Why AI Companies Need a New Model of Leadership

Artificial intelligence is forcing companies to rethink not just how work gets done, but who makes decisions, how success is measured and what leaders are ultimately responsible for. The qualities that help an AI company break through—technical brilliance, ambitious vision and the ability to move fast—can become liabilities when the business grows beyond the reach of a handful of experts. Suddenly, the decisions that once fit in a conference room carry consequences for customers, employees, infrastructure and public trust.

So what happens when the technology moves faster than the leadership model built to support it? Who gets to challenge a launch when the market demands speed? How do companies develop experienced talent when AI agents take on the work that once taught people the ropes? And how can leaders preserve the freedom to experiment without losing sight of the risks?

Members of the Senior Executive AI Think Tank, a curated group of experts in machine learning, generative AI and enterprise AI applications, are confronting these questions from across the AI landscape. Together, they offer a fresh look at what leadership must become as AI companies grow—and why the industry’s next phase may depend as much on how leaders think and organize as on what their technology can do.

“The AI industry has spent years rewarding leaders who build faster, raise more capital and promise bigger breakthroughs. That leadership model has an expiration date.”

– Steve Taplin, CEO of Sonatafy Technology

SHARE IT

Replace Innovation Theater With Measurable Accountability

For Steve Taplin, CEO of Sonatafy Technology, the excitement surrounding new capabilities can obscure whether a product actually solves a business problem.

“The AI industry has spent years rewarding leaders who build faster, raise more capital and promise bigger breakthroughs,” Taplin says. “That leadership model has an expiration date.”

He argues that leaders must distinguish technical achievement from business performance. 

“A successful AI demo is not a successful AI business, and adoption without measurable ROI is just expensive experimentation.”

The remedy, he says, is greater engineering discipline, independent validation, human oversight and clear ownership of results. 

“The industry’s winners won’t be those who deploy AI the fastest. They’ll be those who prove their AI deserves to be trusted.”

Make Governance an Engineering Capability

Raghu Para of Ford Motor Company argues that AI leadership must connect innovation with the operational rigor required to deploy technology responsibly.

“As AI companies scale, leadership must evolve from prioritizing technological breakthroughs and rapid deployment to balancing innovation with accountability, operational discipline and responsible governance,” Para says.

That balance cannot be achieved by researchers and executives working in isolation. 

“AI leadership can no longer remain concentrated among researchers and executives,” he says. “It requires coordinated decision-making across engineering, security, legal, product and domain experts.”

Para recommends establishing clear ownership of AI risks, reliability standards and independent evaluation mechanisms. He also urges leaders to move beyond model benchmarks and adoption rates to assess business value, cost efficiency, explainability and real-world safety.

The goal is to make governance part of the development process rather than a late-stage compliance exercise. As McKinsey’s research on AI organizational change indicates, companies are increasingly redesigning workflows and elevating AI governance as they seek meaningful returns.

Redesign the Talent Pipeline for an AI Workforce

Mahendran Vasagam, Principal Member of Technical Staff at Slack Technologies, sees a fundamental shift in how organizations create value: AI systems are becoming part of the workforce itself.

“AI companies are the first industry where the product is also part of the workforce,” Vasagam says. He notes that AI agents can multiply the output of small engineering teams, undermining traditional measures of organizational scale.

“Ten engineers with agents can now ship what once took 200,” he says. As a result, leaders must rethink how they measure productivity and assign accountability when a single person supervises work performed by many agents.

Vasagam also warns of a less visible consequence: the erosion of entry-level learning opportunities. 

“Juniors became seniors by doing the routine work agents now handle,” he says. “Unless leaders deliberately redesign how people learn, the industry will face a shortage of experienced judgment within a decade, just when it needs it most.”

Pair Relentless Innovation With Ethical Stewardship

Anisha Manvatkar, Sr. Technical Program Leader at NVIDIA, believes AI leaders must serve two complementary roles as their organizations grow.

