Source: Unite.AI
Corey Spencer is Group Vice President and General Manager of Product Management for AI & Platform at UKG. He leads product strategy and innovation across UKG’s AI, platform, and workforce intelligence initiatives, helping organizations turn workforce data into actionable insights that shape positive business outcomes. With more than 20 years of experience, Spencer is a respected thought leader on the future of AI, workforce intelligence, and the application of AI to frontline work.
UKG is a global provider of human capital management, payroll, and workforce management software designed to help organizations manage and support their employees more effectively. Its platform combines human resources, payroll, scheduling, time and attendance, compliance, analytics, and AI-powered workforce insights, serving businesses ranging from small companies to large global enterprises. With 50 years of industry experience, UKG supports more than 80,000 organizations across 150 countries, with tens of millions of people using its solutions each day.
You’ve spent more than two decades helping enterprises make sense of customer, analytics, and workforce data across companies like Adobe, Alteryx, and UKG. What convinced you that frontline workers, rather than knowledge workers, represent one of the most important and overlooked opportunities for AI today?
What convinced me is simple: the frontline is where the economy actually runs.
We’ve seen tremendous progress applying AI to knowledge work. But the largest, most dynamic part of the workforce has often been underserved. Frontline workers represent the majority of workers globally — as much as 80% by some estimates — and they power the industries we all depend on every day: healthcare, retail, manufacturing, logistics, hospitality, and public services.
For many companies, the frontline is also one of the largest parts of the P&L. It is where labor cost, customer experience, compliance, safety, service quality, and employee engagement all come together. And yet, it remains one of the least optimized parts of the enterprise.
Not because the work is simple. Because it is incredibly complex.
Frontline work may be planned by the week or the month, but it changes every minute of every shift. Someone calls out. Demand spikes. A patient load changes. A shipment is delayed. A compliance rule applies differently based on role, location, certification, or hours worked.
Traditional enterprise systems were built to record what happened. The frontline needs technology that can help organizations act while the decision still matters.
That is why AI is so important. Agentic AI can understand context, intent, and constraints, then help orchestrate action across scheduling, time, pay, HR, and operations — with people in control.
For me, the opportunity is closing the gap between workforce planning and workforce execution. When we do that, employees get better experiences, managers get support at a scale no individual can handle alone, and businesses can operate with more speed, precision, and trust.
This is not just a productivity story. It is an execution story.
Frontline employees make up the majority of the global workforce, yet much of the AI boom has focused on office productivity tools. Why do you think the industry underestimated the complexity and importance of frontline environments for so long?
I think the industry underestimated the frontline because it was easier to build AI for the work that was already digital.
Office work happens in documents, emails, meetings, calendars, and collaboration tools. That made it a natural first wave for AI. But the frontline is different. It is physical, distributed, highly regulated, time-sensitive, and constantly changing. And yet, it represents as much as 80% of the global workforce — people who generally do not sit at desks, but who keep businesses, communities, and economies running every day.
The complexity is not just the work itself. It is the data around the work. We’re fortunate because UKG has always been focused on the frontline, so we have a multi-decade head start in this area.
Frontline workforce data is incredibly high-volume and noisy. It comes from schedules, time clocks, punches, pay rules, certifications, labor forecasts, absences, demand signals, compliance requirements, manager decisions, employee preferences, and operational changes happening across thousands of locations and shifts. Those signals are constantly moving, and they are deeply dependent on context — role, location, skill, policy, regulation, demand, and timing.
That has made frontline environments much harder to analyze in a meaningful way. By the time many organizations can understand what happened, the moment to influence the outcome has already passed. A staffing gap, overtime risk, compliance issue, payroll exception, or missed coverage signal does not matter a week later. It matters while there is still time to act.
That is why frontline AI has to be fundamentally different from office productivity AI. It cannot just summarize information or generate content. It has to understand operational context, interpret noisy workforce signals, and help people take action in real time.
For UKG, this is the core opportunity of our business and solutions like the UKG Workforce Intelligence Hub. The frontline does not need another system of record. It needs a system of action — one that helps organizations turn workforce data into better decisions, faster execution, and more trusted outcomes for employees, managers, and the business.
