Las Vegas, October 20-22, 2026. Mandalay Bay.
There is a difference between a conference that talks about the future and one that is already living in it, and HR Tech 2026 landed firmly in the second category.
The conversations on the floor at Mandalay Bay this week had a different texture than prior years. The question was never “is AI going to transform HR?” That debate ended somewhere around 2024. The question this year, asked in nearly every session, every breakout, every hallway debrief between practitioners who flew in from around the country, was considerably more difficult: “We ran the pilot. It worked. Now what?”
That gap between a successful AI pilot and a scaled, operational AI deployment turns out to be wider than most HR leaders expected, and more technically, culturally, and organizationally demanding than any vendor demo suggested it would be. HR Tech 2026, running under the theme “From Change to Transformation,” built its entire program architecture around the premise that the industry has reached the edge of the pilot era and needs frameworks for what comes next.
What the conference delivered was not a playbook, because no clean playbook exists yet. What it delivered was something more useful: a room full of practitioners who had already tried things, learned hard lessons, and were willing to talk about what actually happened. In an industry that produces more aspirational content than honest retrospectives, that candor was the most valuable product on offer.
The Opening Keynote That Set the Right Tone
Kelly Jones, Chief People Officer at Cisco, opened the conference Tuesday afternoon with a keynote that was notable for what it refused to do as much as for what it said.
She did not promise that AI would solve HR’s talent challenges. She did not present a transformation framework that conveniently aligned with a vendor’s product roadmap. What she did was walk through Cisco’s actual experience moving from AI pilots to large-scale execution across the people function, and the picture she painted was more complicated, more instructive, and more useful than the cleaned-up version of the story most executives present when they’re on a conference stage.
The core of her argument was about readiness, and specifically about the organizational readiness gap that sits between “we have an AI tool that works in testing” and “AI is improving decisions and outcomes across our HR operations.” At Cisco, the biggest barriers to scaling were not technical. The models were not the problem. The problem was data quality, change management, and the organizational habit of treating AI as a project rather than as a new operating mode.
Jones made a point that resonated in subsequent sessions throughout the day: the organizations that are successfully scaling AI in HR did their infrastructure work first. They cleaned data. They established governance. They built clear decision rights around which decisions AI informs versus which decisions AI makes. And they communicated relentlessly with the workforce about what AI was doing and what it was not doing. The organizations that skipped these steps are the ones with successful pilots that have not moved into production.
That framing, pilot success as a necessary but not sufficient condition for transformation, became a recurring theme in every substantive conversation at HR Tech this week.
The Science Behind the Human-AI Interface
If Kelly Jones set the operational frame for the conference, Dr. Tomas Chamorro-Premuzic, Chief Science Officer at Russell Reynolds Associates and one of the more rigorous voices in organizational psychology, provided the intellectual architecture on Day 2.
Chamorro-Premuzic’s work sits at the intersection of people analytics and the behavioral science of how humans interact with AI systems, and his keynote was the kind of talk that gets better the second time you read the transcript. His central argument is about a specific risk that HR leaders are not paying enough attention to: the algorithms that HR technology uses to assess, hire, promote, and develop employees are increasingly consequential, but the humans deploying those systems rarely understand them well enough to know when they are working correctly versus when they are optimizing for the wrong thing in a technically sophisticated way.
He distinguished between two categories of AI error in people decisions. The first is the error that is visible, the recommendation that is obviously wrong, the assessment score that does not track with any observable data about the candidate. Those errors get caught and corrected. The second category is the error that is invisible, the algorithm that is working exactly as specified but is specified against flawed assumptions about what good performance looks like, or what factors predict leadership potential, or what diversity of perspective actually contributes to team effectiveness. Those errors persist and compound because nobody flags them.
His prescription was not to avoid AI in people decisions. It was to invest in what he called calibrated human oversight: the organizational capability to understand what signals AI systems are using, to audit outputs against ground truth, and to build feedback loops that catch systematic errors before they become embedded in institutional practices. In a world where AI is making or influencing more hiring, compensation, and development decisions every quarter, the HR function that cannot evaluate the quality of its own AI outputs is flying blind with an autopilot it cannot verify.
