Gen AI has invaded the workplace, but its impact remains elusive. While two-thirds of global companies use gen AI,1 these deployments are proving to be nothing like past enterprise software projects. CEOs increasingly understand gen AI’s potential yet remain unclear on how to create meaningful value from the technology.
Gen AI has the potential to completely change how employees work. Its natural language interface makes it easy to use, while its burgeoning reasoning and agentic capabilities allow it to perform increasingly complex tasks such as interpreting large volumes of information, coding, and answering queries. The most advanced agents are even starting to perform tasks such as creating spreadsheets and navigating web pages. And employees are clearly eager to use it; they are already doing so three times more than their leaders realize.2
But simply putting new technology into people’s hands does not ensure they will use it effectively, nor does it profoundly change the way a company works. Instead, CEOs need to deploy a novel change management approach that mobilizes their people, turning them from gen AI experimenters into gen AI accelerators. This is not a linear process. Change management in the gen AI age asks employees to become active participants rather than just users. It asks employees to experiment, cocreate products, and commit to continual skill development. It also recognizes that not everyone will make the transition smoothly, and that some employees will need additional support. In this article, we outline five steps that CEOs can use to lead their companies through these uncharted waters.
This new type of transformation could usher in an era of tremendous growth—and finally unlock real value from gen AI. But getting there will require CEOs to manage the substantial change that gen AI will bring to their workforces. The best leaders will reconfigure their organizations to be fully gen AI enabled, arming their employees with AI superpowers to free them from tedious and repetitive tasks.
Step 1: Craft a North Star based on outcomes, not tools
It’s tempting to view gen AI as just another tool that employees will be required to use. But CEOs are learning that this view is incomplete—and that employees see AI as more of a capability than a tool. An organization reconfigured around AI will have humans and gen AI agents working together seamlessly. While that future has not yet arrived, forward-thinking CEOs are creating North Star plans today to succeed in this new world tomorrow.
The North Star should be simple enough to be universally understood but bold enough to inspire teams. It should accommodate the evolution of technology and be able to absorb and activate frontier AI models and their new capabilities within a reasonable time frame. The North Star should define how an organization will create value and competitive advantage from gen AI, and what the likely impact will be on the talent life cycle. To lead such a change successfully, leaders may need to deepen their own education—both on what gen AI can do today and the trajectory of what’s likely in the future.
Managing the magnitude of change that such a North Star represents will require two critical components: a well-resourced change management plan and a full reimagining of end-to-end workflows. Leaders will need to consider both the pace of technological change and the ability to mobilize their people toward this future state. For example, a company could start by deploying gen AI agents that complete discrete tasks and then evolve into deploying agent swarms capable of achieving complete business outcomes. In this version of a gen-AI-enabled company, parts of the organization could become minimum viable organizations (MVOs) where agent swarms oversee most work, with as few humans as possible to check their outputs. Other parts of the company would retain larger numbers of human workers equipped with technology-enabled superpowers. When designing a North Star, CEOs can contemplate which parts of their companies lend themselves to each approach, such as high-touch customer functions remaining more human powered, while back-office operations become MVOs.
Whatever the balance between human and machine, the change management approach will need to recognize that the rate of change across an organization will be uneven. The journey of creating this type of blended human and AI organization is beyond what traditional playbooks cover.
Step 2: Build trust with accessible data, governance, and enterprise wisdom
Creating foundational trust in gen AI use throughout the organization is essential. After all, if employees don’t trust gen AI output, they won’t trust the decisions it makes. And, thus, the technology will have little chance of attaining scale. Our research shows that “gen AI high performers,” or those companies that attribute at least 10 percent of their EBITDA to their gen AI usage, are more likely than other companies to invest in trust-enabling activities (Exhibit 1). And when companies invest in building trust in AI and digital technologies, they are nearly two times more likely to see revenue growth rates of 10 percent or higher than companies that do not.3
Building a foundation of trust also requires leaders to treat data accessibility as a first-class work stream and a key component of the change management process. Our experience is that almost every organization has a “marketplace of information” shaped by hierarchies and policies that define where information is housed and who can access it. Gen AI’s ability to consume unstructured data within these marketplaces can be a distinct competitive advantage, helping organizations extract hidden insights and patterns from their data. But training models to process both structured and unstructured data to deliver high-quality, trustworthy answers remains challenging. The business practices and technology frameworks needed to monitor the quality of gen AI’s responses are still emerging, making it critical for leaders to establish trust now.
An effective change management approach needs to set clear expectations among all employees for data governance and usage. Together with the CEO, the CIO and chief data officer (CDO) can prioritize accessible data and then wrap robust AI governance around it. This requires establishing an AI oversight committee, defining policies on acceptable use, setting compliance and risk guidelines in conjunction with risk and legal teams, and ensuring human-in-the-loop checkpoints to spot hallucinations, biases, or data leakage.
In highly regulated industries, building trust in gen AI is especially critical, particularly for employees who handle sensitive client data.
