Gartner recently shared a graphic on LinkedIn called “Start where AI creates the greatest value in HR,” and it does what the title says. It lays out six areas of HR work, from recruiting through employee operations, and rates each process by how much value AI can create there.
I spent some time with it, less for what it rates than for what it leaves out.
The Reward column has four boxes: rewards plan design, rewards communication, recognition, and well-being. Plan design happens before the compensation cycle. Communication happens after it. The cycle itself doesn’t have a box. That’s the part where the budget gets allocated, managers make individual calls, and those calls roll up through the organization for approval. Payroll and benefits administration shows up in a different column, further downstream.
The moment when money gets assigned to named people isn’t on the map.
Whether that was deliberate or not, I think the gap tells you something. I’ve spent four decades in HR technology, a good part of it embedded in comp planning systems, and this is the part of HR where I’d move most carefully.
Where AI earns its place in HR
Start with the part the AI vendors get right, because it’s real.
A lot of HR work is information processing. Records to keep current, time to track, the same benefits question answered over and over, learning content to produce. AI is good at that kind of work and getting better fast. If your team is still doing it all by hand, AI should take a lot of it off their plate.
Where the chart says AI does little
The chart is just as clear about where AI adds little. Gartner’s lowest rating means “AI has no or limited effects on process execution and decision making,” and few solutions are available yet. Four boxes carry it:
- Onboarding
- Culture
- Change management
- Employee relations and case management
None of those four is short on information. What they have in common is that they run on trust. Someone has to reassure, persuade, mediate, or make a call another person will have to live with.
So the pattern is fairly clean, at least for now. AI does well where the work is processing. Where the work runs on trust and judgment, it does little today, and the tools are thin.
Compensation planning is both
That’s why the comp cycle doesn’t sit comfortably on either side.
It’s about as structured as HR work gets: salary ranges, compa-ratios, budgets, eligibility rules, proration, approval chains. It’s dense with data and governed by rules, which looks a lot like the work AI handles well.
It’s also judgment-heavy, consequential, and contested. Every number it produces lands on a specific person. An employee who gets less than they expected will want to know why, and they’ll ask a person, not a model.
The structure is what makes AI useful around the decision. The consequence is what means a person has to own it.
Where AI helps inside the cycle
None of this means AI has no place in compensation. Used well, it makes the people who own the decisions faster and better informed. Some places it fits:
- Flagging outliers and anomalies in a manager’s plan before it’s submitted
- Surfacing possible equity gaps for a person to investigate
- Modeling budget scenarios and what-ifs
- Drafting a first pass at a manager’s rationale, for the manager to review and rewrite
- Summarizing where the cycle stands across a large organization
Every item on that list is assistive. It helps someone see more, or move faster, before they decide. None of them is the decision.
Where the human factor still carries the weight
The allocation itself, who gets what and why, is where I’d hold back, at least for now.
That’s not because the technology is weak. It’s because someone has to own the outcome and be able to explain it to the employee in plain words. A record of what the tool suggested matters, because it’s what lets anyone check that explanation later. But a record isn’t an explanation, and a model can’t sit across the table from the person whose increase it shaped.
The law is pointing the same way. My last article walked through what’s already on the books. Two points matter here:
- Colorado puts compensation in scope. Its new AI law takes effect January 1, 2027. If an AI tool materially influences an employee’s compensation, that employee is owed notice before the tool is used. After an adverse outcome, they’re owed a plain-language description of the decision and the role the technology played in it, within 30 days.
- The employer stays on the hook. Liability for discriminatory outcomes doesn’t depend on anyone intending them, and a vendor’s good faith doesn’t move it. In California, a vendor that helps make these decisions on your behalf can count as your agent, and the rules treat an agent as an employer.
The lawsuits so far are about hiring tools. The theories being tested there are the ones compensation tools will face next. And every one of those obligations assumes a person who can explain the number.
Meanwhile, compensation software vendors have started marketing AI “agents.” One now describes an AI compensation agent that calculates bonuses. The pitch is moving into the box the Gartner map left out.
Four questions before AI moves into your comp cycle
- Is the tool assisting the decision, or making it? If managers accept nearly everything it suggests, in practice it’s making it, whatever the contract calls it.
- Who owns each number? Name a person for every recommendation that becomes a compensation decision. “The system” isn’t an owner.
- Can that person explain a specific outcome to the employee without pointing at the model? If the honest answer to “why did I get this?” is “the tool recommended it,” nothing has been explained.
- Can the reasoning be checked afterward? You want a record of what the tool suggested, what the person decided, and why they differed. That record is necessary. It isn’t sufficient on its own.
The box nobody drew
The map is a useful guide to where AI pays off in HR. Use it that way. Just notice where it stops.
The comp cycle sits in the gap between plan design and communication, and it’s the part of HR where a person still has to stand behind every number. AI can help them get there faster and better prepared.
The cycle may earn its box someday. For now, the organizations that hand AI the allocation first are the ones who’ll find out how these new laws get applied. That’s a test most comp leaders shouldn’t volunteer for.
If a vendor tells you their AI can run the cycle, ask them where it sits on that map.
Sources:
- Gartner, “Start where AI creates the greatest value in HR,” graphic, CTMKT_5230645, © 2026 Gartner, Inc. and/or its affiliates. Shared on Gartner’s LinkedIn page, October 12, 2026, with the webinar “Strategic Roadmap: Redesigning HR for an AI-Infused Operating Model”, recorded July 16, 2026.
- Colorado SB 26-189, signed May 14, 2026, effective January 1, 2027. Ogletree, “Colorado’s New AI Act Targets Automated Decision-Making for Consequential Decisions”; Epstein Becker Green, “Inside Colorado’s Senate Bill 26-189: Impacts and Implications for Employers”.
- California Civil Rights Council automated decision system regulations, effective October 1, 2025. Duane Morris, “AI Catches Up: California Employers’ Regulations on Automated Decision Systems Now in Effect”.
- Dartican, “AI in Comp Is Moving Fast. The Law Is Not Far Behind.” (September 30, 2026), and its sources.
