Stop debating automation vs. augmentation

For the past few years, leaders have debated whether artificial intelligence will automate jobs or augment workers. That framing made sense when AI primarily assisted employees with individual tasks. But as organizations begin deploying agentic systems that can execute entire workflows, a more consequential question is emerging: How should organizations use agentic AI to reimagine workflows that create growth rather than simply reducing costs?

Drawing on our research across industries, we see organizations confronting this question from two complementary perspectives. Katherine C. Kellogg’s research examines organizational transformation, workflow redesign, workforce planning, and the management practices needed to help leaders and employees navigate AI-related change. Rama Ramakrishnan’s work focuses on business strategy, digital innovation, and the use of emerging technologies to create new sources of growth and competitive advantage. 

Together, these perspectives point to a broader leadership agenda: Organizations need to redesign workflows around agentic AI’s distinctive capabilities, pursue strategically valuable opportunities, prepare employees for new roles, and use productivity gains enabled by agentic AI to develop new products, services, capabilities, markets, and jobs.

Move beyond the automation-vs.-augmentation debate

The automation-vs.-augmentation debate focuses on how tasks should be divided between people and AI. Leaders ask which tasks should be automated, with AI performing them in place of people, and which tasks should be augmented, with AI enhancing peoples’ performance of them. 

Agentic AI expands the scope of the leadership challenge. Agents can perform substantial parts of workflows autonomously while, increasingly, people are supervising the systems, governing their actions, managing exceptions, and improving their performance. This changes the design of work, roles, and organizations. The relevant unit of analysis becomes the workflow rather than the individual task. And leaders must now consider how they can transform entire workflows, using agentic AI to create new value, expand organizational capabilities, and open opportunities for growth.

Pursue growth rather than simply increasing efficiency, and do it across workflows

Many leaders choose to begin with automating existing workflows, because efficiency gains generated from these projects can reduce operating costs, demonstrate value, and unlock resources for broader AI investments.

But firms cannot stop there or wait too long before pursuing growth initiatives.

Organizations that concentrate primarily on automating existing workflows may improve their cost position, and if price is the basis of competition, those cost savings can be channeled into price reductions that drive increases in market share. However, AI tools are available to competitors as well, so this is not a durable source of competitive advantage.

What leaders who are focused on automation leave underused are resources that competitors cannot readily acquire: proprietary data, the expertise of experienced employees, and established customer trust. Since every organization has access to the same models, productivity gains alone are table stakes; lasting advantage comes from combining those resources with agentic AI to create new products, services, and business models that rivals cannot replicate. These advantages can stem from reimagining a workflow or from combining multiple related workflows into a single, new workflow that would have been impossible before agentic AI.

Some firms may allocate separate budgets for productivity and growth initiatives, whereas others may need to sequence their investments, using initial savings to fund the skills, infrastructure, and experimentation required for growth. What matters is establishing a deliberate path that gradually tilts the agentic AI portfolio over time from productivity improvements to new value creation.

Prioritize a small number of high-value workflows

Enterprise-scale agentic AI often requires access to core systems, changes to how work is carried out, new safeguards against security risks, and decisions about who is accountable for the technology’s actions. Because each initiative demands substantial investment in systems integration, process redesign, security, and governance, as well as coordination across functions, leaders should concentrate resources on a limited number of strategically important workflows rather than spreading their investments across many pilots.

Our research suggests that organizations should evaluate potential workflows for agentic AI transformation, using particular criteria.

• Strategic importance: Would the initiative address a problem worth solving or an opportunity worth pursuing? If the firm chose not to pursue the initiative but the competition did so successfully, would it put the firm at a significant competitive disadvantage?
• Technical fit: Can agents perform the workflow reliably? Rather than evaluating the reliability of agents in isolation, firms should compare agents to humans performing the same workflows and assess whether the performance gap makes agentification beneficial.
• Foundation model substitution risk: To what extent is the proposed capability likely to be provided by future foundation model improvements, reducing the need for custom development?
• Operational readiness: Does the organization have the necessary data, processes, and infrastructure?
• Compliance and equity risk: What could go wrong, and how consequential would a failure be?
• Return on investment: Would the value created exceed the full cost of deployment, oversight, and risk? 
• Reuse potential: Could the agent’s underlying capabilities or outputs be applied elsewhere? For example, an agent that extracts structured information from transcripts or documents can be reused across functions, and the data it produces can support many other applications. That single agent has become a platform with broader value.
• Growth potential: Can the new process be the basis for a new offering? For example, if agentic AI reduces the time to resolve a customer issue from the industry-standard few days to a few minutes, that speed could support a premium-priced version of an existing product or set the product apart from competitors’ options and grow market share.

Reimagine the future rather than just automating the present

Automating an existing workflow can produce valuable gains. By examining a process end to end (including the flow of information, sequencing of decisions, coordination among functions, allocation of responsibility, handling of exceptions, and points of human oversight), leaders can determine where agents can act autonomously, where people should exercise judgment, and how the two can work together effectively

However, reimagining the future is even more valuable, because it can lead to a fundamentally different operating model. By questioning assumptions embedded in existing workflows and envisioning possibilities that did not make sense before agentic AI, leaders can redesign workflows around what agents can do that people cannot: work at machine speed, perform many tasks at once, retain and draw on context across every step, monitor operations continuously, and coordinate directly with other systems and agents.

