HR technology trends 2026: conversational video enters the stack

Every HR leader knows which conversations matter most: the onboarding call where a new hire decides whether they made the right choice, the coaching session where a struggling rep finally hears feedback that lands, the screening conversation where a great candidate either leans in or drops out.

Those are face-to-face moments. Most organizations still can't deliver them consistently at scale. HR technology has spent two decades automating the work around those conversations: scheduling, tracking, scoring, filing. The conversation itself still depends on having a human available at the right time, in the right language, and with the right context. When that person isn't available, organizations fall back to text-based interfaces, interactive voice response trees, or nothing at all. Organizations now have a new way to deliver presence in those moments.

Across recruiting, learning and development (L&D), and employee experience, human computing is entering the HR technology stack through real-time conversational video. AI humans, full-stack AI entities that see, hear, understand, and respond in live video conversations, are giving HR teams a way to deliver presence at scale without trading off quality for reach.

The HR technology stack in 2026

HR buyers are paying closer attention to the total cost of ownership than they did a few years ago. Global HR technology investment in the first half of 2025 reached $3.55 billion across 119 deals, a 60% increase over the same period in 2024.   

Only 43% of HR professionals and executives rate their organization's HR technology as effective, per SHRM's 2025 State of the Workplace data. One-quarter of HR teams now name updating their technology infrastructure as a 2026 priority, a 10-point year-over-year increase that is among the largest recorded in the Lattice 2026 State of People Strategy Report. Point systems are losing ground as buyers look for integrated platforms that consolidate spend and data. Consolidation cuts tool sprawl, and most human capital management (HCM) systems still center on workflow and record-keeping. 

Agentic AI moves from pilot to production

Agentic AI, systems that pursue multistep, adaptive goals with limited human oversight, is a dominant narrative in HR technology for 2026. These systems reason and act with a considerable degree of autonomy, handling multi-step tasks rather than just one.

Predictions show that by 2026, 40% of enterprise applications will use task-specific AI agents, up from less than 5% in 2025. Execution still lags behind the ambition. While some organizations report scaling an agentic system in at least one function, far fewer are scaling agents broadly across business functions. 

Over 40% of agentic AI projects will be canceled by the end of 2027. Governance is the bottleneck. Many organizations still lack full security approval for their agent fleets and formal AI governance policies. Much of the implementation work sits in data integration, validation, monitoring, and organizational change.

For HR leaders, agent deployment usually starts with governance and workflow redesign. The procurement decision comes after that. Governance is one half of the readiness question. The other is whether the workforce itself is mapped in a way that AI can act on.

Skills-based talent strategy replaces role-based planning

Workforce planning is shifting toward capabilities, location, and development. Headcount by role no longer provides HR teams with sufficient resolution.

Skill gaps remain a major barrier to business change, and workers can expect a meaningful share of their existing skill sets to be reshaped or rendered obsolete by 2030. 55% more organizations now map skills directly to jobs, up from 47% in 2023.

Sequencing matters in these shifts. Taxonomy restructuring can precede and support skills-based hiring, giving the rest of the program a stable foundation to build on. As these systems take on more of the decisions that affect a person's job, the rules governing them tighten.

Data privacy and responsible AI become procurement criteria

Regulatory pressure on AI in employment decisions is accelerating across multiple jurisdictions. The EU AI Act's full high-risk obligations take effect on August 2, 2026, classifying AI systems used in recruitment, candidate evaluation, performance monitoring, and termination decisions as high-risk under Annex III of the regulation.

Employers using vendor-supplied AI are classified as deployers when they use high-risk AI systems, with obligations that include human oversight, input data monitoring and record-keeping/logging under Article 26 deployer duties.

In the United States, NYC Local Law 144 requires an independent bias audit conducted no more than one year before a tool is used and public disclosure of the results. Illinois adds a biometric dimension: the Biometric Information Privacy Act governs the collection of facial scans and voiceprints, directly applicable to AI video systems analyzing facial geometry or voice characteristics. For HR teams evaluating any technology that processes video or voice data, compliance planning for biometric data collection and consent workflows is a procurement prerequisite.

The candidate and employee experience gap

Spend, autonomy, skills, and compliance describe the market. None of them explains why so many HR interactions still leave people feeling unseen. The technology problem in HR is often a shortage of presence: the feeling that someone is genuinely paying attention. Employee trust in company-provided generative AI fell 31% between May and July 2025. Trust in agentic AI systems that act independently dropped 89% over the same two-month period.