For years, enterprise software was built around work that was easier to digitize: desks, documents, meetings, workflows, and knowledge work. But frontline work is different. It is physical, distributed, regulated, time-sensitive, and highly industry-specific.
A retail associate, a nurse, a manufacturing technician, a warehouse worker, and a hotel manager are all frontline employees, but the operating environment around each of them is completely different. Retail teams may need to adjust labor based on traffic, promotions, weather, or local events. Healthcare teams need the right person with the right credential in the right unit as patient needs change throughout the day. Manufacturing teams balance skills, safety, overtime, production schedules, and labor rules.
That is the real complexity. Work may be planned by the week or month, but it changes every minute of every shift. A schedule that looked right on Monday may be wrong by Wednesday. A staffing plan that worked at 8 a.m. may be broken by noon.
Historically, most systems could only record what happened after the fact. The frontline needs more than reporting. It needs action in the moment.
That is why AI is such an important tailwind. Agentic AI can understand context, intent, constraints, and business rules, then help managers and employees take action across scheduling, timekeeping, pay, HR, and operations.
We have spent decades working on these frontline challenges. What is different now is that AI gives us a way to transform them at a new level of speed, precision, and industry relevance.
The opportunity is not to make frontline work look like office work. It is to build technology that respects how frontline work actually happens — and helps people and businesses make better decisions in real time.
Building AI for a retail associate, nurse, warehouse worker, or hospitality employee is very different from building AI copilots for desk workers. What are the biggest technical and operational differences between those environments?
The biggest difference is that frontline AI cannot just live in a document, inbox, or chat window. It has to live in the flow of work.
For desk workers, AI often helps create, summarize, search, or analyze information. For frontline workers, AI has to understand what is happening in the operation right now — who is working, who is qualified, who is available, what demand looks like, what rules apply, and what action should happen next. That is a very different technical problem.
So the AI has to be contextual, industry-aware, and action-oriented. It has to understand roles, policies, labor rules, pay implications, schedules, certifications, and business priorities — and it has to do that with trust, explainability, and human oversight.
The operational difference is just as important. Frontline managers do not have time to interpret dashboards while the shift is moving. Employees do not want another system to navigate. AI has to be embedded, fast, intuitive, and useful in the moment.
That is why frontline AI is not simply a copilot for work. It is intelligence that helps organizations sense, decide, and act in real time.
You’ve spoken about the importance of real-time workforce data, including shifts, scheduling, payroll, and staffing signals. Why is real-time execution especially critical for frontline AI systems?
The challenge is not whether an organization has the data. Most enterprises are creating more workforce data than ever. The challenge is translating those signals into timely recommendations, guidance, and action — while the decision still matters.
That is why real-time execution is so critical for frontline AI. AI needs to understand what is happening, what constraints apply, and what the best next action should be. Do we adjust a schedule? Fill an open shift? Alert a manager? Prevent a compliance issue? Fix a payroll exception before it becomes an employee experience problem?
These may sound like small decisions, but across thousands of employees, shifts, and locations, small improvements compound quickly. Better coverage, fewer exceptions, smarter labor decisions, and faster manager action can create meaningful impact for employees and for the business.
This is the purpose of the UKG Workforce Intelligence Hub: to help organizations turn workforce data into better decisions and faster action, especially for the frontline.
We have a uniquely rich view across HR, pay, and workforce management — the areas where workforce strategy becomes operational reality. When that data is combined with industry and peer benchmarks, labor cost insights, and AI that can surface the most relevant takeaways and recommended actions, leaders get a much clearer view of what is happening, why it matters, and where to act.
For executives, that is the real value. It is not another dashboard to review. It is a tool to help manage labor more precisely, identify risks earlier, benchmark performance more effectively, and make decisions with greater confidence in the moments that matter.
Many organizations are exploring agentic AI, but there’s concern around autonomy, governance, and trust. What guardrails are necessary before enterprises can allow AI agents to take action in workforce management systems?
