Why a board-ready assessment looks different from a vendor demo
You have three weeks until the board meeting. The company's PE sponsor asked a direct question at the last update: "Which of our roles are exposed to AI, and what are we doing about it?" You said you'd have an answer. Now you're staring at a headcount report in one tab and a dozen open browser tabs on AI workforce trends in another, and none of it adds up to something you could defend if a board member asked, "how did you get this number?"
This is the gap between reading about AI and workforce change in the abstract and producing something specific to your organization. Macro research from McKinsey, the World Economic Forum, and Goldman Sachs is useful context, but none of it tells you whether your underwriting team or your logistics coordinators are the ones to look at first. Gartner's 2024 HR research found that 86% of HR leaders have not implemented strategic workforce planning, and a related Gartner survey found that 66% say their workforce planning is limited to headcount forecasting or that they struggle to show its ROI — which is another way of saying most HR functions have never had to build this kind of artifact before.
This article lays out a repeatable, five-step method for running an AI workforce readiness assessment at the company level — from building a clean role inventory through producing a document your board and your CFO can both trust.
Step 1: Build a clean role inventory before you touch exposure scoring
Every defensible assessment starts with an unglamorous step: a complete, deduplicated list of roles across the organization, not job titles pulled from an org chart that hasn't been updated in two years. The inventory should capture, at minimum, the role title, the department, headcount in the role, and — ideally — a short description of the core duties as actually performed, not as written in a five-year-old job posting.
This matters because every downstream step depends on mapping each role to the right occupational reference. A role inventory with ambiguous or duplicate titles ("Analyst," "Senior Analyst," "Analyst II" all describing the same work) will produce inconsistent mappings and an assessment that falls apart under scrutiny. Spend the time here. It is the single highest-leverage step in the process, and it is also the step most teams try to skip.
If you need a structured template for this stage, our guide to building a company-specific AI workforce exposure assessment walks through the inventory format in more detail, including how to handle hybrid roles that span two functions.
Step 2: Map roles to O*NET occupations and tasks
Once the role inventory is clean, each role needs to be mapped to a standardized occupational reference — not an internal title, but an external, independently maintained taxonomy that lets you compare a logistics coordinator at your company to the same role anywhere else in the economy. This is what O*NET (the Occupational Information Network, maintained by the U.S. Department of Labor) is built for.
ONET currently covers 1,016 occupational titles, with 923 of them carrying full descriptive data, representing more than 55,000 individual jobs in the U.S. economy, according to the ONET Resource Center. Each occupation is described by roughly 277 standardized descriptors — including tasks, skills, knowledge areas, and work activities — which the Department of Labor updates on a rolling basis, with a primary annual update in the third quarter. O*NET occupations sit inside the Bureau of Labor Statistics' Standard Occupational Classification system, which organizes the economy into 867 detailed occupations, 459 broad occupations, 98 minor groups, and 23 major groups under the 2018 SOC structure.
The mapping step means assigning each role in your inventory to its closest ONET-SOC occupation code, then pulling that occupation's task list as the raw material for scoring. Most roles map cleanly to a single occupation. Some — particularly hybrid or newly created roles — will require judgment calls, and those calls should be documented, not hidden, because a board member or auditor may ask about them later. For a deeper walkthrough of how the ONET taxonomy is structured and why it is the right reference point for this work, see what O*NET is and how it works.
This section incorporates information from ONET. Used under the CC BY 4.0 license. ONET is a trademark of USDOL/ETA.
Step 3: Score exposure across four dimensions, not one
This is where most informal "AI risk" exercises go wrong: they try to assign a single exposure number to a role based on a gut sense of how "automatable" it sounds. A defensible methodology scores each task within a role across multiple independent dimensions and then combines them using the task's actual importance to the occupation, as documented in O*NET.
