Rising costs, aging workforces, and relentless competition from native AI and tech-forward competitors are pressuring companies to increase productivity. But while billions have been spent on automation and artificial intelligence, few organizations have achieved transformational, enterprise-wide value. So far, most companies and industries have had to settle for something narrower: faster reports, fewer coders, and modest micro-productivity gains.

What’s missing? A critical focus on linking workflow modernization to workforce modernization. Workflow modernization requires deep process reengineering, simplification, and targeted technology application. Workforce modernization relies on smarter teaming, more robust strategic workforce planning, and dynamic reskilling and redeployment. The two are inextricably linked, but too often, workforce modernization trails behind workflow redesign. If technology’s impact on people is treated as a downstream change management challenge, AI won’t get past micro-productivity improvements. Instead, companies will settle for less ambitious goals, less creative technology deployment, and lower adoption and scaling—leading, in turn, to disappointing ROI, workforce disengagement, and technology skepticism.

Why AI investments stall at micro-productivity

There is a better way. Forward-looking companies are converging on four high-gain moves:

1. Deploy AI in a human-centric way that transforms workflows and modernizes workforces in parallel to boost engagement, experimentation, and adoption.
2. Invest in building the technology and HR capabilities underlying a next-generation continuous improvement engine that dynamically links workflow and workforce as technology continuously changes the allocation of work between humans and agents.
3. As part of the AI transformation, systematically address the accumulated workflow debt that has built up over time in knowledge-based workflows.
4. Reimagine and strengthen your employee value proposition and experience in order to continue to attract, retain, and motivate difference-making talent.

Later in this article, we’ll show how a leading global bank applied these moves by redesigning workflow and roles together to compress a 60- to 100-day process with more than 10 handoffs into a one-day cycle—accelerating learning and adoption.

Well before AI’s widespread adoption, it was clear that the skillful management of human capital is as important to a company’s productivity, sustained performance, and competitive advantage as its management of financial and physical capital. Research done for our book Time, Talent, Energy (HBR Press) showed that weak management systems and poor deployment of human capital drain companies of up to 40% of their productive power.

Human capital is even more crucial in the age of AI. Humans are the primary source of novel innovation and the creators of the intangible assets that represent 92% of S&P 500 market value, according to Ocean Tomo. The companies realizing the largest productivity gains will pursue technological innovation while also unleashing the trapped productive power of their people.

The firms we’ve worked with that have embraced this idea and the four high-gain moves enumerated above are demonstrating a 10% to 15% productivity lift depending on the domain, translating to 10% to 25% EBITDA gains that are growing as the programs scale. This scale of impact has delivered significant return on the underlying investment and effort.

When aimed at the right problems, AI can unlock a technology-fueled, human-centric productivity engine that delivers transformation at scale.

A smarter way forward

From disconnected initiatives to end-to-end workflow redesign

Today, most organizations’ efforts to modernize their workflows and workforces are disconnected, one-off efforts. Finance and procurement rethink sourcing and shoring; tech focuses on automation; operations chase lean efficiency; and HR tackles reskilling, right-sizing, and engagement. This fragmentation creates duplication, missed interdependencies, resistance, and uneven adoption. In the worst cases, they leave destabilized and demoralized workforces facing death by a thousand cuts.

Three years of experimentation since generative AI became accessible with tools like ChatGPT have shown us a better path. Frontier companies are now moving beyond scattered pilots to redesign workflows from end to end using fit-for-purpose technologies that balance off-the-shelf and custom-built applications.

The questions frontier leaders ask

Their executives don’t ask, “What’s the AI use case?” They ask, “What work should stop, simplify, or move in order to better serve customers?” and “What can AI make 10 times better?”

What focus looks like in practice

Those questions force choices. Leaders can’t modernize everything at once, and they can’t automate broken work. The companies pulling ahead stop funding scattered pilots and bet on a critical few end-to-end workflow rebuilds, ones with clear outcomes and single-threaded ownership. They stop waiting for perfect data. They don’t build everything custom. Instead, they ship fit-for-purpose data products for the workflows that matter and default to buying or partnering first, only building bespoke when it truly differentiates their offering. They stop optimizing the current state.