Doing that is not easy. Human imagination is needed to help people overcome what psychologists call functional fixedness: the tendency to interpret new possibilities through the lens of familiar tools and ways of working. To look beyond existing workflows and imagine what might become possible when human constraints no longer apply, leaders must combine knowledge of internal workflows with an understanding of agentic AI capabilities and granular insights about the nature of competition in their organization’s markets and product categories. They must challenge themselves and their employees to ask “What configuration of people and agents will allow us to create the greatest value in this area of the business?”

Some organizations are creating structured opportunities for this work by using cross-functional design sprints. These efforts bring together business leaders with technical specialists who understand what agentic AI makes possible, and employees who understand existing operations. Teams examine external models, challenge current practices, and experiment with new combinations of people and agents.

External learning is also important. Startups and other early adopters are conducting experiments that can signal where new opportunities are emerging. Organizations that remain close to these experiments can identify valuable problems before the solutions and markets become obvious.

The central question is not simply “How can agentic AI perform this existing workflow more efficiently?” It is “How would we design this workflow or a set of related workflows, and the related business, product, service, or market, if we were starting today with agentic AI?”

Redesign jobs alongside workflows

As organizations automate and reimagine workflows, they must also redesign jobs.

When agents perform larger parts of workflows, employees shift from carrying out routine activities to supervising systems, resolving exceptions, exercising contextual judgment, and improving AI-enabled operations. New roles are emerging around agent ownership, governance, performance monitoring, and workflow orchestration.

These transitions require deliberate workforce planning. Leaders need to communicate transparently about how work is changing, identify future roles, determine how employees can move into them, and provide the necessary opportunities for reskilling.

Generative AI Business Sprint

Attend Online

REGISTER NOW

Employee voice can improve the quality of the redesign. People who currently perform the work understand its exceptions, dependencies, informal practices, and points of failure, and the workarounds they can use when the official process falls short. Their knowledge can help organizations make these workarounds explicit so that workflows built around AI agents still function when conditions depart from the norm and are able to identify where human oversight remains especially valuable. 

At the same time, employees may be apprehensive that their contributions to AI projects will increase their risk of losing their jobs. Workforce planning, job redesign, and reskilling should therefore be part of agentic AI initiatives from the beginning. They should not be addressed only after workflows have already been reimagined and redesigned around agentic AI’s distinctive capabilities.

Treat workforce strategy and growth strategy as inseparable

The effects of AI on employment will be shaped partly by the growth choices leaders make.

If they choose a strategy centered on efficiency, they will reduce the amount of labor required to deliver existing products and services. In contrast, if they choose a growth strategy, they will create demand for new products, services, human and organizational capabilities, and jobs. Some of the resulting work will be automated, while some will require human judgment, coordination, relationships, and expertise.

For this reason, a growth strategy is also a workforce strategy.

Leaders cannot guarantee that every employee will move into a new role or that the quantity and composition of work will remain unchanged. They can, however, influence workforce outcomes through the opportunities they choose to pursue, the jobs they design, and the investments they make in employee upskilling and reskilling.

This makes workforce planning and business strategy mutually reinforcing. Workflow redesign determines how the organization creates value, job redesign determines how people contribute to that value, and growth strategy determines whether the organization creates new demand, new capabilities, and new work for its people. 

For this reason, leaders must explicitly focus on using agentic AI to reimagine workflows that create growth rather than simply reduce costs. The defining leadership challenge of our time is turning agentic AI-enabled productivity gains into new value for the business and new possibilities for work.

Key Takeaways for Leaders

• Move beyond the automation-vs.-augmentation debate. Now that agentic systems can execute entire workflows rather than just assisting employees with individual tasks, examine how people, agents, workflows, and business strategy can be combined to create value.
• Pursue growth as aggressively as efficiency. Don’t stop at automation or wait too long before pursuing new products, services, revenue streams, and markets. Because AI tools are available to competitors as well, automation is not a durable source of competitive advantage.
• Prioritize a small number of high-value workflows. Enterprise-scale agentic AI often requires access to core organizational systems, changes to how work is carried out, new safeguards against security risks, and decisions about who is accountable for its actions. Because each initiative demands substantial investment, select workflows based on strategic importance, technical fit, foundational model substitution risk, operational readiness, compliance and equity risk, ROI, reuse potential, and growth potential.
• Reimagine the future rather than automating the present. Human imagination is needed to overcome the tendency to interpret new possibilities through the lens of familiar tools and ways of working. Use cross-functional experimentation and external learning to envision the possibilities created by agentic AI and redesign workflows around what agents can do that people cannot.
• Redesign jobs alongside workflows. People who currently perform the work are critical to developing AI agents, because they understand exceptions, dependencies, informal practices, and points of failure. But employees may be apprehensive that their contributions to AI projects will increase their risk of losing their jobs. Engage in workforce planning, transparent communication, and reskilling to show people where the growth vectors are.
• Treat workforce strategy and growth strategy as inseparable. Use AI-enabled productivity gains to create new sources of demand, opportunity, organizational capability, business value, and possibilities for work.

Katherine C. Kellogg is the David J. McGrath Jr. Professor of Management and Innovation at the MIT Sloan School of Management. Her research focuses on helping knowledge workers and organizations develop and implement predictive, generative, and agentic AI systems in everyday work to improve decision-making, collaboration, and learning.

Rama Ramakrishnan is professor of the practice in AI and machine learning at MIT Sloan. His interests center on the application AI to drive business growth and productivity gains, and in the creation of intelligent products and services.