# AI and Jobs: Which Tasks Will Change

Author: Marin Niehues, Niehues Consulting · Published: 2025-09-08 · Updated: 2026-11-16 · AI Strategy & Business Case
Source: https://www.niehuesconsulting.com/blog/ai-and-jobs-which-tasks-change/

AI changes tasks before it changes whole occupations: routine knowledge work is automated or supported, while accountability for judgment and results remains with people. For the executive team, this creates a delivery task. What matters is how quickly and how orderly the organization translates the new division of work into roles and processes. This article provides an overview and does not constitute legal advice.

## The analysis starts with tasks

**The question of what AI means for jobs can only be answered at the level of individual tasks. Every role consists of activities with very different potential for automation, and that is where management has to intervene.**

An occupation is made up of dozens of individual tasks. Anyone who asks about jobs at the level of job titles therefore receives answers that are of no use for a concrete decision. The analysis becomes meaningful once each role is broken down into its tasks: which of them follow fixed rules, which require judgment, and which depend on the relationship with customers or colleagues?

Research has worked with exactly this approach for years. A study of the US labor market concluded that for around 80 percent of workers, at least 10 percent of tasks are potentially touched by large language models. For roughly 19 percent, at least half of their tasks are affected. What is measured is exposure, meaning the technical possibility of doing a task faster or differently with a language model. Whether a job disappears does not follow from that. The study does show how far the change reaches into knowledge work.

## How many jobs in Germany are exposed to AI?

Around 38 percent of employees subject to social security contributions in Germany worked in 2022 in an occupation in which at least 70 percent of tasks could technically be performed by computers or computer-controlled machines. In 2019 the share was 34 percent. The Institute for Employment Research (IAB) explicitly attributes the increase to new capabilities of generative AI.

The direction of the shift is notable. The so-called substitution potential rose most sharply in expert occupations, which usually require a university degree, by almost ten percentage points to around 36 percent. In skilled occupations it stands at around 62 percent, in helper occupations at around 57 percent. Programming and writing, long core competences of highly qualified knowledge work, are now regarded as technically automatable to a large extent.

**Exhibit 1: Substitution potential is highest in skilled occupations and rose most in expert occupations**

| Label | Value |
| --- | --- |
| Skilled occupations | around 62% |
| Helper occupations | around 57% |
| Expert occupations | around 36% |

Source: Grienberger, Matthes, Paulus, IAB-Kurzbericht 5/2024
Note: substitution potential by skill level; in expert occupations up by almost ten percentage points
Graphic: [SVG](https://www.niehuesconsulting.com/blog/ai-and-jobs-which-tasks-change/grafik-1.svg) · [CSV](https://www.niehuesconsulting.com/blog/ai-and-jobs-which-tasks-change/grafik-1.csv) · [JSON](https://www.niehuesconsulting.com/blog/ai-and-jobs-which-tasks-change/grafik-1.json) · Niehues Consulting, [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/deed.en). Data: Grienberger, Matthes, Paulus, IAB-Kurzbericht 5/2024; the source's terms apply.

Two qualifications belong to any sound reading of these figures. First, the potential describes technical feasibility. Whether a task is actually replaced is decided by economics and quality, along with legal limits and the acceptance of employees. Second, potentials are rarely exploited in full in practice. For the executive team the figure therefore indicates the scale of the coming change, and it is no forecast for the headcount plan.

**What the studies show about the labor market**
38 percent | of employees subject to social security contributions in Germany worked in 2022 in occupations with high substitution potential, up from 34 percent in 2019
87.9 percent | average substitution potential of manufacturing occupations in 2022, the highest of all occupational segments
13.5 percent | average substitution potential of social and cultural service occupations in 2022, the lowest value
80 percent | of US workers for whom at least 10 percent of tasks are potentially touched by language models (exposure, not replacement)
170 million | new roles worldwide by 2030 according to an employer survey, across all drivers, AI included
92 million | roles displaced over the same period, across all drivers

## Which tasks does AI change first?

AI first changes tasks that consist of processing text and information along recognizable patterns. In practice, three groups can be distinguished:

- Information work with a clear template, such as summaries and standard correspondence as well as the pre-checking of documents. Here AI takes the first pass, and a person reviews and approves.
- Knowledge work involving judgment, such as analyses and proposals. Here AI shortens the preparation, and the decision stays with a person.
- Work built on relationships and accountability, such as negotiation and leadership. Here mainly the surroundings of the task change, because preparation and follow-up become faster.