Instead, they start with a clean sheet, remove low-value work, collapse handoffs, reset decision rights, and cap exceptions. Committee vetoes are replaced with tiered risk guardrails. Reskilling, redeployment, incentives, and trust are built into their business case from Day 1.

Redesigning workflow without modernizing the workforce falls short

Even when companies adopt an end-to-end approach to workflow reimagination and redesign, workforce modernization is too often addressed only downstream. This is most pronounced when companies view AI and automation as an efficiency tool (“doing the same with less”) rather than a productivity engine (“doing more with the same”). To really unlock AI’s potential, companies need a unified process that integrates technology, operations, and HR in a synchronized, continuous cycle by deeply connecting workflow and workforce modernization with integrated workforce planning (see Figure 1).

Figure 1

Workflow and workforce modernization should be synchronized at every step:

Prepare leaders and teams. Leaders are often the bottleneck in AI-powered transformations. Many companies search for an almost otherworldly blend of bold, breakthrough vision, deep process knowledge, technology fluency, and workforce planning experience. Because there are few such superhumans, cross-functional teaming of technology, business process owners, finance, and HR is essential. That will be the core unit of change and force multiplier of these efforts.

Don’t automate workflow debt. Organizations that get real returns from AI start with a clean sheet when it comes to their most critical workflows. They don’t try to tweak today’s processes, chase benchmarks, or waste time on small gains in efficiency. They set bold, integrated targets for customer experience, service quality, cost, and speed, and then work backward to design the workflows that can deliver those outcomes. Their future-state design eliminates low-value work, simplifies the path for customers and employees, and determines what must be done by humans, what should be done by humans assisted by AI, and what can be run autonomously by AI agents. These redesigned, streamlined workflows become the anchor for strategic workforce planning and role design in a human-machine environment.

Technology leaders talk about “tech debt,” the complexity of legacy systems that slow everything down. Most companies also carry substantial workflow debt—unnecessary work that has accreted around meetings, approvals, handoffs, exceptions, and one-off policies, making even simple tasks hard to execute. If a “simple” change takes weeks or months and requires multiple handoffs, you’re looking at workflow debt.

AI amplifies whatever system it’s dropped into. If workflow debt isn’t addressed, AI and automation multiply complexity instead of productivity. AI agents quickly make this visible. Humans can work around fuzzy rules; agents can’t. They need clear rules, stable handoffs, and aligned decision rights. When those are missing, agents either push work back to people or produce outcomes that erode trust. Getting process and work design right is therefore a prerequisite to deploying AI agents at scale, not a problem you can fix afterward.

Prepare the hybrid workforce. The future workforce includes four types of workers:

• humans using off-the-shelf AI tools;
• “super humans” made much more productive by custom AI tools;
• autonomous and semi-autonomous agents; and
• humanoid robots.  

These groups will operate in integrated workflows. Humans will focus where they add the most value: traditional human-centered activities such as cultivating relationships with customers, suppliers, and peers and developing and mentoring employees; and activities that can “change the business,” where a human in the loop is critical, specifically those centered on innovation and growth, technology and complex decision governance, and strategy and business transformation. Agents and robots will shoulder more “run the business” activities. Intelligence dashboards track how the boundaries of work evolve and shift between these groups over time.

Given the rate of change, technology evolution, and the constantly moving frontier between humans and agents, AI-powered workforce transformation will need upgraded operational and strategic workforce planning systems, learning and development capabilities, and change management approaches and capacity.

Build trust. Adoption depends on trust in technology, employer intentions, and leadership. Users won’t rely on black-box systems whose reasoning they can’t see or influence. ChatGPT and Claude have been broadly adopted in part because they are tools humans provide context to and that users can evaluate and tweak until they get output they understand and trust. Interaction builds confidence. 

