Measuring AI's impact on a workforce means breaking roles into their component tasks, scoring each task's exposure to AI capability, and rolling those scores back up into role- and workforce-level estimates under more than one adoption scenario. Done properly, it produces a task-by-task map of where AI changes work, how much time that could free up or shift, and what that means for headcount, hiring, and role design.
This is a how-to, not a definitions piece. If you want the background on what an AI workforce impact assessment is and why task-level analysis beats job-title-level guessing, that's covered elsewhere. Here, the focus is the process itself: the six steps you'd actually run, in order, whether you're doing it yourself or asking a vendor to walk you through it.
Why Task Level, Not Job Level
Job titles are too coarse to measure AI impact accurately. Two people with the same title can spend their weeks on completely different mixes of tasks, and AI capability doesn't respect job titles either. It's good at some tasks and weak at others, often within the same role.
GoFIGR's method addresses this by working at the task level throughout: break roles into their component tasks, assess each task's exposure to AI capability on its own terms, then aggregate the task-level results back up into role-level and workforce-level impact. That aggregation step matters. It's what turns a pile of task scores into a number a workforce planner can actually use.
The Six-Step Method
Step 1: Map the Roles
Start by listing the roles you want to assess, whether that's a single function (HR, finance, customer support) or the full workforce. For each role, note headcount and any obvious sub-variants (a generalist HR role looks different at a 50-person company than at a 5,000-person one).
This step is mostly inventory work. The goal is a clean list of roles to feed into the next step, not an analysis yet.
Step 2: Break Roles into Tasks
For each role, decompose it into its component tasks. A single HR generalist role might break into 15-25 discrete tasks: scheduling interviews, drafting offer letters, running onboarding sessions, answering policy questions, processing leave requests, and so on.
This is the step most impact assessments skip, and it's the one that makes the biggest difference to accuracy. GoFIGR's HR research mapped 1,800 tasks across 100 HR roles to do this at scale, which is roughly 18 tasks per role, a useful benchmark for how granular this step needs to be.
Step 3: Assess Each Task's Exposure to AI
For every task, assess how exposed it is to current and near-term AI capability. This isn't a yes/no judgment. It's a graded assessment of whether AI can perform the task, assist with it, or barely touch it at all, based on what the task actually requires (judgment, physical presence, relationship-building, structured information processing, and so on).
This is where GoFIGR's research found that 84.3% of HR tasks show some AI impact within three years under a Conservative adoption scenario. That figure is a task-level finding, not a claim that 84.3% of HR jobs disappear. Most of those tasks are affected in the sense of being changed, assisted, or partially automated, not eliminated outright, and the number depends on the adoption scenario used (more on that in Step 5).
Step 4: Estimate Time Impact per Task
Exposure alone doesn't tell you much on its own. The next step is estimating how much time each task currently takes and how much of that time could shift if the task's exposure is realized (through automation, AI assistance, or task redesign).
This converts an exposure score into something with practical weight: hours per week, per role, per team. A highly exposed task that only takes twenty minutes a month matters less than a moderately exposed task that eats a full day a week.
Step 5: Run Multiple Adoption Scenarios
This is the step that separates a usable assessment from a single scary or reassuring headline number. AI capability existing is not the same as AI being adopted, and adoption is not the same as it actually changing how work gets done. Those are three different gaps, and each one slows down or speeds up how much impact shows up in practice.
GoFIGR's method runs multiple adoption scenarios, typically conservative, moderate, and aggressive, rather than producing one point forecast. A conservative scenario assumes slow uptake and heavy human oversight. An aggressive scenario assumes fast uptake and deeper process redesign. Reporting a range instead of a single number gives workforce planners something they can actually plan against, and it makes the method's assumptions visible instead of hidden inside one number.
Step 6: Turn Results into Workforce Decisions
The output of Steps 1-5 is a dataset: roles, tasks, exposure scores, time estimates, scenario ranges. Step 6 is where that dataset turns into decisions: which roles to redesign, where to slow or pause hiring, where to redeploy freed-up time into higher-value work, and which skills to build or buy for the next 12-24 months.
This is also the step most organizations rush or skip, because it requires connecting the analysis to actual workforce planning, not just producing a report that sits in a folder.
Assumptions Worth Stating Up Front
Any credible version of this method rests on a few explicit assumptions. First, that tasks (not job titles) are the right unit of analysis, because capability and exposure vary within a role. Second, that "AI can do this task" and "AI is doing this task at your company today" are different claims, which is why Step 5 exists. Third, that the output is a decision input, not a prediction of exact headcount numbers; scenario ranges are meant to bound uncertainty, not eliminate it.
Where a stat is cited, it should be traceable to a specific study and scenario, not presented as a universal law. The 84.3% figure above, for instance, is specific to HR tasks, a three-year window, and a conservative adoption scenario from GoFIGR's own research, and it will differ for other functions or other scenarios.
Running This Yourself vs. With Support
Organizations can run a lightweight version of this process internally, especially for a single team or function, using the six steps above as a checklist. The harder parts in practice are usually Step 2 (task decomposition takes real time and domain knowledge) and Step 5 (building credible scenarios instead of guessing).
For an individual, quick first look, GoFIGR's free AI Impact Assessment is a self-serve starting point. For a full workforce or organization-wide view, the enterprise diagnostic runs this six-step method with support, through a booked AI briefing, and produces the role-by-role, scenario-based output described above rather than a single top-line number.
Frequently Asked Questions
How long does an AI workforce impact assessment take?
It depends on scope. A single role or small team can be mapped and assessed in a matter of days once the task list exists. A full organizational assessment, covering many roles across multiple functions with adoption scenarios, is a larger undertaking, generally weeks rather than days, mostly because of the time needed to accurately decompose roles into tasks.
What's the difference between AI exposure and AI impact?
Exposure is whether a task could plausibly be done or assisted by current AI capability. Impact is what actually happens once you factor in adoption speed, organizational readiness, and how much the task's time cost matters. A highly exposed task with low time cost and slow adoption may have minimal real impact for years.
Why use multiple adoption scenarios instead of one number?
Because AI capability existing, AI being adopted, and AI actually changing daily work are three separate gaps, each with its own pace. A single forecast hides those gaps behind one number. A conservative-to-aggressive range shows planners the boundaries of likely outcomes instead of a false point estimate.
Do I need to assess every task in a role, or can I sample?
For an accurate picture, decompose the role fully. Sampling tasks tends to miss the low-visibility, high-frequency tasks (scheduling, data entry, routine follow-ups) that often carry a lot of the time cost and exposure, even though they don't show up when someone describes their job in a sentence.
Does a high exposure score mean those jobs will be automated away?
Not necessarily. Task-level exposure measures whether AI can perform, assist with, or partially handle a task, which is different from a role disappearing. Most roles are a bundle of tasks with a mix of exposure levels, so the more common outcome is role redesign and time reallocation rather than outright elimination.
Can I run this assessment myself, or do I need a vendor?
The method itself (map roles, break into tasks, assess exposure, estimate time, run scenarios, decide) can be run internally with enough time and task-level knowledge of the roles involved. Where teams typically want support is in the task decomposition and scenario-modeling steps, which is what a fuller diagnostic, like the one run through GoFIGR's booked AI briefing, is built to handle at scale.
Helena Turpin spent 20 years in talent and HR innovation where she solved people-related problems using data and technology. She left corporate life to create GoFIGR where she helps mid-sized organizations to develop and retain their people by connecting employee skills and aspirations to internal opportunities like projects, mentorship and learning.