A transparent four-dimension rubric scores every task on:
- Cognitive routine — how standardized and rule-based the thinking involved is (0–100)
- Physical routine — how repetitive and specifiable the physical component of the task is (0–100)
- Social/judgment — how much the task depends on interpersonal nuance, negotiation, or contextual judgment (0–100)
- Creative — how much the task requires original or non-formulaic output (0–100)
Each O*NET occupation lists its component tasks along with an importance weighting — some tasks matter far more to the job than others. A worked example: if a role has three tasks weighted at 50%, 30%, and 20% importance, and each task carries its own score across the four dimensions, the role-level exposure score is the importance-weighted average of those task scores, not a simple average. A task that's 90% automatable on paper but represents only 5% of the role's importance should not dominate the result, and a defensible methodology prevents that from happening.
For the full scoring logic, including how importance weighting interacts with each dimension, see our breakdown of the task-based AI exposure methodology.
Step 4: Roll task scores into role and department heatmaps
Once every role in the inventory has a four-dimension exposure score, the next step is aggregation — rolling individual role scores up into department- and function-level views. This is the layer that actually gets presented to a board, because no executive wants to review 400 individual role scores; they want to see which functions carry the most exposure concentration and why.
A well-built heatmap should let a reader drill down from the department level into the specific roles and even the specific tasks driving a high score, because a leadership team's first question after seeing a red cell on a heatmap is almost always "which tasks, specifically, and why." An aggregation that can't answer that question in the same view isn't finished yet — it's a chart, not an assessment.
The exposure score is a research input to a management decision. It is not a verdict about any person's job.
This distinction matters enough to repeat in whatever form you present the output: a Monitor rating describes task composition today, not a prediction about tomorrow's organization chart.
Step 5: Turn exposure into Monitor, Review, and Redeployment Candidate actions
A heatmap without a next step is just a more colorful spreadsheet. The final step in a readiness assessment translates exposure scores into a small number of clear action tiers your leadership team can actually act on — not a verdict, but a prioritization layer for human judgment:
- Monitor — exposure is present but modest relative to the role's judgment and creative components; revisit at the next assessment cycle.
- Review — exposure concentration is high enough in specific tasks to warrant a structured conversation about workflow redesign, tooling, or training.
- Redeployment Candidate — a meaningful share of the role's weighted task composition overlaps with skills that transfer to other open or adjacent roles in the organization, and that overlap is worth exploring with the employee and the function lead.
The redeployment layer is where O*NET earns its keep twice over: it doesn't just describe the role you're assessing, it describes the skill and task profile of every other occupation in its database, which is what makes a credible internal redeployment match possible in the first place. This connects to a broader trend the World Economic Forum has flagged: in its Future of Jobs Report 2025, WEF found that of every 100 workers needing training through 2030, 59 will need it — 29 are expected to be upskilled in their current role, 19 reskilled or redeployed internally, and 11 are unlikely to receive adequate training. A readiness assessment that identifies Redeployment Candidates with specificity is one practical way to push more workers into that 19-of-100 group rather than the 11-of-100 group.
Building the board-ready output and keeping the assessment repeatable
The deliverable a board actually wants is not a spreadsheet of scores — it's a short narrative, backed by a methodology they can interrogate, that answers three questions: where is exposure concentrated, what is the company doing about the highest-priority functions, and how will this be reassessed as the business and the technology both change. The methodology itself — the role inventory, the O*NET mapping, the four-dimension scoring, the importance weighting, the action tiers — is what lets you answer "how did you get this number" with a straight answer instead of a shrug.
Because O*NET's descriptor set is updated on a rolling basis and your own role mix changes with every reorg, a readiness assessment is not a one-time project. Building the process as a repeatable cycle — ideally run at the same cadence as your workforce planning calendar — is what turns a one-off board deliverable into an ongoing capability your HR function owns.
If you want a structured, ready-to-use version of this five-step process, our Workforce AI Readiness Assessment Guide packages the inventory templates, the ONET mapping worksheet, and the scoring rubric into one document, and the Workforce AI Readiness Complete Kit extends that into board-presentation templates and a redeployment-matching worksheet. For teams that want to run this process directly against their own role list rather than building the spreadsheets by hand, our self-serve workforce assessment tool automates the ONET mapping and four-dimension scoring steps described above, so you can move from role inventory to a department heatmap in hours rather than weeks of analyst time.