Morgan Stanley, for instance, worked with OpenAI to train a gen AI assistant on more than 100,000 of the bank’s research reports. But the bank did not roll the assistant out firmwide until rigorous evaluation frameworks proved the AI’s answers met advisers’ quality standards. The result: Once deployed with proper guardrails, the “AI @ Morgan Stanley Assistant” quickly reached 98 percent adoption by the firm’s wealth management teams, democratizing access to expertise.4
The most trusted gen AI platforms are those grounded in an organization’s own context, revealing how answers are derived and the sources used. When deploying any gen AI model, a company should augment it with institutional knowledge such as proprietary research, customer interaction logs, or decades of engineering know-how. Companies can also consider boosting this internal knowledge base with dynamic external information. When gen AI delivers knowledgeable answers that users can trust, they will be far more apt to incorporate it into their daily workflows.
Step 3: Reimagine workflows to evolve toward AI teams
Bolting gen AI onto existing processes will not encourage widescale uptake and will deliver incremental, if any, impact. That’s because gen AI is not a tool like most enterprise software and is instead a capability—providing a new way of thinking, working, and creating. Instead, leaders can put gen AI at the center of workflows, entirely reconfiguring how work takes place in their organizations. To do this, it’s critical to use a two-in-the-box approach to change, where business and technology teams work together to define the new way of working. Business teams ensure the new working mode delivers the expected business results, while the technology team ensures the feasibility of architecting the technical change.
Such a reimagining can evolve over three phases to allow people to adapt to new ways of working. This evolution can progress from stand-alone AI agents that humans use to complete very discrete tasks, to groups of AI agents that complete full end-to-end processes overseen by humans, to fully automated agentic swarms that act independently as MVOs to deliver full business outcomes. Even in this last phase, people are not eliminated. A few employees will be needed to ensure MVOs function safely and effectively, and many others will be redeployed to higher-value activities. Effective change management during this three-phase evolution cannot begin without first building trust among employees that gen AI is a core capability required for their jobs.
In the first phase, companies clarify a few end-to-end workflows that could be transformed with AI, such as procurement to payment, hire to retire, or idea to product. Then, they identify where to insert stand-alone gen AI agents into these workflows to help employees complete discrete tasks, but not the entire workflow. This is the phase when gen AI acts most like a “tool,” since employees use it to complete specific tasks.
In the second phase, companies expand the use of gen AI agents, creating groups of agents that work together to complete all the tasks in the end-to-end workflow, learning as they evolve to get better and better at the process. In this phase, humans still manage the agents. While the agents themselves do all the work, humans oversee their operations and outputs.
In the third phase, more gen AI agents are added to create an agentic swarm that could reach full autonomy, operating as MVOs with no human intervention. In this third phase—which holds vast potential but has yet to be proved at scale—humans would still have oversight over the agents but would not be involved in the end-to-end work at all. Instead, they would be freed up to focus on higher-value work. Thus, the company achieves its North Star goal.
In the first two stages, involving employees directly in the process maximizes the change’s impact. Leaders can invite them to create their own agents and provide feedback on areas where gen AI could be woven into their workflows. The strongest leaders set the expectation that all employees are part of the process to make gen AI work—and that it is essential to the company’s future success. Selecting the right work processes to automate first can increase employee buy-in because these improvements will make their jobs easier. Leaders can select work streams where value is clear, feasibility is high, and the amount of investment required to make the change happen is manageable.
Once work streams are augmented by gen AI, employees need formal training on how to use them. This helps reduce employees’ anxiety and builds their confidence to do work in new ways. When employees receive adequate training on gen AI tools, they use the tools more and more frequently as their skill levels rise (Exhibit 2). What’s more, our research shows that 48 percent of US employees would use gen AI tools more often if they received formal training, and 45 percent would use gen AI tools more frequently if they were integrated into their daily workflows—citing these two factors as the most motivating out of many.5
Integrating gen AI into daily workflows moves it from a hobby to a habit—and encourages employees to see it as a team member, not a tool. Consider how our own organization drove usage of the McKinsey internal gen AI platform called Lilli. Senior leaders make a point of role modeling uptake by asking in every team meeting: “Have you asked Lilli?” New hires are taught how to use Lilli during onboarding, while regular risk trainings for the entire firm include reminders on how to use it appropriately. Engagement with the platform was also accelerated by a continual rollout of new features that our teams could use on a day-to-day basis, such as the ability to generate PowerPoint slides in the firm’s template, build spreadsheets, and generate draft client proposals. Since Lilli’s launch in July 2023, 92 percent of McKinsey’s global staff have used the AI platform, and 74 percent now use it regularly, saving more than 30 percent of their time on information gathering and synthesis. To date, Lilli has answered nearly 19 million prompts. By making Lilli part of every job at McKinsey—and setting the expectation that Lilli is simply part of the team—usage rapidly became normalized.
