Assess AI disruption by tracing a path from tasks to workflows, from workflows to customer demand and company costs, and from those changes to the value the company can actually capture. Task exposure is an early signal—not a forecast of lost revenue, job cuts, or investment returns.
What company-level AI exposure means
A company’s exposure is not simply the share of its employees in occupations that AI could affect. The relevant question is how AI might change the work the company performs, the customer outcomes it sells, and the economics of delivering them. A task may be technically automatable yet too risky, costly, or difficult to integrate to replace a real workflow. Even when adoption occurs, customers, employees, suppliers, or competitors may capture more of the resulting value than the company does.
Keep four questions separate throughout an assessment:
- Capability: What can current AI systems do for the tasks involved?
- Adoption: Can the capability be used reliably, economically, and within the relevant workflow?
- Market effect: How might adoption change competition, customer behavior, demand, or prices?
- Value capture: Which company, customer, worker, or supplier gains the economic benefit?
There is no source-established, validated cross-industry score that answers all four questions for a company. A task-exposure index can help identify where to investigate, but it is not a company valuation or a probability of disruption.
#1 Best Overall
Start with the company’s economic engine
Before estimating AI risk, map how the company creates and earns value. Use filings and other company disclosures to identify its major products and services, customer groups, pricing basis, recurring versus transactional revenue, and principal costs. For each offering, state the customer outcome it delivers: for example, a completed transaction, a decision, a piece of work, or access to a trusted service.
Then identify what supports that outcome and may affect the company’s ability to defend it:
- Proprietary data, specialized expertise, or access to scarce inputs
- Distribution, customer relationships, or an established user base
- Trust, regulatory standing, or accountability for the result
- Integration with customers’ systems and processes
- Switching costs, network effects, or service quality
These features are hypotheses to test, not automatic protection from AI. Ask whether AI makes the feature more valuable, easier for competitors to reproduce, or less important to customers.
Map AI-exposed tasks to real workflows
Build two maps: one of significant work inside the company and another of the customer workflows its offerings support. Prioritize activities that are high-volume, costly, central to the customer outcome, or material to quality and risk. For each activity, classify the plausible effect as automation, assistance to a worker, faster or better output, a new AI-enabled offering, or a route for customers to bypass an intermediary.
Do not equate an AI-generated output with a completed job. Evaluate the full workflow, including data access, handoffs, human review, exceptions, accuracy requirements, and integration. A system that drafts a response may assist a service team without replacing its responsibility for resolving the customer’s problem. Conversely, a tool that lets customers complete a task themselves may pressure an intermediary even if that intermediary does not use AI to automate its own staff.
Use exposure indices as screening evidence
The International Labour Organization (ILO) published a 2025 GenAI exposure index based on occupational task descriptions. Its development combined a representative sample of 29,753 tasks in Poland’s occupational classification, perceived automation-potential input from 1,640 employed people, expert discussion, and model predictions. The study also used 52,558 data points on automation potential for 2,861 tasks. The ILO reports that clerical work remains highly exposed and that exposure is rising for some digitized professional and technical roles.
Rank #3
The ILO’s global estimates put one in four workers in an occupation with some GenAI exposure, while 3.3% of global employment falls in its highest exposure category. In that highest gradient, the estimates are 4.7% of female employment and 2.4% of male employment. The ILO also reports overall exposure of 11% of employment in low-income countries and 34% in high-income countries. These are occupational exposure estimates, not estimates of a particular company’s revenue at risk, probability of failure, or likely job losses.
The ILO’s 17 April 2026 brief explains why such estimates need careful interpretation: indices use static descriptions of current tasks, may embed subjective assumptions, and do not establish economic feasibility or account fully for institutional barriers, workflow changes, or adjustments in employment, wages, and demand. The ILO states that exposure measures “offer risk assessments about potential job transformations but cannot be interpreted as predictions of job displacement, productivity gains or reskilling needs.” Treat the index as a prompt to investigate company-specific workflows, not as a percentage to apply to revenue or headcount.
Test whether adoption is feasible and real
For each priority workflow, distinguish technical possibility from deployable use. Check whether the company or its customers can provide the necessary data, meet accuracy and security requirements, handle exceptions, and integrate the system into existing processes. Consider regulatory acceptance and the cost of oversight as well as the model’s capabilities.
