Why the work activities list is where exposure analysis actually starts
A People Analytics Lead sitting down three weeks before a board meeting does not need another article about AI and the future of work. She needs to know what happens inside the 40 roles in her operations department, at the level of what people actually spend their day doing. The board has asked for a company-specific answer, and the honest answer starts with a list most HR teams have never opened: the O*NET work activities list.
O*NET — the Occupational Information Network, maintained by the U.S. Department of Labor's Employment and Training Administration — breaks every occupation down into the general categories of things workers do, the more specific versions of those categories, and the individual tasks underneath them. This article explains what that list contains, how it's structured, and how a transparent exposure methodology uses it to move from "AI might affect HR broadly" to "here is what the AI exposure of a Payroll Specialist role actually looks like, and why."
What's actually in the O*NET work activities list
ONET's occupational database covers 1,016 occupational titles, with 923 of those carrying full descriptive data, representing roughly 55,000 job titles in the U.S. economy (ONET Resource Center, USDOL/ETA, 2019). Each occupation is described using roughly 277 standardized descriptors covering skills, knowledge, work context, and — most relevant here — work activities and tasks (U.S. DOL, 2025). The database is maintained on a regular update cycle, with the primary annual update typically arriving in the third quarter (U.S. DOL, 2025).
The work activities themselves come in two layers:
- Generalized Work Activities (GWAs) — broad categories such as "Getting Information," "Analyzing Data or Information," or "Interacting With Computers." These are the same 40-odd categories across every occupation in the database, which is what makes cross-role comparison possible in the first place.
- Detailed Work Activities (DWAs) and Tasks — the specific, occupation-level statements underneath each GWA. For a Payroll Specialist, "Analyzing Data or Information" resolves down to detailed tasks like "process and record employee time and payroll data" or "verify computations against source documents."
This structure matters because it's what lets an exposure methodology stay grounded in what the job requires rather than what the job title implies. Two roles titled "Analyst" can have almost no task overlap. Two roles with very different titles — a Bookkeeping Clerk and a subset of a Financial Analyst's duties — can share substantial task overlap. The GWA/DWA/task hierarchy is what surfaces that.
For a deeper walkthrough of how O*NET itself is structured and where this data comes from, see what O*NET is and how it works.
Every work activity carries an importance rating — and that's the input that matters
A work activity list by itself is descriptive, not analytical. What turns it into something scoreable is the importance rating attached to each task and activity within an occupation — a measure, on O*NET's standard scale, of how central that activity is to doing the job well.
This is the detail HR analysts most often skip, and it's the one that changes the output the most. A task rated low-importance for an occupation shouldn't move that occupation's exposure score much even if the task itself is highly automatable; a task rated high-importance should. Treating every listed task as equally weighted produces a flattened, misleading picture — a role can look far more or less exposed than it is if importance weighting is ignored.
We cover the mechanics of this weighting — including how importance ratings interact with O*NET's separate frequency and level scales — in more depth in O*NET task importance ratings explained. It's worth reading before building anything on top of the raw activity list.
Before you touch the activities: map the job title correctly
None of this works if the underlying occupation mapping is wrong. ONET data is organized under the 2018 Standard Occupational Classification system, which defines 867 detailed occupations rolling up into 459 broad occupations, 98 minor groups, and 23 major groups (U.S. BLS). A job title inside a company rarely maps one-to-one onto a single SOC/ONET occupation — "Senior Operations Coordinator" might split across two or three O*NET occupations depending on which duties actually consume the role's time.
Getting this step wrong upstream means every downstream exposure score is built on the wrong foundation, no matter how careful the task-level analysis is. The full process — how to take a messy internal role list and land it on the correct O*NET occupation codes — is covered in how to map job titles to O*NET occupations.
A worked example: from activity list to a role's exposure profile
Here's a simplified, illustrative walkthrough — round numbers, for teaching the method, not a real occupation's published data.
Say a role's O*NET-mapped occupation lists four generalized work activities relevant to a given exposure dimension, each with an importance rating on a 0–100 scale after normalization:
- Processing structured data inputs — importance 90
- Reviewing standardized documents against rules — importance 75
- Communicating routine status updates — importance 40
- Resolving novel client disputes — importance 60
If the first three activities score high on cognitive routine exposure (a task dimension in a four-part rubric — cognitive routine, physical routine, social/judgment, and creative, each scored 0–100) while the fourth scores low on cognitive routine but high on social/judgment, an importance-weighted rollup will pull the role's cognitive-routine score up (because the highest-importance activities are the most routine ones) while keeping its social/judgment score meaningfully above zero (because the dispute-resolution activity, while lower-importance, isn't trivial).
The output of that rollup is not a verdict. It's a profile across four dimensions that places the role into one of three categories — Monitor, Review, or Redeployment Candidate — as an input for a People Analytics Lead's own judgment, not a replacement for it. The role in this example, with meaningful weight on both cognitive-routine and social/judgment activities, would likely land in "Review": worth a closer look, not a role anyone should describe as slated for automation. For the full four-dimension scoring logic, see the task-based AI exposure methodology.
Where analysts get this wrong
Three recurring mistakes show up when teams try to do this manually in a spreadsheet:
- Treating the activity list as flat. Pulling every GWA for an occupation and averaging them, without importance weighting, produces scores that don't reflect what the role actually spends time on.
- Skipping the SOC mapping step. Matching on job title text instead of verifying the underlying occupation code introduces error before scoring even begins.
- Using stale data. Because ONET updates its content regularly, with a primary release typically each Q3, an analysis built on an old download can miss recent revisions to task lists or importance ratings. Confirm the release version — the current production release is ONET database v30.3 — before treating any pull as current.
Turning the list into something repeatable
The ONET work activities list is public, well-documented, and free to use — this site incorporates information from ONET, used under the CC BY 4.0 license; O*NET is a trademark of USDOL/ETA. What it doesn't come with is a pre-built importance-weighting engine, a SOC mapping workflow for messy internal job titles, or a four-dimension exposure rubric that turns the raw list into a department-level heatmap.
That's the gap a structured, self-serve tool closes for teams who don't have the analyst-hours for a manual O*NET build-out. If you want the reference workflow we use internally — including the exact mapping and weighting steps referenced above — the O*NET Task-Mapping Methodology Handbook walks through it step by step. For teams ready to run this against their own role list, pricing outlines the available tiers.