In workforce operations, AI agents need clear boundaries because the decisions are deeply human. They affect schedules, pay, compliance, coverage, fatigue, fairness, and employee trust. That means agents cannot simply optimize for speed or efficiency. They have to operate within policy, permissions, and purpose.
The most important guardrails are context, control, and accountability.
Context means the AI understands the employee, the role, the location, the labor rules, the business need, and the downstream impact. Control means the agent acts only within defined permissions, with the right level of human approval based on risk. Accountability means every recommendation or action is explainable, auditable, and reversible.
Some actions may be appropriate for an agent to execute automatically, like surfacing an open shift to qualified employees or alerting a manager to an overtime risk. Higher-impact actions, especially those involving pay, compliance, discipline, or material schedule changes, should require human review.
The goal is not to remove managers or HR teams from the process. It is to give them intelligent support at a scale no individual can manage alone.
Trusted agentic AI should turn intent into action — but always with transparency, governance, and people in the loop.
Every capability should undergo rigorous model validation, security review, privacy assessment, and ongoing monitoring before it reaches production.
UKG has discussed AI applications that can rebalance schedules, detect payroll anomalies, and accelerate hiring. Which of these use cases are already delivering measurable operational value today, and which are still early-stage?
The use cases delivering value fastest are the ones that remove friction from decisions managers and teams already have to make every day.
Payroll anomaly detection is an example everyone can likely relate to. Payroll is one of the highest-trust processes in any business, and it is especially complex for frontline and hourly employees. A missed punch, overtime issue, pay rule exception, or compliance problem can create hours of manual work and directly affect employee trust. AI can help identify issues earlier, explain what looks wrong, and guide teams toward resolution before payroll closes.
There’s some much more advanced examples we are very excited about though. One is real-time labor and schedule optimization. This is where UKG Dynamic Workforce Operations is so important. The value is not simply seeing that a store, unit, warehouse, or hotel is understaffed. The value is understanding what changed, what constraints apply, and what action a manager can take in the moment — whether that is filling an open shift, avoiding overtime risk, or protecting coverage. These decisions are made today by gut, often hours too late. Now, managers can get a recommendation backed by data before a human even knows there is a problem.
The third is frontline support in the flow of work. Through the UKG Frontline Worker Network, workforce data can help connect employees to support around wealth, health, and everyday essentials when those needs are most relevant. That is still an emerging category, but it is an important one because employee experience cannot live in a separate portal no one has time to use.
Hiring acceleration is also becoming very practical, especially in high-volume frontline environments where speed matters. Our AI-native UKG Rapid Hire has reduced the time from candidate interest to scheduled interview from weeks and days, to hours and even minutes in some cases. We’ve reduced repetitive work across sourcing, screening, scheduling, onboarding, and first-day readiness for frontline hiring at scale.
The real opportunity is not isolated automation. It is a workforce operating platform that can sense what is changing, recommend the next best action, and help organizations move from intent to action with the right governance and human oversight.
In sectors like healthcare, manufacturing, and logistics, small workforce disruptions can have immediate downstream consequences. How do you design AI systems that can respond quickly without creating additional operational risk?
Speed matters, but in frontline environments, speed without control is risk.
In healthcare, manufacturing, logistics, and other complex sectors, a small workforce disruption can ripple quickly. One absence can affect patient coverage. One skills gap can slow a production line. One missed staffing signal can delay deliveries or create overtime exposure.
So the goal is not to build AI that simply moves faster. The goal is to build AI that moves faster within the right constraints.
That starts with context. The AI has to understand who is qualified, what rules apply, what coverage is required, what compliance risks exist, and what the downstream impact of each action could be.
It also requires risk-based autonomy. Some actions can be recommended or executed with very low risk, like alerting a manager, identifying qualified replacements, or surfacing an open shift. Other actions should require human review, especially when they affect pay, compliance, safety, or material schedule changes.
The best frontline AI systems are designed to be explainable, auditable, and reversible. Managers need to know why a recommendation was made. Employees need to trust the process. And the business needs confidence that AI is improving execution without introducing new risk.
For us, the design principle is simple: AI should help organizations sense, decide, and act in real time — with people, policy, and governance built in from the start.