Step 4: Rethink organizational structures with a mix of MVOs and augmented teams
As gen AI becomes embedded into workflows, CEOs will need to decide how different parts of the enterprise are structured. Only some business units may evolve into MVOs, which are extremely lean and highly automated workflows. Other parts of the company, meanwhile, will remain in step two of the AI evolution, giving human teams digital superpowers so they can accomplish more than ever. Both trajectories will require CEOs to modify organizational structures, communicating clearly with employees about how the changes will impact them.
MVOs will work best when handling repetitive or logic-based work. For example, consider a routine back-office process such as invoice processing. With gen AI, a company could automate invoice matching, approvals, and entries with near-zero human touch, maintaining just a small team to handle exceptions.
To enable an MVO, companies must, of course, invest in robust AI operations and monitoring. But just as importantly, CEOs must rethink their talent strategies. The few people running an MVO need to be highly skilled in managing AI systems, data analysis, and exception handling. Roles such as AI workflow optimizer or automation product owner may emerge as critical positions. And team members who formerly completed tasks in the now-MVO workflow will need to be reskilled to deliver value in other parts of the organization. While not every part of the business can or should become an MVO, CEOs should identify areas where a lean, AI-first model could dramatically cut costs and improve speed or accuracy.
On the other hand, some functions should never evolve into full MVOs but instead stop at phase two and remain augmented teams. Sometimes, equipping humans with AI superpowers makes them vastly more productive, but the human touch is still needed. This is already happening in many companies. Sales teams, for instance, are already using gen AI to analyze customer data and get tailored suggestions for upselling or personalized marketing content in seconds. This could allow a single salesperson to manage a larger portfolio with much higher conversion rates. Customer service representatives are using gen AI to resolve issues more quickly and consistently. But taking humans entirely out of the loop in these customer-facing roles would not make sense, as it could erode the customer experience in ways that damage the company’s brand.
Step 5: Empower employees to learn and become change agents
A gen AI reconfiguration of workflows will not succeed unless leaders bring the whole workforce along on the journey. Our research on large-scale tech transformations shows the power of involving employees in the process. On average, only about 2 percent of employees are directly involved in a typical transformation effort. Yet organizations that expand their participation see far better outcomes. In fact, companies involving at least 7 percent of employees in transformation initiatives double their chances of delivering positive excess total shareholder returns (TSR), with the highest performers involving 21 to 30 percent of employees.6
Inviting employees to become gen AI ambassadors is even more important than in a typical technology transformation because we are all still learning the extent of gen AI’s use cases and capabilities. Everyone at every level of an organization can learn together. However, some “superusers” can become powerful change agents who boost overall uptake. In any organization, these employees should be identified and supported to drive cultural change. Our research shows that the most enthusiastic adopters of gen AI are millennial managers. Some 62 percent of employees aged 35 to 44 report high levels of expertise with AI, compared with 50 percent of 18-to-24-year-old Gen Zers and 22 percent of baby boomers over 65.7
CEOs can thus encourage these millennial change champions to mentor their peers in gen AI adoption and lead practice groups to share tips and tricks. Most critically, CEOs should lead by example, visibly using gen AI tools in their own work. This leads to more of a middle-out approach to change versus top-down or bottom-up.
At McKinsey, we established an adoption and engagement team to involve everyone in gen AI. First, we conducted segmentation analysis to classify user types and then tailored training to specific groups. Some approaches included creating Lilli Clubs composed of superusers who gathered to share tips and make suggestions on how to improve the platform and one-on-one training sessions for the most senior leaders. Lilli was also designed with multiple self-guided capabilities so users could “ask Lilli” how to best use it.
When it came to launching gen AI agents, we encouraged all colleagues to create their own, which in turn created a snowball effect of more employees taking part. While there are about 150 centrally developed agents, the firm’s federated development model (with built-in risk controls) has led employees to create nearly 17,000 additional agents so far. Lilli walks users through the agent building process and generates its own custom model instructions. Thus, employees are engaged directly in redesigning their own workflows. And what’s unique about a gen AI project like Lilli is how different it is from a traditional top-down, centrally controlled enterprise software rollout. Each new agent created comes with no additional tech debt or complexity for the technical organization to manage.
Singtel is another example of a company that has taken a skills-first approach to rolling out gen AI tools. In October 2024, Singtel created an AI Acceleration Academy in conjunction with Nanyang Technological University and the National University of Singapore to train more than 10,000 employees across many roles how to leverage data and gen AI in their workflows.8
For CEOs, the charge is clear: Plan for a company-wide reconfiguration today so that humans and AI together can achieve extraordinary outcomes tomorrow. This type of fundamental transformation won’t happen overnight, but business leaders who define a North Star can put their companies on a course for change.
Bringing people along for the ride is critical. Change management in the age of gen AI requires leaders to foster a culture of experimentation, where employees are not just passive recipients of new technology but active participants in it. By creating a work environment where gen AI is seen as a superpower rather than a threat, companies will have a far better chance of generating value from their investments. True transformation occurs when gen AI becomes an invisible but indispensable coworker.
