Table of Contents
Worldwide AI spending will reach $2.5 trillion in 2026, a 44% jump from 2025, even as Gartner research finds that only one in 50 AI investments delivers transformational value and only one in five produces any measurable return on investment. The volume of money pouring into AI tools has decoupled from the volume of value coming out.
The mismatch is now visible inside HR itself. By mid-2025, 61% of HR leaders were implementing AI according to Gartner’s global benchmarks, up from just 19% two years earlier. Yet i4cp data shows 68% of CHROs identify AI-related workforce initiatives as a top priority, while only 12% of large organizations report mature, sustained investment in the skills-based practices AI is supposed to enable. Adoption has raced ahead of execution.
Former BlackRock HR head Jeff Smith, who spent more than a decade leading human resources at the global asset manager after senior roles at Time Warner and AOL, has been warning HR leaders against vendor-led decision-making for years. “It is critical to have exceptional technology to make processes better and more efficient, for risk management, and to help provide data and insight to make decisions,” he says, then names the constraint that most CHROs underweight. “The technology landscape has never been more complicated, and the constant emergence of new vendors makes it more so. So it is not enough to just go with one of the established players without evaluation.”
The Hype Cycle Catches Up to HR
Gartner places agentic AI at the Peak of Inflated Expectations on its 2026 Hype Cycle, with only 17% of organizations having deployed AI agents to date, but more than 60% expecting to within two years. The pattern matches every prior technology wave HR has weathered, including the rush to applicant tracking systems in the early 2000s and the scramble for engagement platforms in the 2010s.
What separates this cycle is the gap between investment confidence and outcome confidence. A Gartner survey published in April 2026 found that only 39% of technology leaders believe their current AI efforts will improve financial performance. Organizations that achieve meaningful AI outcomes invest up to four times more in foundational areas like data quality, governance, talent, and change management than those that do not.
The implication for HR is uncomfortable. Buying tools is the easy part of the cycle, while building the foundations that let those tools deliver is what most companies skip.
Why Default Vendor Selection Backfires
Most HR technology decisions follow a predictable shortcut. A board or CEO asks what HR is doing about AI. The CHRO calls the firm’s biggest existing vendor. The vendor pitches its newest module. The contract is signed before any rigorous comparison, and implementation begins before the underlying problem has been clearly defined.
Jeff Smith treats this pattern as the central error of HR tech procurement. Defaulting to established players sidesteps the diagnostic work that determines whether the chosen tool will actually solve the company’s specific problem, and it locks in switching costs that make later course corrections expensive. The vendor benefits from incumbency. The HR function inherits the gap between what was sold and what gets delivered.
i4cp founder Kevin Oakes describes the dynamic in similar terms. “Whenever a disruptive technology enters the mainstream, there’s immense pressure to act quickly, often amplified by investor expectations, media narratives, and competitive anxiety,” he notes. “Effective leadership during these periods isn’t about keeping up with the hype; it’s about creating a future-ready organization that focuses on culture, mindset, and experimentation.”
The shortcut is rarely a single decision. It is the accumulation of small choices made under pressure to appear current, with each one foreclosing alternatives the company never properly considered.
BlackRock and Time Warner Alum Jeff Smith on Building a Selection Discipline
Smith’s framework for HR tech procurement starts with a question most companies skip: What specific problem are we trying to solve, and how would we know if it were solved?
Define the problem before the platform
Tools follow problems, not the other way around. “I think getting the basics right and executing them is far more important before you are innovating,” Jeff Smith argues. “Pay people right, have great hiring practices, develop your leaders, have a culture of feedback, and ensure leaders know their expectations. Have good, solid processes, then innovate on top of that where it is important to the business and where it is going to work because there is a foundation to innovate on top of.” Companies that buy AI for performance management while their feedback culture is broken are layering technology on top of the dysfunction. The investment magnifies what was already there rather than fixing it.
Evaluate context, not category leadership
Vendor rankings describe what works on average across hundreds of companies, not what fits a specific company’s data, workflow, regulatory environment, or culture. Jeff Smith insists on context-specific evaluation, including the practical question of whether the tool will work for the real users in the real environment rather than the demo conditions sales teams stage. This is where most procurement processes get short-circuited, because rigorous evaluation takes months and the pressure to act is measured in quarters.