This landed differently with different practitioners in the room, and the split was interesting. Leaders from organizations with strong data science and analytics functions generally felt they were already doing some version of calibrated oversight. Leaders from smaller organizations or functions without dedicated analytics capacity were more honest that they were largely trusting vendor claims about model validity. Both groups left with work to do.
The Microsoft View on What “Future-Ready” Actually Requires
Prerna Ajmera, General Manager of HR Digital Strategy and Innovation at Microsoft, closed the conference Thursday afternoon with a keynote she titled “The Future is Unbound: There is No Finish Line,” and the title is worth taking literally.
Ajmera’s core argument was that the organizations struggling most with AI transformation in HR are those that conceived of it as a project with a completion state. They set a target, deployed a technology, measured adoption, declared success, and stopped. The problem is that the technology keeps moving, the workforce keeps changing, and the competitive environment keeps shifting. An AI deployment that was genuinely leading-edge eighteen months ago may be table stakes today and underperforming in six months.
The mental model she was pushing against was HR technology as infrastructure build, something you plan, execute, and maintain. The mental model she was advocating for was HR technology as continuous learning system, something that gets better as it is used, that incorporates feedback from actual outcomes, and that requires a team with the capability and authority to keep improving it rather than just keeping it running.
At Microsoft, she described the internal HR function as operating something closer to a product development cadence than a traditional technology deployment model. New capabilities are piloted in small cohorts, evaluated against real outcomes, iterated, and then scaled when the evidence supports scaling. The team that builds this capability includes HR practitioners, data scientists, and technology specialists working in the same room against shared goals, not in separate functions aligned by quarterly business reviews.
The audience response to this framing was mixed in the most informative way. The HR leaders who found it energizing were those who already had organizational air cover to experiment and iterate. The HR leaders who found it uncomfortable were those operating in organizations where HR technology decisions go through five-layer approval processes and take eighteen months from conception to deployment. For that second group, the real challenge is not figuring out what good AI-enabled HR looks like. It is negotiating the organizational conditions under which continuous improvement becomes possible.
From the Floor: Six Things Practitioners Were Actually Talking About
The expo floor at HR Tech 2026 opened immediately after the opening keynote and ran through all three days, and the conversations happening between sessions were frequently more substantive than the formal agenda.
Several threads emerged as consistent across dozens of conversations with practitioners:
The skills architecture problem is getting more urgent, not less. Companies that started building skills inventories two years ago are discovering that the skills that matter are changing faster than their inventories can track. Several practitioners from manufacturing and healthcare were discussing this as a genuine crisis: their skills data is already outdated by the time it is collected, and the AI tools they bought to match talent to roles are working off stale inputs.
Workforce readiness for AI is uneven in ways that create real management complexity. Within the same organization, some teams have adopted AI tools aggressively and self-reported significant productivity gains, while other teams in the same building have low adoption rates and are skeptical about what the tools actually do. The organizations navigating this best are the ones that treat the adoption gap as a management challenge rather than a technology challenge.
The governance question is more complicated than most vendors acknowledged during the sales cycle. Specifically: when an AI system makes a recommendation about a promotion decision, or a performance rating, or a skills gap assessment, who is accountable for that recommendation? The legal and compliance functions are asking this question in ways that HR technology teams are not always equipped to answer.
Data quality remains the biggest barrier to AI deployment that practitioners did not anticipate when they started. Multiple sessions returned to the Sanofi example presented in the formal program: a two-year HR transformation that had to sequence data cleanup and governance before AI experimentation, because the AI tools simply did not produce useful outputs on messy data. This is obvious in retrospect. It was not obvious in the sales conversation.
The frontline workforce is being underserved by most HR AI deployments. Most of the products on the expo floor are optimized for knowledge workers. UKG’s sessions on CHRO strategy for frontline workers addressed a gap that is genuinely large: the roughly 50% of the workforce that does not sit at a desk is not well-served by the tools that most HR technology vendors are building.
