What 258 HR Leaders Told us About AI in Talent Decisions

Two years ago, the conversation about AI in HR was about whether to use it. In our Q1 2026 research of 250+ senior HR leaders at organizations with 1,000 or more employees in North America and Europe, the conversation has moved on. 48% of respondents are exploring or piloting AI in skills and career processes, another 31% are running it operationally or at scale, and 57% feel more confident about expanding AI than they did two years ago.

What changed is the specificity of what HR leaders now ask for. Two years of deployments produced enough evidence for HR teams to write their own procurement spec, with explicit requirements for what an AI system must do before it goes near a talent decision.

The spec HR has written for itself

Three requirements sit at the top of the research data.

• 38% of respondents named bias testing or fairness audits before deployment as a required capability, the single most-cited requirement in the research.

• 35.7% require explainable logic that lets managers and employees see why a specific recommendation was made.

• 34.1% require human oversight at the decision points where AI is involved.

Below those, ongoing bias monitoring after deployment (32.9%), legal and compliance sign-off (28.7%), audit trails for AI-influenced decisions (28.3%), and documented decision rules (26.7%) round out the core list.

These are the questions HR is now bringing into vendor demos.

• Has the system been bias-tested before deployment, and can you show the results?

• Can a manager open any specific recommendation and read the logic behind it?

• Where are the human checkpoints in the workflow?

• How is bias monitored on an ongoing basis once the system is live?

• What does the audit trail look like for a recommendation that gets challenged?

• What is the process when an employee disagrees with what the system surfaces?

63% of respondents said AI should support and inform human decisions with logic that can be inspected and reviewed, and only 17% wanted AI to generate recommendations autonomously. HR buyers have moved past the autonomous-by-default vision of 2024 to a more specific set of requirements, with explicit checkpoints, explicit explanations, and explicit ways to challenge what the system produces.

The accelerators built governance in from day one

• 21% of respondents are accelerating their AI use and investment heading into 2027.

• 22.5% report that governance was designed into their AI systems before deployment. 

These two groups overlap meaningfully. The organizations that built bias monitoring, explainability, and audit trails in from day one have systems that work under review, and that gives them room to expand AI into more of the talent process.

A larger group of 29% is proceeding cautiously with a governance-first approach. They are putting governance frameworks in place now, often onto live systems, and expanding more slowly while they do that work. The 25% of respondents who paused an AI initiative in the past 24 months sit largely inside this same group.

Most of the shelved deployments were ones built fast, before the organization had a way to explain a recommendation to a regulator, a works council, or an employee asking how the system reached its conclusion.

The AI systems still in production are the ones that were built to be defended from the start. These systems have bias testing before launch, ongoing bias monitoring once live, explainable recommendation logic, human decision points, audit trails, and a defined path for employees who want to challenge a recommendation. Each of these is now a default requirement that has to be in place before deployment.

Fuel50 was built this way. Our skills ontology is maintained by I/O psychologists who are accountable for what goes into it, and the AI is bias-tested before any deployment. Every recommendation comes with logic a manager or employee can read, and employees have a defined route to flag or challenge what they see.

Adoption follows when the AI is connected to real skills and real opportunities

• 36.4% of respondents named ease of access through familiar systems as a top driver of employee adoption of AI tools.

• Another 36.4% named integration with daily tools like Teams, Slack, and email.

• 34.5% named manager endorsement and active participation. 27.1% named a clear connection between recommendations and real opportunities.

• Among respondents whose organizations have deployed AI tools, 34% reported that the tools reduce friction in talent decisions, with the strongest gains in organizations that integrated AI into existing manager and employee workflows.

These answers point in the same direction. AI delivers value when it works inside the systems people already use and when the recommendations it makes lead to something the employee can act on. A career tool that does not know what skills an employee already has, what roles are open, or what learning will close the gap will produce generic suggestions that get dismissed. Built on a full skills inventory and connected to internal mobility, learning, and gigs, the same tool can move career conversations to specific next steps. Those steps are tied to roles the company is actually trying to fill.

54% of managers in the research said they feel moderately, very, or overwhelmingly burdened by their role in delivering career guidance. The AI tools that take some of that load off live inside the systems managers already use, run on accurate skills data, and link their recommendations to opportunities the company can actually offer.

What the next 12 months look like

Asked about their posture toward AI in talent decisions going into 2027, respondents distribute into four groups. 29% are proceeding cautiously with a governance-first approach. 21% are accelerating their AI use and investment. 19% are holding steady at current levels, and 7% are reassessing based on results so far. 23% have no plans to use AI in these decisions.

The 21% accelerating offer the clearest read on what is working. They deployed AI early, learned what worked, and now have the governance and skills data infrastructure to scale further across the talent process. The procurement spec they helped establish, which includes bias-tested AI, explainable recommendations, human oversight at decision points, audit trails, and a defined appeal process for employees, is becoming the market norm. These requirements now appear as table-stakes line items in vendor evaluations across industries and company sizes.

The full Q1 2026 report, The State of AI Readiness in Talent Decisions, includes the cross-cuts by industry, company size, and HR role, the complete data on which AI capabilities are getting funded in the next 12 months, what HR leaders are willing to pay for, and what would most increase their willingness to expand AI in talent decisions. You can download it here.