The Six-Question Evaluation Filter
Companies that consistently extract value from HR technology investments tend to run candidate tools through a similar set of questions before signing anything:
- Problem fit: Does this tool address a problem we have actually defined and quantified, or are we backing into the problem after seeing the solution?
- Foundation readiness: Do we have the data quality, process clarity, and skills baseline this tool requires, or are we expecting the platform to compensate for foundations we have not built?
- Implementation honesty: Do we have the change management, training, and managerial support time the deployment will demand, or are we assuming adoption will happen on its own?
- Vendor durability: Will this vendor exist in three to five years with the roadmap we are buying into, or are we betting on a category leader whose moat looks weaker on closer inspection?
- Switching cost: What does the exit ramp look like if this tool fails to deliver, and how much of our workflow, data, and integration architecture will be locked in?
- Human consequences: What does this tool do to the manager-employee relationship, to candidate experience, to the day-to-day texture of work, and is that change one we want?
The questions are not original to any one framework. The discipline lies in actually asking them before the contract gets signed, with answers documented and revisited as the deployment unfolds.
Where AI Helps and Where It Hurts
Jeff Smith’s view on AI in HR is consistently pragmatic. “AI should augment human abilities,” he says. The framing matters because it draws a line between use cases where automation creates value and those where it destroys it.
Routine analysis, pattern recognition, scheduling, document drafting, candidate screening, and large-scale data interpretation are domains where AI demonstrably reduces effort and improves consistency. Companies deploying AI thoughtfully in those zones report meaningful productivity gains and free their HR teams to spend more time on the work only humans can do.
The opposite pattern emerges when AI is applied to judgment-heavy work. Performance assessment, succession decisions, compensation calibration, conflict resolution, and culture-shaping conversations all depend on context, relationship, and ethical judgment that current systems cannot replicate reliably. Companies that automate those domains tend to discover the limits the hard way, through bad decisions delivered at scale before anyone catches the error.
Gartner has named one consequence of indiscriminate AI adoption “workslop,” meaning the abundance of fast but poor-quality output produced by or with AI when employees are pressured to apply it everywhere without time to assess whether the output is fit for purpose. The cost of work slowdown falls on the people downstream of the automation, who spend their time correcting outputs that were supposed to save them effort.
What Governance Looks Like in Practice
Technology governance is the discipline most HR functions underbuild. The standard pattern is to procure first and govern second, with policies written after deployment to address problems that have already surfaced. Companies treating governance as a procurement input rather than a procurement output produce different results.
Strong HR tech governance covers four areas: data privacy and consent, model transparency and explainability, escalation paths when systems produce harmful or biased outputs, and ongoing monitoring of whether the tool is delivering the value its business case promised. Each area deserves its own owner, its own review cadence, and its own visibility to the executive team.
Smith treats governance as an extension of the people-first principle, not a separate compliance exercise. The question is not whether the tool is technically permitted to do something, but whether deploying it that way protects the trust between the company and its workforce. The protection is what makes future technology adoption easier rather than harder, because employees who have seen governance work are more willing to engage with the next round of tools.
The Discipline That Compounds
The companies that emerge from the 2026 AI cycle in stronger HR positions will share a pattern: they refused to treat technology adoption as a status signal. They defined problems carefully, evaluated tools against context, deployed with realistic implementation budgets, and governed outcomes deliberately. None of those steps requires special insight. They require restraint at moments when restraint feels expensive.
Restraint is what most companies fail to muster, and the failure compounds. A poorly chosen tool consumes change capacity that the next deployment needs. An unaddressed governance gap erodes employee trust that the next initiative depends on. A vendor lock-in narrows the options three years out, when the right answer is no longer available without a costly migration.
Smith’s view comes back to the same point that anchors his approach to every HR question: technology serves people, not the other way around. “AI should augment human abilities,” he says. The companies internalizing that principle build technology stacks that get more useful over time. The companies treating it as a slogan watch their tools accumulate without producing the outcomes the original investment promised, while the gap between AI spending and AI value continues to widen across the broader market.