“Successful leaders will need to continuously fill the hopper with new ideas and new frontiers,” Manvatkar says. Innovation cannot become an excuse for complacency, particularly in an industry where capabilities and competitive boundaries change quickly.

But the other responsibility is equally important. Leaders must become “full-time champions, stewards and mentors of teams regarding ethics and values,” she says.

Manvatkar warns that ethical boundaries can blur without deliberate attention. Leaders therefore need to embed values into everyday decisions, mentor teams through ambiguous situations and create an environment in which employees recognize emerging ethical concerns before they become costly failures.

The lesson is not to slow innovation indiscriminately. It is to ensure that the people advancing AI capabilities also understand the responsibilities that accompany them.

“Leadership has to shift from being the smartest person in the room to building a room that argues well.”

– Hastimal Jangid, Co-Founder of RankRabbit AI

SHARE IT

Build Leadership Teams That Challenge Assumptions

Hastimal Jangid, Co-Founder of RankRabbit AI, believes the industry must reconsider who gets to influence consequential decisions.

“Early AI companies were led by the people who understood the technology best,” Jangid says. “That worked while the product was a research project.”

As AI businesses expand into areas involving security, employment, regulation and public trust, technical knowledge alone is insufficient. Leaders need teams that can examine problems from different perspectives and challenge assumptions before a decision becomes difficult to reverse.

“Leadership has to shift from being the smartest person in the room to building a room that argues well,” Jangid says.

That requires giving safety, policy and other specialist voices genuine authority—not simply inviting them to comment after a direction has already been set. It also requires leaders willing to change course when evidence contradicts their assumptions, even when doing so delays a launch.

“Companies that build this early will be believed when something goes wrong,” he says. “The ones that don’t will spend years explaining themselves to regulators and customers.”

Distribute Decision-Making Before Growth Creates Bottlenecks

Daria Rudnik, Team Architect and Executive Leadership Coach at Daria Rudnik Coaching & Consulting, argues that AI companies cannot scale successfully if every important decision still flows through a small executive group.

“Right now, too many important decisions still go through a small group at the top,” Rudnik says. What works in a compact research team becomes a bottleneck when an organization grows.

“AI tools help teams work faster,” she says. “But if teams still need approval for every important call, they just get stuck at the top sooner.”

Her solution is to establish clear decision rights, give accountable owners authority to act and encourage employees to raise problems early. Executives still set direction and define boundaries, but they should not become the default approval point for every operational choice.

“Companies that do this will scale faster and keep their most experienced people. Those are usually the people who burn out first.”

Replace Command-and-Control With Adaptive Teams

Uttam Kumar, Engineering Manager at American Eagle Outfitters, argues that AI-enabled organizations need structures that allow teams closest to the work to respond quickly to changing conditions.

“Modern retail AI enterprises must completely rethink traditional command-and-control leadership hierarchies in favor of agile, agentic team structures,” Kumar says.

For retailers, that means empowering domain leaders to test and refine solutions for physical stores without waiting for multiple layers of approval. Local teams often understand operational constraints and customer needs that centralized decision-makers may overlook.

Kumar’s principle is that “speed of adaptability is the ultimate competitive advantage in dynamic markets.”

That does not mean abandoning oversight. Leaders must define strategic objectives, acceptable risks and performance standards while allowing teams to determine how best to achieve them.

“This fundamental management shift will transform the industry by accelerating product deployment cycles from months to days across complex brick-and-mortar ecosystems.”

Develop Leaders Who Can Connect the Whole AI System

Dhyey Mavani, a Generalist in Agentic AI at Abridge, emphasizes the growing importance of cross-disciplinary leadership.

“AI companies will need leaders who can integrate research, engineering, product, infrastructure and policy rather than optimizing each in isolation,” Mavani says.

As AI systems become more capable, decisions in one area can create consequences elsewhere. A model improvement may increase operating costs; a product feature may introduce reliability concerns; a deployment decision may raise questions about safety or societal impact.

“The hardest decisions increasingly involve trade-offs across capability, reliability, economics and societal impact,” he says.

