This article explains how to use the AI Impact Analyzer in the AI Transformation Hub. The AI Impact Analyzer groups the positions of your employees into job families and shows how AI could change the way your teams work: where it could save time, free up capacity, and shift roles.
You can use the AI Impact Analyzer to see:
- The automation potential of each job family on four levels: low, medium, high, and very high.
- Which tasks in a job family AI can speed up and which tasks stay manual.
- How to get started with a task. This can be with a chat assistant (light setup) or with AI connected to your tools (extended setup).
- How many of your positions are in job families with high or very high automation potential.
The AI Impact Analyzer is in beta. Personio is still improving it and uses your feedback to decide on changes. To share feedback, go to AI Transformation and click Report feedback.
Before you start
Permissions
To see the AI Impact tab, you need an Administrator role.
Job families
The AI Impact Analyzer matches the position titles of your employees to job families. Personio does this automatically, and you can’t change the matching.
If you’ve set up your job architecture, the analyzer shows your job family names. However, you don’t need to set up your job architecture to use the analyzer. If you haven’t, it shows Personio-defined job family names. For example, Personio’s job families include Recruiting and Learning & Development, within the Human Resources job family group.
Understand how the analysis works
- Personio automatically matches the position titles of your employees to job families. It uses its own job title model. You can’t change the matching.
- Each job family has a list of tasks. The tasks come from O*NET, the standard US framework that describes occupations and their tasks.
- Each task gets an Automation potential label. This is a Personio estimate of how much time AI saves on the task. Tasks with medium or high automation also get an AI Setup label. This comes from external research combined with Personio estimates and shows what you need to get started.
- Personio calculates the automation potential of each job family from the automation potential of its tasks.
Personio doesn’t use an AI model on your employee data for this analysis. It updates the analysis every night.
Understand the two task labels
| Label | Value | Meaning |
| Automation potential | High automation | AI saves 80% or more of the time. A person only checks the result and signs off. |
| Medium automation | AI saves 50% to less than 80% of the time, but a person still does a meaningful share. | |
| Manual | AI saves less than half of the time. Examples: in-person, physical, or judgment-heavy work. | |
| AI Setup | Light setup | A chat assistant is enough. This means a large language model (LLM) on its own, used in a chat interface. You need a license, training, and data rules. |
| Extended setup | An LLM on its own isn’t enough. You first need to connect AI to your tools, systems, and documents. You need an integration. |
Manual tasks have no AI Setup label. The AI Setup only tells you how to get started. It doesn’t change the automation potential of a task or the level of a job family.
Where the labels come from
- The AI Setup label is based on a research study (Eloundou et al., 2023) combined with Personio’s own estimates. In it, every task was rated on whether a large language model (LLM) could cut the time by at least half at the same quality and what it takes to get there.
- How much time AI saves (medium automation or high automation) is a Personio estimate on top of the AI Setup label.
How Personio calculates automation potential
The level of a job family shows which share of its tasks AI can speed up. Personio counts the tasks with medium automation or high automation:
Share = (high automation tasks + medium automation tasks) ÷ all tasks of the job family
| Level | Share of tasks with medium or high automation |
| Very high | 90% or more |
| High | 75% to less than 90% |
| Medium | 50% to less than 75% |
| Low | Less than 50% |
The thresholds are the same for every job family and every organization. The analysis shows the level only. It never shows the share as a number.
For example, a job family has 20 tasks, and 12 of them have medium or high automation. The share is 60%, so the job family has medium automation potential.
Sources
- Eloundou, T., Manning, S., Mishkin, P., Rock, D. (2023). GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models. arxiv.org/abs/2303.10130
- Task ratings: github.com/openai/GPTs-are-GPTs (file full_labelset.tsv)
- Includes information from O*NET Online by the U.S. Department of Labor, Employment and Training Administration (USDOL/ETA). Used under the CC BY 4.0 license. Personio has modified all or some of this information. USDOL/ETA has not approved, endorsed, or tested these modifications.
- Personio Job Title Model (job families and job family groups) and Personio automation potential estimates.
Explore the AI Impact Analyzer
To explore the AI Impact Analyzer, go to AI Transformation and click the AI Impact tab. Then, follow the steps below.
Read the introduction
The first time you open the page, a window opens. It has three parts:
- Where to start: Personio ranks job families by how many of their tasks AI can speed up. Start with the ones at the top of the table.
- How to start: Open a job family to explore its tasks, automation potential, and AI setup.
- What are the estimates based on? General guidance based on external research and Personio’s estimates. It uses typical task descriptions of these job families matched to your positions and does not see your processes or tools.
The window also shows the notes from the disclaimer. Click I understand to continue, or click Read the full methodology first.
Take the product tour
A product tour walks you through the page.
Check the summary
At the top of the page, you can see the Job family automation potential summary. This gives a high-level overview of all job families with at least three positions. It shows four tiles, grouped by automation potential: low, medium, high, and very high. Each tile shows the number of job families and the number of positions at that level.