Return on investment conversations are getting more sophisticated, but not fast enough. Most organizations are measuring AI ROI in HR through process efficiency metrics, time-to-fill, cost-per-hire, time spent on administrative tasks. A smaller number are trying to measure downstream workforce outcomes: retention, performance improvement, internal mobility rates. The gap between those two measurement approaches is wide, and the organizations measuring downstream outcomes consistently find different results than those measuring process efficiency.
What the Agenda’s Architecture Tells You
The six focus areas that organized the HR Tech 2026 program are worth reading as a diagnostic of where the field thinks the important problems actually are.
AI and automation in people strategy appears first, which reflects both its centrality and the degree to which it cuts across every other focus area. But the sessions under this heading were notably different from prior years. Where 2024 and 2025 sessions asked what AI could do in HR, the 2026 sessions assumed AI is present and asked how to make it work reliably at scale.
The CHRO leadership track reflected the growing recognition that the Chief People Officer role is evolving in ways that most CHRO development programs have not caught up with. The people in those seats are increasingly expected to have working fluency in data systems, AI governance, and organizational change management in addition to the traditional HR expertise. Kelly Jones at Cisco is an example of what this looks like when it works. The track was honest about the fact that it is genuinely difficult to find and develop people who combine those capabilities.
The skills and learning track was the most crowded of the six, which tells you something about where practitioner anxiety is concentrated. The fundamental challenge is that the skills required to operate effectively in AI-augmented workplaces are changing faster than most organizations’ learning infrastructure can accommodate. Several sessions were explicitly about the organizational design question: should skills development be a centralized HR function, a distributed team responsibility, or some configuration between the two?
Key Takeaways
The era of AI pilot euphoria in HR is over. The organizations still announcing successful pilots as proof points are behind the curve. The relevant metric now is whether a pilot scaled to production, whether it is producing measurable workforce outcomes, and whether the organization has the capability to keep improving it. HR Tech 2026 was built for the organizations trying to answer those questions honestly.
Data quality is the unglamorous prerequisite that nobody put in the conference session title but that kept appearing in every substantive conversation. AI in HR does not work on bad data. It produces confident wrong answers on bad data, which is worse than no answer. Organizations that invest in data infrastructure before AI tools will outperform those that do the reverse.
Calibrated human oversight of AI people decisions is a capability that most HR functions do not yet have. Chamorro-Premuzic’s framing of invisible algorithmic errors is the framework HR leaders should be using to evaluate their own AI deployments. If you cannot explain what your AI recommendation system is optimizing for and verify that the outputs track with real performance outcomes, you have a governance problem.
The frontline workforce gap is real. Most HR technology built for AI augmentation was designed for knowledge workers. The half of the workforce that does not sit at a desk needs different tools, different deployment models, and different measurement frameworks. This is a market gap that is starting to get attention but has not yet been adequately addressed.
Continuous improvement capability is the differentiating factor among organizations that are successfully scaling AI in HR. The organizations treating AI deployment as a one-time infrastructure project will fall behind the organizations that have built the product development operating model to keep iterating. The build is not the point. The loop is the point.
Notes for HR Technology Leaders
If your organization has AI pilots in HR that have not scaled to production in the past twelve months, the HR Tech 2026 sessions from IFF, Cloudflare, and Sanofi are worth reviewing for the specific diagnostic frameworks each used to identify what was blocking scale. The Cloudflare session in particular, titled “Why HR AI Pilots Stall and How to Move Them into Production,” is structured around a set of diagnostic questions that most organizations can run against their own situation.
The Pitchfest and Startup Pavilion on the expo floor surfaced several vendor approaches to the skills tracking problem that are meaningfully different from the legacy competency model infrastructure that most organizations currently operate on. If skills currency is a real challenge in your organization, the new class of dynamic skills intelligence platforms is worth a serious look.
On governance: if your organization does not have a documented process for auditing AI recommendations in HR decisions against actual outcomes, building one should be on the agenda before deploying additional AI tools. The compliance and legal exposure from ungoverned AI in hiring and performance decisions is growing.
HR Tech 2026 ran October 20 through 22 at Mandalay Bay Las Vegas. Session recordings and on-demand content are available through the conference platform at hrtechnologyconference.com. The next HR Tech Conference is scheduled for fall 2027.