Mavani believes this complexity could favor smaller leadership groups composed of technically sophisticated operators with broad judgment. These leaders must understand enough about each function to recognize trade-offs, ask informed questions and align teams around shared outcomes.

Manage AI Infrastructure Like a Long-Term Operation

Bhubalan Mani, Leader in Supply Chain Technology and Analytics, says AI businesses must coordinate computing capacity, memory, energy and infrastructure investments that cannot be adjusted as quickly as software road maps.

“AI leaders must think like plant managers, not product managers,” Mani says.

He argues that infrastructure decisions require disciplined planning rather than a reflexive response to the latest demand forecast. 

“Gartner projects data center power demand rising from 132 GW this year to 290 GW by 2030,” he says. “That is capacity planning, not sprint planning.”

Mani recommends sales and operations planning discipline, including scenarios for both oversupply and shortages. He warns that small changes in demand expectations can trigger excessive investment across supply chains.

“The next AI crisis may arrive as a write-down, not a bug,” he says.

“AI leaders should spend more time with the people who encounter failures first. ”

– Nikhil Jathar, CTO of AvanSaber Inc.

SHARE IT

Give Frontline Teams the Authority to Stop Failures

Nikhil Jathar, CTO of AvanSaber Inc., argues that leadership must shorten the distance between an AI failure and the person empowered to fix it.

“AI leaders should spend more time with the people who encounter failures first,” Jathar says. “As companies grow, customer problems can pass through several management layers before reaching someone who can act.”

He points to Toyota’s practice of empowering workers to stop production when they detect a problem as a useful model. 

“AI teams need comparable authority to pause a faulty release, backed by leaders who reward early reporting,” he says.

Jathar also recommends pairing headline performance results with a review of failed customer tasks. That practice can reveal weaknesses that aggregate metrics obscure and help leadership connect technical progress with actual service quality.

As AI becomes embedded in daily operations, the ability to identify, escalate and correct failures quickly will become a defining leadership capability.

How to Rethink AI Leadership

• Measure business outcomes, not just technical progress. Define clear return-on-investment targets and require evidence that AI systems deliver reliable, useful results.
• Build governance into development workflows. Assign risk owners, establish reliability standards and involve engineering, security, legal and product teams early.
• Redesign career development around human judgment. Create supervised learning opportunities that help junior employees build the expertise automation may otherwise bypass.
• Make ethics part of everyday leadership. Pair ambitious innovation goals with active mentorship and clear expectations for responsible behavior.
• Give constructive dissent real authority. Ensure safety and policy specialists can influence decisions and that leaders respond when evidence challenges a preferred plan.
• Clarify decision rights across teams. Delegate operational authority within defined boundaries so employees can act without unnecessary escalation.
• Empower teams closest to customers and operations. Let domain leaders adapt solutions quickly while holding them accountable for agreed-upon outcomes.
• Develop cross-functional executive judgment. Bring technical, commercial, infrastructure and policy perspectives together when decisions involve competing priorities.
• Plan infrastructure against multiple scenarios. Use demand forecasts, capacity constraints and downside scenarios to avoid costly overinvestment.
• Make failure reporting a leadership priority. Give frontline teams the authority to pause problematic releases and review failures alongside successful performance metrics.

The Next AI Breakthrough May Be a Leadership One

The AI industry knows how to build faster. The harder question is whether it knows how to grow wiser. As AI systems take on more responsibility and companies make increasingly consequential decisions, leadership will be tested not just by what organizations can create, but by what they can recognize, prevent and learn from along the way. The next defining moment for an AI company may not be a breakthrough model or a record-breaking launch. It may be a leader who pauses a release, empowers an overlooked voice or chooses long-term trust over short-term momentum.

That shift creates an opportunity to redefine what winning looks like in AI. The companies that endure will be those that make sound judgment a competitive advantage—building organizations where innovation and accountability reinforce each other, people can challenge assumptions and speed never comes at the expense of responsibility. In an industry racing to shape the future, the leaders who matter most may be those who know when to accelerate, when to question and when to stop.