 of 6

The previous chapter ended with an uncomfortable question: if the new Junior arrives with AI and delivers something close to what a Mid-level used to deliver, what is left for the Mid-level?

Picture someone with four years at the company. They know the systems, solve problems without needing explanations and are the person everyone messages when something gets stuck. One morning they review the report put together by the Junior who joined two months ago and realize it is good. Not perfect, but good. Roughly what they would have delivered, in half the time.

The edge AI narrowed

One of the most cited studies on AI at work, by Brynjolfsson, Li and Raymond, followed more than 5,000 customer support agents who started using an AI assistant. Average productivity rose 14%. But the gains were very uneven: the least experienced improved 34%, while the impact on the most experienced was minimal.

The authors explain it this way: AI captures how the best people work and hands it to those who are just starting. For the company, that is great news. For the person in the middle, less so, because their currency was exactly that: doing fast and well what the Junior did not yet know how to do. AI shortened their edge from below.

The sandwich

The Mid-level is squeezed from both sides.

From below, they take on the work of the Juniors the company no longer hires. A study by IESE Business School, analyzing data on 138 million workers in AI-exposed sectors in the United States, found that firms cut Junior positions by 4% and increased mid-level positions by the same proportion, but with lower starting salaries for those roles. There are more seats in the middle, and they pay less.

From above, they take on the work of the managers who leave. In 2024 Gartner projected that, through the end of 2026, one in five organizations would use AI to flatten their structure and eliminate more than half of their middle management positions. When a layer disappears, someone inherits its decisions, and it is almost always whoever sits right below.

The result is someone who reviews what AI produces, corrects the Junior, covers for the boss who is no longer there and, on top of that, delivers their own work. With the same title and, often, the same salary.

Job openings have already picked a side

Job postings tell the same story. According to Indeed data, the share of openings for people with two to four years of experience fell from 46% to 40% between mid-2022 and mid-2025. Over the same period, those asking for at least five years rose from 37% to 42%.

The market is asking for someone who already knows how to decide alone. The Mid-level who still needs to be told what to do, even if they do it very well, sits in the zone where AI competes the hardest.

Executing is no longer enough

The Mid-level who executes The Mid-level who grows
Owns tasks Owns an outcome
Value comes from speed Value comes from judgment
Knows a bit of everything Masters a topic that matters to the business
Uses AI to go faster Uses AI to do what was not possible before
Reviews the Junior's work Trains the Junior while reviewing

In practice, what separates one from the other:

  • Owning a complete outcome. A process, a client or a system that is "theirs" end to end. If something breaks, they are the first to know why.
  • Specializing in something AI cannot solve alone. Knowledge of the business, the internal rules, why things were done the way they were. That is not in any model.
  • Being the quality filter. With more AI-generated content, someone has to decide what ships and what does not. That someone gains weight.
  • Training while reviewing. The Mid-level who explains why they corrected something is doing the Senior's job before having the title, and it shows.
  • Translating between business and technology. AI answers well-formed questions well. Whoever can turn a messy business problem into a good question is worth more than whoever only executes the answer.

Promotion has a different exam now

The step up to Senior used to be earned by accumulation: years, projects, flight hours. Today AI compresses the flight hours, so the exam is different. What gets evaluated is whether you can make decisions with incomplete information, own the consequences and help the people behind you grow.

The middle level is not disappearing, but its role is changing. Whoever keeps measuring their value by speed competes head to head with AI. Whoever measures it by judgment gets to use it.

If you are at that point in your career, or you lead a team full of overloaded Mid-levels and are not sure how to reorganize it, send me a message and let's talk.

Next chapter

If the Mid-level absorbs the work from below and from above, what is left for the Senior? And if nobody trains Juniors, where do the next Seniors come from? In chapter 4 I look at the level everyone takes for granted, which has its own crisis.

The full series

  1. If nobody hires Juniors, where will the Seniors come from?
  2. How does a Junior gain experience if AI does their job?
  3. The Mid-level now does the work of three. On the salary of one (you are here)
  4. The Senior with AI thinks they are faster. The data says otherwise
  5. The four faces of resistance to AI
  6. In today's job interview everyone has AI. The winner is whoever can explain what they did

Sources

Share article

Next business step

Connect this article with a relevant service and a real MacarenoNet case to move from insight to execution.

Recommended service

Digital Transformation Services

Consulting, implementation and continuous improvement on Microsoft stack.

View service

Recommended case

Project Dashboard

Real-world case with measurable execution and visibility gains.

View case