There’s growing concern among workers that AI could replace jobs, particularly in frontline industries. How do you balance automation with empowering employees rather than creating fear or uncertainty around AI adoption?
I think we have to separate automation hype from operational reality.
Over the last decade, we have seen plenty of examples where replacing frontline work proved much harder than expected. Self-checkout has had mixed results. Fully automated fast food has been more pilot than paradigm. Drone delivery still feels, for most people, like something from the future.
That is not because the technology is not powerful. It is because a large portion of frontline work remains deeply human.
Frontline employees handle exceptions, judgment, care, safety, service, empathy, and trust. A nurse is not just completing a task. A retail associate is not just moving inventory. A restaurant worker is not just fulfilling an order. These roles sit at the intersection of people, operations, compliance, and customer experience.
AI can remove the friction around frontline work: confusing schedules, missed punches, pay questions, open shifts, compliance complexity, staffing gaps, and administrative tasks that pull managers away from their teams. It can help employees get answers faster, help managers make better decisions, and help organizations operate with more precision.
The future of frontline AI is not about taking people out of work. It is about giving people better tools, better guidance, and more support in the moments that matter.
You’ve described AI as moving enterprise software from systems of record toward systems of intelligent action. How do you see workforce platforms evolving over the next five years as agentic AI becomes more embedded into day-to-day operations?
I would actually reframe that slightly. The question is not what workforce platforms will look like five years from now. The question is which organizations are building the foundation today to be leaders five years from now.
The pace of change in AI is too fast to treat this as a distant transformation. The advantage will go to organizations that move now from disconnected systems, clunky suites, and point solutions toward a true workforce operating platform — one that connects HR, pay, time, scheduling, workforce intelligence, and execution.
This is another example of where the UKG Workforce Intelligence Hub becomes important. As AI becomes more embedded in daily workforce operations, intelligence cannot sit off to the side as static reporting. It has to become part of how work gets understood, prioritized, and acted on. Organizations need an intelligence layer for frontline workforce operations to bring the right content into the right moment.
That matters because workforce operations are full of marginal gains and marginal losses. One missed staffing signal, one payroll exception, one compliance issue, one open shift, or one delayed manager decision may look small in isolation. But across thousands of employees, locations, and shifts, those moments compound quickly.
Agentic AI gives organizations a way to close that gap. It can understand what is changing, interpret the business context, recommend the next best action, and help people execute in real time with the right governance and human oversight.
Over the next five years, the leaders will not be the organizations that simply added AI features to existing workflows. They will be the ones that rethought workforce management as a system of intelligent action.
For executives, this is the urgency: the frontline is already moving in real time. Your technology needs to move with it. The organizations building that capability now will have a significant head start.
Looking ahead, do you believe the companies that win in enterprise AI will be the ones with the best models, or the ones with the deepest operational and workforce data tied directly to real-world workflows?
The best models will matter, but I do not believe models alone will define the winners in enterprise AI.
In the enterprise, intelligence is only valuable if it understands the context of work and can help people take the right action. That is especially true in workforce operations, where a recommendation may depend on who is scheduled, who is qualified, what rules apply, what demand looks like, what pay implications exist, and what risk or compliance issue may be created downstream.
A model without that operational context can generate an answer. A platform with the right workforce data, business rules, workflow context, and governance can help drive an outcome.
This is why an intelligence layer, which is what we’ve created with our Workforce Intelligence Hub, is so important. AI sitting on top of a system has limited value. Intelligence should be embedded into the flow of work, able to interpret signals, understand operational constraint, and surface decisions that matter most.
That is where I think the market is headed. As models become more powerful and more accessible, differentiation will come from how well AI is grounded in real-world workflows and how safely it can move from insight to action.
For the frontline, that distinction is critical. Work changes constantly, and decisions have immediate consequences for employees, managers, customers, patients, and the business.
The companies that win in enterprise AI will not simply have AI that sounds intelligent. They will have AI that understands how work actually happens — and can help organizations operate with more speed, precision, trust, and humanity.
Thank you for the great interview, readers who wish to learn more should visit UKG.