The order of change follows task structure and hardly follows hierarchy. Case handling with a high share of text changes earlier than a leadership task, and within every role the standardized parts shift first. Office jobs and case handling are therefore affected early and broadly. What is replaced there are primarily work steps, and the position changes its profile. Whether this leads to fewer positions is decided by two factors: the growth of the business and whether the freed-up time flows into tasks that create value.

For divisional heads this means that change starts in each team at a different point, and it starts where employees already use language models. Whoever knows these points already holds the starting list for the analysis.

## A framework for the analysis

Every task has three possible treatments. It is automated, it is supported by AI, or the role is reshaped around the capacity that is freed up. Four screening questions decide which treatment fits.

**Four screening questions per task**
- How often does the task occur, and how uniform is it?
- What does an error cost, and who notices it?
- Are the required data and documents available in usable form?
- How will the business units respond to the change?

The first two questions determine the degree of automation, the third the effort of implementation, the fourth the pace of introduction. Frequent, uniform tasks with low error costs suit automation with sample checks. Tasks with high error costs remain support cases in which a person approves. This classification produces a priority list with a business case for each application, which feeds into the portfolio of AI management.

A simple calculation shows the scale for each role. For every task, multiply its share of the role's time by the expected time saving. A case handling team that spends 40 percent of its time on correspondence and saves half of that time gains 20 percent of its capacity. With ten employees in that role, this equals two full positions on paper. This figure is the starting point for deciding what the capacity will be used for. It becomes robust only once a prototype has measured the saving in the real workflow.

**Three ways to steer the change in work**
Criterion | Employee initiative | Central technology rollout | With Niehues Consulting
View of the work | Individual tasks without the overall picture | Tool in the foreground | Task analysis per role with a business case
Benefit in the workflow | Hard to measure | Measured only after rollout | Prototype proves the benefit in the real workflow
Freed-up capacity | Absorbed by coordination | Unplanned | Assigned to a value-creating task in advance
Acceptance in business units | Left to chance | Often delayed | Enablement and responsibilities from the start
Transfer into operations | Open | Technically regulated | Handover to internal or external teams

## Which new jobs does AI create?

AI mainly drives growth in roles that work with data and software or embed AI applications in workflows. A global employer survey by the World Economic Forum names big data specialists and fintech engineers, followed by AI and machine learning specialists as well as software and applications developers, as the fastest-growing occupations to 2030. According to the same survey, clerical roles decline fastest, among them cashiers and ticket clerks, administrative assistants and executive secretaries as well as data entry clerks.

Across all drivers, meaning technology as well as demographics, the green transition and the economic cycle, the employers surveyed expect 170 million new and 92 million displaced roles worldwide by 2030. For an individual company the balance says little, because the new roles arise in other places and with other profiles than the ones that disappear. Inside the company, new tasks often emerge first as part of existing roles: business ownership of an AI application, maintenance of the knowledge base, quality review of outputs.

**Exhibit 2: Employers expect more new than displaced roles worldwide by 2030, across all drivers**

| Label | Value |
| --- | --- |
| New roles | 170 million |
| Displaced roles | 92 million |

Source: World Economic Forum, The Future of Jobs Report 2025
Note: Drivers besides technology include demographics, the green transition and the economic cycle
Graphic: [SVG](https://www.niehuesconsulting.com/blog/ai-and-jobs-which-tasks-change/grafik-2.svg) · [CSV](https://www.niehuesconsulting.com/blog/ai-and-jobs-which-tasks-change/grafik-2.csv) · [JSON](https://www.niehuesconsulting.com/blog/ai-and-jobs-which-tasks-change/grafik-2.json) · Niehues Consulting, [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/deed.en). Data: World Economic Forum, The Future of Jobs Report 2025; the source's terms apply.

## How should a company use the time AI frees up?

A company uses freed-up time effectively when it defines before the rollout which task the capacity gained will take on and measures that task against a metric. Suppose a case handler drafts a report in half the time: value is created once someone has defined what the time saved is used for. Without that definition the saving dissolves into additional coordination and never shows up in a metric.