Rank #4
Look for evidence of adoption and outcomes rather than relying on product launches or AI announcements. Useful indicators include customer usage, renewals, implementation time, realized savings, quality outcomes, regulatory acceptance, and willingness to pay. Label each item as observed operating evidence, a management statement, a third-party estimate, or analyst inference. A forecast or stated target should not be presented as a realized gain.
Trace effects through revenue, costs, and investment
For each material workflow or offering, build scenarios around the main economic channels. State the mechanism, likely timing, evidence, and uncertainty; do not combine offsetting effects into a single unexamined “AI opportunity” or “AI risk.”
| Channel | What to examine |
|---|---|
| Demand or price pressure | Could customers need less of a vulnerable offering, obtain a similar result from an AI substitute, or negotiate a lower price? |
| New or expanded demand | Could an AI-enabled product serve a new use case, improve customer outcomes, or increase the market’s size? |
| Delivery costs and productivity | Could the company reduce labor or other delivery costs, and would those savings remain after review, support, and exception handling? |
| Investment and operating costs | What spending is needed for infrastructure, models, data, energy, components, integration, training, and ongoing operations? |
| Pricing power and margins | Can the company charge for the improvement, or will competitors and customers push the benefit into lower prices? |
Track the economics of the company’s own AI offerings separately from their existence. Paid customers and sustained usage matter more than a feature announcement; margins matter alongside revenue, because compute, training, inference, energy, support, and capacity costs can absorb gains.
Recommended Free Tools
Microsoft’s fiscal 2026 Form 10-K illustrates the kinds of disclosures to inspect, not a universal benchmark. It discusses competition from free applications and open-source products that may mimic features, with possible pressure on sales volume and prices. It also describes substantial AI infrastructure and operational investment ahead of fully developed revenue streams, uncertainty about adoption and demand, capacity utilization, training and inference costs, component and energy costs, and pricing pressure. These are company disclosures and management risk statements, not independent verification of projections or evidence that the same economics apply to other companies.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Assess whether the company can respond and capture value
After mapping exposure and adoption, assess whether the company can turn an AI-related change into a durable economic benefit. Consider whether it can access the data and distribution that matter, integrate AI into established products and workflows, and mobilize the technical and organizational capacity to implement and maintain the change. Weigh customer trust and applicable rules alongside execution capability.
Compare the value of the improvement with the cost of compute, labor, oversight, and capital. Ask whether the company can retain that value through pricing, customer retention, or lower costs—or whether customers, suppliers, and competitors are better positioned to capture it. Also test the reverse risk: the same tools may lower barriers for entrants or make the company’s product easier for customers to replace. An AI feature is not, by itself, evidence of a defensible advantage.
Compare companies on the same basis
When evaluating peers, use comparable reporting periods and the same questions for each company. Avoid ranking one company on announced products and another on realized results. A consistent comparison can use the following dimensions without pretending that they add up to a validated score:
Free tools Windows power users keep installed
One-click scans. No signup required.
| Dimension | Evidence to compare |
|---|---|
| Task and workflow exposure | Which company activities and customer workflows could change, and how material are they to the offering? |
| Substitutability | Can AI replace the customer outcome or let customers bypass the company, rather than merely assist an employee? |
| Adoption and willingness to pay | What usage, renewals, implementation results, and paid demand are observed? |
| Revenue and margin pressure | What evidence points to demand changes, price competition, cost savings, or new revenue? |
| Investment burden | What AI-related capital and operating costs are disclosed, and how do they compare with realized returns? |
| Defensibility and implementation | How do data, distribution, integration, switching costs, trust, and organizational capacity affect the response? |
| Governance and constraints | What regulatory, customer, workforce, or other impacts could limit adoption or require mitigation? |
Keep evidence quality visible beside each comparison: observed result, management-stated expectation, third-party estimate, or analyst inference. Do not collapse exposure, readiness, and resilience into one number unless the assumptions and limits of that score are explicit. Revisit the comparison when capabilities, customer behavior, company disclosures, or regulation change.
Include governance and implementation impacts
Business exposure is only part of the assessment. The OECD’s 2026 guidance for responsible AI due diligence sets out a six-step approach: embed responsible business conduct in management systems; identify and assess impacts; prevent and mitigate them; track implementation; communicate actions; and cooperate in remediation where appropriate. Applied to an AI assessment, this means checking who may be affected by the company’s adoption, how risks are handled, and whether the company can track and explain its response—not just whether the technology can be deployed.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
