HR software has acquired a lot of sparkle icons. The harder question is whether those AI features actually change the work HR teams do.
AI adoption is rising. AI value is not settled.
SHRM’s 2026 State of AI in HR found that 39% of organizations currently have AI adopted in their HR functions, while another 7% intend to launch AI in HR during 2026.
The same study found that 56% of HR functions do not formally measure the success of their AI investments, and only 16% use their own ROI metric.
That is the tension worth paying attention to. It is getting easier to add AI to HR. It is still surprisingly difficult to prove what the AI changed.
The useful AI test: does it change a workflow?
A generated job description is useful. A summary of a long HR case can save time. A policy assistant can answer a routine question. But those are individual steps.
A stronger test is end-to-end: can the system identify a need, use the right workforce data, recommend or complete the next step, and escalate an exception to a person? That is where AI starts changing the operating model rather than just speeding up typing.
HR needs a seat at the AI table
SHRM found that 52% of organizations do not involve HR directly or through cross-functional collaboration in developing their overall AI strategy and vision.
That should worry HR leaders. AI changes jobs, skills, performance expectations, access to employee data and the employee experience. HR does not have to own every AI program, but it should not discover the people implications after deployment.
Three useful levels of AI maturity
- Assist: AI drafts, summarizes, searches and recommends. Most organizations start here.
- Orchestrate: AI works across structured HR data and workflows, helping route requests, surface exceptions, match candidates, or support managers.
- Act with guardrails: AI can execute bounded actions inside approved workflows, with permissions, auditability and clear escalation.
Not every HR process needs the third level. In fact, most should not start there. The sensible approach is to begin with a narrow process where both the value and the failure mode can be measured.
One more governance number worth knowing
SHRM found that about half of organizations that currently use or are about to pilot AI have workforce AI-use policies, but only about a quarter of those organizations feel their policies are clear and future-proof.
A policy is not governance by itself. Governance starts when the organization can answer who can use the system, what it can access, what it can do, who reviews exceptions, and how success is measured.
What this means for RightlyHR
RightlyHR describes native AI as part of a broader platform built around connected HR processes, real-time insights and configurable workflows. That is the right architectural direction: AI works best when employee data, organizational context and workflow rules are already connected.
Five questions to ask any HR AI vendor
- What workforce data does the AI actually use?
- Can it act, or does it only recommend?
- What happens when the underlying data is incomplete?
- How are permissions and approvals enforced?
- How will we know whether the capability created value?
