Problem
For decades, the HR function inside professional services firms has been a people-and-paper operation: a partner sponsor, an intake form, a calendar invite, and a human generalist shepherding the request through policy. That model is breaking. Small and mid-sized law firms, accounting practices, and consultancies that added headcount to ride growth waves from 2020 through 2024 now find their people costs growing faster than revenue, while the actual work of HR (answering the same PTO question for the eight-hundredth time, screening the same stack of resumes, reconciling the same benefits data across carriers) crowds out the strategic work partners actually need.
Why It Matters
Gartner’s 2026 HR Priorities Survey, drawn from 426 CHROs across 23 industries and four global regions, identified AI as the single largest spend line in HR and the priority expected to deliver the highest productivity impact: roughly 29 percent of all AI-driven productivity gains in the coming year will come from an HR-led redesign, not from engineering or finance. That is a striking number, because it puts the HR function on the same strategic footing as the rest of the C-suite for the first time in a generation.
Deloitte’s parallel research underscores the demand side. Its 2025 Global Human Capital Trends report, polling nearly 10,000 leaders across 93 countries, found that 6 in 10 workers already treat AI as a coworker rather than a tool. The 2026 edition, surveying roughly 9,000 leaders across 89 countries, reports that 7 in 10 business leaders now identify speed and nimbleness as their primary three-year competitive strategy. For a people-heavy professional services firm, those two facts collide: the workforce expects AI to be a teammate, and the firm expects the workforce to move faster. HR is the function that has to reconcile that collision without breaking trust.
The cost of getting the transition wrong is not abstract. The Klarna episode of 2024 to 2025 is the cautionary tale of the year: a fintech widely celebrated as “AI-first” for replacing customer-service headcount with AI agents had to publicly reverse course by mid-2025 and rehire human staff after measurable declines in service quality and customer satisfaction. The same pattern is showing up inside HR functions that have automated too aggressively, losing institutional knowledge, damaging the employee experience, and discovering that “AI-first” and “human-last” are not the same thing. Conversely, firms that simply bolt AI onto existing processes, keeping the same headcount, the same org chart, the same approval chains, are finding that they have added cost without gaining speed.
The AI Approach
An AI-first HR operating model is not “HR with AI tools sprinkled on top.” It is a deliberate redesign of how the function delivers value, with four layers that build on one another.
1. The service layer: conversational HR. The first job is to remove the repetitive work that consumes 60 to 80 percent of HR generalist time. AI agents, internally branded, governed, and trained on the firm’s own policies, handle tier-one questions on PTO, benefits, payroll, onboarding paperwork, and policy lookup. The benchmark here is IBM’s AskHR, which now resolves roughly 94 percent of routine HR inquiries and handles around 2.1 million conversations a year, cutting human HR support volume by about 75 percent. A 40-person professional services firm does not need IBM’s scale, but it can absolutely deploy a similar agent to free its HR lead from being the in-house answers desk.
2. The decision layer: AI-augmented people decisions. Resume screening, candidate ranking, skills inference, compensation benchmarking, and attrition risk modeling all benefit from AI that is grounded in the firm’s own data. The key word is augmented: AI produces a shortlist, a flag, or a recommendation, but a human partner or HR lead makes the call. This layer is where 2026’s regulatory scrutiny is sharpest, particularly in the EU AI Act context, so auditable models and human-in-the-loop checkpoints are non-negotiable.
3. The workforce-design layer: redesigning roles for human-AI collaboration. This is the hardest layer and the one most often skipped. It asks: which tasks in each role are automatable today, which augment well, and which require human judgment that creates client value? The answer reshapes job descriptions, hiring profiles, and career paths. A paralegal whose document review is 70 percent automated is no longer a document reviewer; they are a matter-strategy associate. An accountant whose book-close is automated is an advisory consultant. Reclassifying these roles, and paying for the higher-value work, is where the AI-first promise actually converts into margin.
4. The governance layer: a small, empowered center of excellence. None of this works without a dedicated owner. Gartner’s 2026 research calls out that roughly 24 percent of CHROs are now directly funding AI technologies inside HR, up sharply from prior years, and the fastest movers are standing up a tight HR-AI center of excellence with explicit authority over tooling, data, policy, and change management. For a smaller firm, this may be a single HR-ops manager plus a part-time data steward. The point is that “everyone owns AI” means nobody owns AI.
The through-line: AI-first is not the elimination of HR. It is the elimination of HR’s transactional drag, redirected toward the strategic work, workforce planning, culture, leadership development, change management, that partners and clients actually pay for.
Real-World Examples
IBM’s AskHR: the gold standard for tier-one automation. IBM deployed its watsonx-powered AskHR assistant internally and now reports that it handles about 2.1 million HR conversations a year, automating roughly 94 percent of routine inquiries and reducing human HR support load by approximately 75 percent. The freed capacity was reinvested into higher-value work like talent strategy and workforce redesign, not eliminated. Crucially, IBM kept a clear escalation path to human HR for the 6 percent of cases that genuinely needed it.
Klarna: the canonical “AI-first” reversal. In 2024, the buy-now-pay-later company became the press’s favorite example of an AI-first organization, replacing customer-service contractors with OpenAI-powered agents. By May 2025, CEO Sebastian Siemiatkowski publicly told Bloomberg the company was hiring humans again because AI-only service produced “lower quality” and damaged the customer experience. Klarna did not abandon AI, it still uses it heavily, but it learned that headcount reduction is not the same as operating-model redesign, and that quality, trust, and brand can erode faster than they can be measured.
Professional services catching up. Thomson Reuters’ December 2025 strategic-innovation report found that law firms with a published, visible AI strategy are about twice as likely to report revenue growth than peers without one. The 2026 ABA Legal Industry Report tracked a meaningful jump in AI adoption inside law firms over the prior year. Early-mover firms, typically mid-sized Am Law 200 and Top 100 accounting practices, are using AI for document review, due-diligence summarization, time-entry automation, and client-intelligence briefings, with HR functions following the same playbook internally: a service-layer agent first, decision-layer tools second, role redesign third.
Action Steps
- Audit one quarter of HR transactions. Pick a 90-day window and log every request. You will find that 60 to 80 percent of volume is the same ten questions. That list is your AI-agent roadmap.
- Ship one tier-one use case in 60 days. Stand up an internal AI agent for the single most common HR question, measure deflection rate and satisfaction, iterate. Speed of learning beats scope.
- Reclassify three roles. Rewrite the job description for the human-plus-AI version. Update compensation and hiring profile. This is where AI-first converts into margin.
- Name an owner. A single HR-AI owner with authority, budget, and a quarterly partner report. “AI is everyone’s job” is a recipe for nobody’s job.
- Write the human-AI principles. One page: where AI decides, where humans decide, where both decide together. Your governance floor and trust foundation.
Call to Action
If your firm is hiring faster than it can train, or training faster than it can bill, the bottleneck is not the labor market. It is the operating model. Schedule a free 30-minute consulting call and we will map where implementing AI can buy you back the hours you need to keep pace with growth, and where moving too fast without a redesign will hurt you.