Trust will govern experimentation, adoption, and ultimately the scale of AI’s impact. So, it’s important that employees trust their employer and believe the company is investing to augment—not replace—them. That means reskilling, redeployment, and clear communication of intent. Leaders must actively protect their culture and employee value proposition, keeping people focused on purpose, not just productivity.

Capture the value and set new standards of performance. New workflows must be embedded in enterprise management systems and technology platforms, as well as new labor staffing rules created for strategic workforce planning models. These foundations establish measurable standards for continuous improvement.

What does this look like when it’s done well? A leading UK banking group’s recent AI-enabled redesign of customer engagement shows how paying down workflow debt while redesigning roles and teams can dramatically accelerate experimentation and adoption.

How a leading UK banking group scaled customer engagement while modernizing workflow and workforce together

Consider a UK banking group’s deployment of AI to strengthen customer engagement. Bank employees had strong ideas and a robust digital foundation to build on, but complicated internal processes made testing or deploying any new idea a 60- to 100-day process involving more than ten handoffs.

How the bank rebuilt workflow and workforce

Rather than try to whittle a few days off the total, leaders radically reframed the challenge, asking, “What if the team could do it in one day?” Within four months, the bank’s data products and digital channels teams had built a new workflow, one that combined their tech foundation with agentic AI capable of complex reasoning and problem solving.

What changes in roles, not just tools

They didn’t address workflow in a silo. They knew that to run this workflow, they would need a different workforce too. Shifting from large, specialized teams of as many as 40 people to three- to four-person groupings of “full-stack” problem solvers, they reduced handoffs and tapped into talent with diverse skills and well-rounded experience. New roles include “engagement leads” who handle work that once required journey managers, marketing leads, data analysts, messaging consultants, and marketing leaders. Other new positions, such as “customer lifecycle leads,” “data science and experimentation leads,” and “compliance leads,” similarly streamline work that was once spread over multiple topic experts.

Finding these full-stack problem solvers isn’t easy, however. It requires both reskilling and hiring. But employees have embraced the project in part because the bank’s approach emphasized training and empowerment, not simply headcount cuts. Adoption also relieved pressure on stretched teams.

The changes are helping the bank’s marketing evolve into a continuous learning system in which AI uncovers insights into customer behavior and humans discover what resonates with customers.

A perpetual productivity engine, led by humans, powered by AI

For centuries, humans have used better tools to aid them in their work. But AI is different. Its promise isn’t simply doing the same work faster with fewer people. It’s building a hybrid system in which humans and machines continuously learn from each other. That requires next-generation continuous improvement systems and management models that compound learning and performance, deploying technology that scales at low marginal costs.

Tech-forward companies that integrate workflow and workforce modernization can create perpetual productivity engines led by humans and powered by AI. But most companies need multiple passes through their workflows to build this capability: first, introducing select, smart technology to standardize and simplify workflows; then, using agentic AI to reshape the allocation of work between humans, agents, and robots.

Two reinforcing learning loops power this perpetual productivity engine (see Figure 2).

• In the human-agent loop:

• Humans learn from machine data–driven insights powered by deep machine learning.
• Machines learn from human experimentation and context.
• Each interaction improves both sides’ performance and accelerates the next cycle of learning.

• In the data-systems loop:
  • Enriched data built from internal and external data, ingested and synthesized by AI, powers insights and accelerates learning.
  • Organizations codify learnings into enterprise management, process, and workforce planning to enable enterprise-wide scaling.
  • New best practices are captured and embedded in workflows and management routines.
  • New performance and efficiency standards anchor continuous improvement.
  • Strategic workforce planning and talent systems visualize and manage a hybrid workforce, identify capabilities required in agents and humans, and measure total cost and productivity.