That pattern shows up across the employee lifecycle. Application abandonment rates can be high, and candidate ghosting remains a meaningful burden for recruiting teams. Onboarding research found that the most valued aspect of onboarding for employees was the human element: meeting people, forming social ties, learning from colleagues. Text-based interfaces handle information and process guidance. Coaching, retention, and skills development usually depend on a richer interface.

Conversational video as the next interface layer

A richer interface is exactly what the learning science points to. Research on interactive video has explored its potential for learning, but findings vary across studies and instructional contexts. Non-interactive video did not significantly improve learning outcomes over no video at all. Interactivity drives the learning advantage more than the presence of a face on screen.

Real-time conversational video involves a system that perceives the person, adapts, and responds during the conversation. Non-interactive video is a broadcast medium. Conversational video is a face-to-face interaction where the AI perceives the person it's speaking with, responds to their emotional state, and adapts in real time.

Tavus is the human computing company building full-stack AI humans that see, hear, understand, and respond in real-time conversations. Within an HR context, its Conversational Video Interface (CVI) gives HR teams a way to deploy AI humans across recruiting, onboarding, and L&D without building real-time conversational video capabilities from scratch.

These AI humans operate through a closed-loop behavioral stack. A conversational flow model governs when the AI human speaks and when it waits. Sparrow-1 predicts who owns the conversational floor and holds it open, giving the trainee time to formulate a follow-up question.

A large language model (LLM) layer reasons about what to say next. A multimodal perception system fuses the other person's tone, expression, and hesitation into a unified understanding of their state. When Raven-1 fuses the trainee's halting tone with a furrowed brow, catching a mismatch between verbal confirmation and visible confusion, the LLM adjusts the explanation's complexity.

An insurance company can deploy the AI human for compliance training. The AI human walks a trainee through a simulated claims explanation, grounded in the company's actual policies through Knowledge Base retrieval. A real-time facial behavior engine renders responsive facial behavior, including active listening cues like nodding and micro-expressions. Phoenix-4 renders an attentive, patient expression throughout.

Objectives and Guardrails define conversation boundaries, completion criteria, and compliance rails before the session closes. Guardrails prevent the AI human from offering coverage advice outside its defined scope, triggering escalation to a human specialist if the conversation drifts into territory that requires licensed judgment.

When the same trainee returns for a second session two days later, the personality layer's memory and evolution system retains what they struggled with previously. It picks up the onboarding journey without repeating material that has already been mastered.

Often, the alternative to this AI human is a non-interactive learning management system module that may show strong completion metrics but weak knowledge retention, or a one-hour webinar that trainees may only half-watch while multitasking.

Building the human-centric HR stack

Adding presence to the stack only pays off if people remain at the center of its deployment. Organizations taking a technology-first approach to AI are 1.6 times more likely than those taking a human-centric approach to fail to exceed their AI ROI expectations. The finding points back to the role of people in shaping workplace outcomes and experiences, even when technology supports the work.

Stack architecture needs a clear division between where AI works autonomously and where human judgment holds. Transactional tasks, answering common benefits questions, scheduling interviews and routing routine inquiries are well suited for automation.

Conversations involving clinical judgment, performance evaluation, and sensitive employee relations require human oversight and accountability. Interoperability shapes the long-term value of this stack. An effective architecture lets specialized AI agents work alongside one another across vendor environments.

Systems that connect to existing applicant tracking systems, learning platforms, and human resource information systems  (HRIS) infrastructure through flexible APIs are more likely to hold their value over time than monolithic single-vendor platforms.

CVI's modular pipeline, including bring-your-own-LLM compatibility and Function Calling that triggers actions in external systems mid-conversation, reflects the same approach. Infrastructure product teams build on what can be adapted to their specific workflows.

The implication for HR leaders

The 2026 HR technology market shows two realities at once. 39% of HR professionals report that AI is currently adopted in their HR functions, 7% intend to launch AI initiatives this year, and 31% are in organizations without plans to do so.

The distance between those groups is where the opportunity sits. For the majority of HR organizations still observing, the entry point is identifying the two or three high-volume conversations where presence drives outcomes: where candidates decide whether to stay in the process, where new hires decide whether they belong, where employees decide whether the feedback they received was fair.

In many organizations, those moments still route people to a text-based interface, a hold queue, or nothing at all. A claims adjuster trainee in their second week, sitting across from an AI human that remembers what confused them last time and adjusts its explanation accordingly, experiences presence: the sense that someone is paying attention to them, understanding where they are, and meeting them there. That feeling has always been the foundation of effective HR. The constraint was always scale. Now it isn't.

See it for yourself. Book a demo.