How GenAI Is Reshaping Business: Key Lessons from the 2025 Wharton Global Forum

Aparna Chennapragada, Chief Product Officer at Microsoft, argued that AI is reversing traditional software development logic. In the past, product teams built detailed user interfaces and workflows first, and only then layered intelligence into the system. Now, she says, we’re seeing the opposite: “The model is eating the product.”

Instead of complex interfaces, products are becoming simpler “shells” built around powerful language models that handle everything from generating code to interpreting data to completing tasks. She calls this the shift to a “thin product, thick model” approach. For organizations, this means it’s time to stop thinking about AI as a plugin and start redesigning tools around it. It also means designing not just for human users, but also for AI agents that are consuming, interpreting, and acting on information.

Even though academic research shows meaningful productivity gains from generative AI in certain domains, many companies are struggling to translate that into tangible business value. Sajjad Jaffer, Founding Partner at GrowthCurve Capital, called this the “GenAI paradox”—where organizations are investing in AI, but seeing few short-term financial returns.

He compared it to the “Solow paradox” from the 1980s, when economists noted that computers were everywhere “except in the productivity statistics.” Just as it took nearly a decade for IT investments to show up in performance metrics, Jaffer argues we may need a similar runway for generative AI. To succeed, businesses must complement AI investments with changes to their workflows, training, and organizational structure.

As foundation models (like OpenAI’s GPT-4 or Google’s Gemini) become broadly available, companies can no longer rely on access to the best model as a unique advantage. Instead, panelists emphasized that proprietary data, especially customer interaction data or domain-specific records, is becoming the new moat – a means of protecting advantages from your competitors.

“Memory is a moat,” said Chennapragada. The longer a system works with a customer, the more it can personalize and improve performance. Lin Yuan, Chief AI Officer at Newfront, explained how industries like insurance have decades of unstructured, messy data (like PDFs and claims documents) that were previously unusable. With LLMs, that data can now fuel specialized models tailored to their domain, offering deeper insight and cost-effective automation.

Jaffer made a powerful point: “Data is a latent off-balance sheet item. It’s an asset, and a liability.” While many firms have valuable data sitting unused, unlocking its potential depends on leadership culture. He emphasized the importance of “data-first CEOs”—leaders who see data as core to the business strategy rather than an IT issue.

Ramon Richards, Chief Technology Officer of T. Rowe Price, added that tech leaders must also invest in non-technical priorities like data privacy, governance, and employee mindset shifts. “We don’t have unlimited capacity,” he said. “Prioritization matters.” To make real progress, organizations need leaders who champion experimentation while balancing risk.

As generative AI gets integrated into everyday workflows, the concept of management is evolving. Hosanagar predicted that every employee will soon be responsible for overseeing their own team of AI agents. These agents, digital assistants that complete tasks, write reports, generate content, or triage customer requests, will operate alongside humans, not just as tools, but as semi-autonomous collaborators.

The implication? Future professionals must learn how to direct, orchestrate, and evaluate the performance of non-human workers. That means new forms of training, management, and organizational behavior, skills we typically don’t teach in business schools today.

Xing Liu, Senior Partner at HongShan Capital Group, predicted the rise of “one-person unicorns”– ultra-lean companies powered by a single founder and dozens (or even hundreds) of AI agents. Thanks to generative AI and remote collaboration tools, startups are being born globally on day one, with teams distributed across time zones, enabled by powerful models.

This model challenges every traditional assumption about firm size, org charts, and productivity. Liu noted that AI-native startups are already operating with global talent, low overhead, and enormous leverage. For incumbents, the lesson is clear: smaller players are moving faster, and winning may mean adapting to their playbook.

For the latest episodes and updates on Creative Intelligence, be sure to subscribe on YouTube, Spotify, or Apple Podcasts.