To the right of the page, you can see the By positions card. A donut chart shows how many of your positions are in job families with high or very high automation potential.
Find a job family
In the center of the page, you can see the Impact on job families table. Use the toolbar to do the following:
- Search by job family or job family group.
- Filter by automation potential level. The options are All levels, Very high, High, Medium, and Low.
- Group by the following values: No grouping, Job family group, or Department.
- Sort job families by the following values: Automation potential (default), Positions, or Job family (A to Z).
The table contains the following columns:
- Job family
- Number of positions
- Automation level
- Automation potential by task. This column shows a bar that splits its tasks into high automation, medium automation, and manual, with counts.
Open a job family
In the Impact on job families table, click a row to open the side panel. This lets you drill down into each job family in detail. The side panel shows:
- The automation potential level.
- Mapped positions: The position titles from your company data that are matched to this job family.
- Automation potential by task: A bar and counts for high automation, medium automation, and manual.
- The AI Setup: A tag that tells you how to get started for light setup and extended setup.
- Task list: All tasks of the family, each with its corresponding labels. Use the icons under Tasks to filter the task list by automation potential and AI setup:
- Under Automation potential, select High automation, Medium automation, or Manual.
- Under AI Setup, select Light setup or Extended setup. The setup filter applies to medium automation and high automation tasks.
When you open a job family, all tasks mapped to this job family appear. Click Show all tasks if no task matches your filters.
Tasks are ordered high, medium, and then manual. Within each group, light setup comes before extended setup.
Read the methodology
To see how results are calculated, click Methodology & sources at the top of the page.
What the analysis includes and doesn’t include
Not every position or job family appears in the analysis. A mapped position is a position that Personio matches to a job family.
Included
- Your positions, taken from the free-text position field in your employee data. People on leave are included in the position counts.
- Job families with at least three mapped positions.
- The tasks of each job family, from O*NET. Each task has an Automation potential label. Medium and high automation tasks also have an AI Setup label.
Not included
- Job families with fewer than three mapped positions. A note under the table shows how many job families are hidden.
- Positions that Personio can’t match to a job family. A note under the table and under the donut chart in the By positions card shows how many positions this affects.
- Positions that Personio matches to the fallback job family "General." This job family is for unclear or very general titles, for example, "Manager" or "Working Student." These positions aren’t mapped, so the analysis leaves them out.
Keep in mind
This analysis gives you an orientation on how AI could affect the tasks in your job families. Please keep the following in mind:
- The results are estimates for orientation. They aren’t a forecast of job changes or headcount, and they aren’t a recommendation to reduce roles.
- The analysis doesn’t rate individual people. Don’t use the results to evaluate, select, or monitor individual employees.
- The results are based on O*NET, a description of US occupations, matched to your positions. The analysis doesn’t see your processes or tools.
- Depending on your country and company, you may need to inform or consult employee representatives before you act on these results.
- Decisions about roles, staffing, or tools remain the responsibility of your company.
FAQ
What is the AI Impact Analyzer designed for?
It gives you an orientation on how AI could change the way your teams work: where it could save time, free up capacity, and shift roles. Your company decides what to do with the results.
Is the result a forecast or a recommendation to reduce roles?
No. The results are estimates for orientation. They aren’t a forecast of job changes or headcount.
Can I use the AI Impact Analyzer to evaluate employees?
No. The analysis doesn’t rate individual people. Don’t use it to evaluate, select, or monitor individual employees.
Who can see the AI Impact Analyzer?
Anyone with an Administrator role can access the AI Impact Analyzer.
What does “automation potential” mean?
It shows which share of the tasks in a job family AI can speed up (tasks with medium or high automation). The more tasks AI can speed up, the higher the level.
Does the AI Setup change the automation potential?
No. The AI Setup only tells you how to get started with a task.
Why don’t I see a job family in the list?
The list shows only job families with at least three positions in your account. If a job family has fewer positions, it isn’t shown.
Are these the same job families as in my job architecture?
It depends. If you’ve set up your job architecture, the AI Impact Analyzer shows your job family names. If you haven’t, it shows Personio’s own job family names. In both cases, Personio matches the position titles of your employees to job families automatically, and you can’t change the matching
Why isn’t one of my positions included?
Titles that are very general, such as "Manager" or "Working Student," often can’t be matched to one job family. Personio then matches them to the fallback job family "General," which is left out of the analysis.
Why doesn’t the analysis show a number?
The analysis shows levels only. The thresholds are the same for every job family and company.
Can I change how a task is rated?
No.
Does Personio use my employee data with an AI model for this analysis?
No.
How often does Personio update the analysis?
Every night.
Do I need to involve my works council or employee representatives?
Depending on your country and company, you may need to inform or consult employee representatives before you act on these results.
What should I do if I find a bug or unexpected behavior?
Click Report feedback in the AI Transformation Hub. Describe the issue and what you did before it happened.