Figure 2

How leaders run the perpetual productivity engine

Leaders who build a self-reinforcing productivity engine run transformations with a small set of operating rules:

• Iterate the workflow like a product. Run modernization as a continuous release cycle, not a one-time transformation. The goal is steady compounding—for example, small changes shipped frequently, measured, and refined.
• Simplify and standardize before automating, especially with agents. AI scales what you give it. If the workflow is inconsistent and has too many variants, handoffs, approvals, and exceptions, automation will multiply that complexity instead of boosting productivity.
• Instrument from end to end and turn exceptions into learning. Perpetual productivity requires telemetry to know what happened and feedback to understand why it happened. Design the workflow in such a way that every human override or exception becomes data from which the system can learn.
• Codify improvements quickly so learning becomes the operating system. The goal isn’t simply insight; it’s institutionalizing insight. Quickly bake what works into workflows, standards, training, and tooling so performance improves by default, not by heroics.
• Keep rebalancing work between humans and machines. As capabilities improve, continually review and update what work is human-only, what is human-assisted, and what is machine-run. This is how productivity compounds as automation expands safely and trust and performance rise.

The difference between AI pilots and a perpetual productivity engine comes down to the way leaders run the work. Consider order-to-cash as an example. AI-forward companies start by fixing the plumbing: standardizing invoice formats and payment terms, cutting unnecessary handoffs, and clarifying decision rights. This makes the process stable enough to automate. With that stable foundation, these companies then modernize in tight cycles, shipping a steady stream of small improvements instead of waiting for a big-bang transformation. One week it might be auto-triage for missing purchase orders; the next, new dispute reason codes; followed by smarter routing for credit holds. Each change is tied to cycle time, first-pass invoice accuracy, dispute aging, and cash collection. 

Feedback loops then compound the benefits of this system. Because an Al-centered company instruments the full funnel—from invoice sent to cash received—and treats every exception as a signal, when a collector overrides an AI recommendation, the reason is captured and fed back into the workflow and the model. What works gets locked in fast, and performance improves by default. As confidence grows, work is rebalanced: Copilots draft and summarize, agents handle low-risk follow-ups and payment plan nudges within limits, and humans focus on the complex disputes and highest-value accounts.

Through constant feedback and adaptation, AI moves beyond simply replacing effort to amplifying human and artificial intelligence, and compounding productivity and innovation.

Reaching a mature state in which workflow and workforce dynamically evolve is a multiyear journey, but the reward is a new, continuous improvement backbone that will make it possible for companies to bend the experience curve in their industry, resetting the rate at which unit costs fall and performance rises.

Many companies are beginning to unlock trapped workforce potential through the smart application of AI and other automation. But it’s not the number of AI use cases that matters so much as taking a human-centric approach that accelerates experimentation, adoption, and engagement while simultaneously redesigning high-value workflows end to end. Organizations that deliver on both dimensions generate total shareholder returns (TSR) 2.3 times that of companies that don’t (see Figure 3).

Figure 3

The practical implication is straightforward. TSR and productivity leaders manage both sides of a balanced productivity scorecard—the “hard” side of workflow outcomes and the “human” side of workforce health—with equal rigor. And they treat divergence as a signal to redesign, not a reason to push harder.

A balanced productivity scorecard tracks both productivity drivers—physical capital, human capital, and technological innovation—and the management practices that boost productivity, such as time management, talent and team development, and energy management (see Figure 4).

Figure 4

Technology constantly reshapes work. Entire blue-collar dominant workforces—from agriculture to manufacturing—have contracted sharply as automation scaled. In the early 1900s, farmwork employed nearly 40% of US workers. Now, it’s under 1%.

AI is maturing and scaling faster than any technology in history and has the potential to be the most transformative productivity driver in generations. If AI triggers for knowledge-based white-collar work even a fraction of earlier displacements, the implications for company strategy, work processes, and talent will be profound.

More than just another technology cycle, AI is a test of leadership and vision. The winning companies will be the ones that do the hard work of paying down workflow debt, simplifying and standardizing what remains, promoting ambitious but focused applications of AI, and modernizing their workforce in parallel to thrive in a hybrid system.

Smarter machines deliver incremental gains. Smarter systems—where human and machine learning reinforce each other—deliver exponential ones.