Four tasks have to be completed in delivery. First, responsibilities must be adjusted, because accountability for results shifts together with the tasks. Second, capability building has to target the concrete task. Third, employee representatives should be involved early: in Germany, where a works council exists, it must be informed about planned AI use, and applications that are objectively capable of monitoring performance or behavior are subject to co-determination. Fourth, every change needs a metric, such as lead time or error rate.

The delivery perspective demands that each of these tasks has an owner and a date. Only then does the analysis become a change that reaches the daily work of the teams.

## Niehues Consulting as a partner for implementation

[Niehues Consulting](/) accompanies organizations from the analysis of tasks to introduction in regular operations. The approach follows the Niehues Delivery Model with the phases Understand, Build and Introduce. For the change in jobs, three of its screening questions matter most: Which business result should change measurably? Which roles, decision paths and interfaces does operations need? Are users enabled and processes adapted?

In AI management, the work comprises business analysis and the business case, as well as the selection and sequencing of applications, steered with metrics. Prototypes prove the benefit in the real workflow with real data and are matured until the business units accept them. With delivery management and organization design, roles and decision paths take the form that changed tasks require.

The projects of Niehues Consulting range from VC-funded scale-ups to DAX 40 groups and have taken place in 20+ countries on six continents. The entry point is the free initial consultation of 30 minutes: [arrange an initial consultation](/#kontakt).

**What the executive team should take away**
Analyze roles by task and assign each task to one of the three treatments. Define in advance what freed-up time will be used for. Involve business units and employee representatives early, and measure every change against a metric. This turns a technology question into a managed change of work.

**Sources**
- Grienberger, Matthes, Paulus: Vor allem Hochqualifizierte bekommen die Digitalisierung verstärkt zu spüren, IAB-Kurzbericht 5/2024, 2024
- Eloundou, Manning, Mishkin, Rock: GPTs are GPTs: Labor Market Impact Potential of Large Language Models, Science, 2024
- World Economic Forum: The Future of Jobs Report 2025, 2025

## Frequently asked questions

### Will AI take away jobs?

AI mainly changes tasks, and whether that leads to job cuts is decided by each company through how it uses the freed-up capacity. Worldwide, employers expect 170 million new and 92 million displaced roles by 2030 across all drivers, AI included. Individual roles such as data entry and administrative assistance shrink markedly, while others grow.

### Will AI replace office and administrative jobs?

In office and administrative jobs, AI first replaces individual work steps such as data capture and summarizing as well as standard correspondence. The position changes its profile as a result: checking and deciding gain weight, as does resolving exceptions. Where volumes stagnate and no new tasks are planned, staffing needs fall.

### Which jobs are safe from AI?

The technical automation potential is lowest in occupations involving personal care, physical work in changing environments and responsibility for people. In Germany it stands at around 14 percent for social and cultural service occupations such as education and social work, and at around 27 percent for medical health occupations. Hardly any occupation can be replaced completely, while almost all will change.

### How many jobs in Germany are affected by AI?

Around 38 percent of employees subject to social security contributions worked in 2022 in occupations in which at least 70 percent of tasks could technically be automated. The figure describes technical feasibility and not expected job losses. Considerably more employees are touched through individual tasks.

### What new jobs will AI create?

Roles around data and AI grow fastest, such as big data specialists, AI and machine learning specialists, and software and applications developers. Inside companies, new tasks also emerge within existing roles, such as business ownership of an AI application or quality review of its outputs.

### Can an employer in Germany make redundancies because of AI?

Yes, if the introduction of AI rests on a business decision that permanently removes the need for the work and no continued employment in a vacant position is possible, including after reasonable retraining. Where the Protection Against Dismissal Act applies, a social selection must also be carried out, and an existing works council must be consulted before every dismissal. If the change affects larger parts of the workforce, negotiations on a reconciliation of interests and a social plan follow.

### Does the works council have to be involved when AI is introduced?

Yes, wherever a works council exists. The German Works Constitution Act requires it to be informed about planned AI use, and any application that is objectively capable of monitoring performance or behavior is subject to co-determination. Early involvement therefore belongs in every rollout plan.

### How does a company prepare for changing tasks?

With a task analysis for each role, a prototype in the real workflow and a clear plan for responsibilities and enablement. Transfer into regular operations follows, with metrics that let the executive team track progress.
