Building a More Competitive and Safer AI Ecosystem

Artificial intelligence is often framed as a race: faster models, bigger investments, larger datasets and more powerful infrastructure. But beneath the headlines lies a more consequential question for business leaders, policymakers and investors alike: Who gets to compete?

A growing share of the AI ecosystem is controlled by a relatively small number of organizations with access to the world’s largest compute resources, proprietary datasets and distribution channels. This means the debate is no longer simply about what AI can do but about whether the next wave of innovation will emerge from an open marketplace of ideas or from a handful of dominant ecosystems.

To explore that question, we asked members of the Senior Executive AI Think Tank—a curated community of leaders specializing in machine learning, generative AI, digital transformation and enterprise AI applications—what single rule they would change to improve AI competition.

While their recommendations differ, a clear theme emerges: The future of AI should be shaped by innovation, trust and customer value rather than lock-in, opacity or concentrated control. The following insights offer a timely look at how technology and business leaders believe a more competitive—and in many cases safer—AI ecosystem can be built.

Break Down Data Silos and Let Customers Move Freely

Salim Gheewalla, Founder and CEO of utilITise, spends his time helping organizations simplify technology operations and eliminate unnecessary complexity. Drawing on more than 15 years of experience leading large IT and cybersecurity organizations, he believes the biggest obstacle to healthy AI competition is data lock-in.

“AI should compete on outcomes, not on who can lock up the most data,” Gheewalla says.

He argues that many organizations possess valuable data but struggle to use it effectively because it remains trapped inside disconnected platforms and proprietary ecosystems.

“Giving businesses greater control over moving and utilizing their own data would create a more competitive market, accelerate innovation and reduce dependency on a handful of dominant providers,” he says.

For Gheewalla, portability is not merely a competition issue. It is also a risk-management issue.

“It wouldn’t just create a healthier AI ecosystem—it would also create a safer one by reducing concentration risk and giving customers more choice.”

Hold Companies Accountable for Customer Outcomes

Ramendra Rout of Five9 approaches the issue from the customer experience perspective.

While many organizations celebrate AI adoption metrics, Rout argues that successful implementation should ultimately be measured by customer outcomes.

“The quality of the outcome is as important as efficiency,” Rout says.

He points to organizations that have replaced customer-facing functions with basic chatbots while failing to improve customer experiences.

“Several businesses have done a shoddy job of replacing their front office with basic chatbots in the name of AI,” he says. “That barely helps end customers in getting their issues resolved.”

Rather than focusing solely on technical standards, Rout proposes a governance mechanism that gives customers a meaningful voice.

“There should be a governing forum for customers to raise concerns where businesses have simply created a machine-enabled front office and every interaction hits a brick wall,” he says.

Such accountability, he argues, would discourage superficial AI deployments designed primarily for public relations or investor optics.

“This will ensure that companies really deliver on the promise of AI and not merely implement AI, gather PR, and satisfy investors and markets, all at the expense of customer experience and workforce reduction.”

Make Switching Providers Easy

Rishabh Dave, Head of Product at BuildOps, believes competition flourishes when customers can leave.

“I would require AI platforms to support standardized data portability and interoperability,” Dave says.

His recommendation echoes broader concerns throughout enterprise software markets where switching costs can discourage innovation and reduce customer choice.

“Companies should compete on the quality of their products, not on how difficult they make it for customers to leave,” he says.

Dave also sees a direct relationship between competition and trust.

“When customers can move freely, providers must continuously earn trust through performance, reliability and transparency rather than relying on lock-in.”

Ultimately, he argues, the healthiest market rewards excellence rather than barriers.

“In the long run, the healthiest AI market is one where the best products win—not the most closed ecosystems.”

“Greater transparency would create a more competitive ecosystem and a safer one because innovation scales best when participants can compete on capability.”

– Richie Adetimehin, AI Advisory and Transformation Delivery Consultant at Visani America

SHARE IT

Make Trust a Competitive Advantage

Richie Adetimehin, an AI Advisory and Transformation Delivery Consultant at Visani America who helps enterprises govern AI adoption at scale, focuses on transparency.

“If I could change one rule, it would be this: Any AI system above a defined capability threshold must disclose the provenance of its training data, major model dependencies and decision accountability structure,” he says.

According to Adetimehin, competition in AI depends on three primary ingredients: “compute, data and trust.”

“Today, trust is the least transparent,” he says. “Incumbents often win by controlling scarce resources.”

Greater disclosure requirements would help level the playing field.

“Greater transparency would create a more competitive ecosystem and a safer one because innovation scales best when participants can compete on capability.”

“Today, a handful of firms control the raw material of intelligence, and whoever owns the data owns the future. That isn’t a market. It’s a moat.”

– Aditya Vikram Kashyap, Vice President of Firmwide Innovation at Morgan Stanley

SHARE IT

Open the Data, Not the Models

Aditya Vikram Kashyap, Vice President of Firmwide Innovation at Morgan Stanley, advocates one of the boldest proposals among the group.

“I’d mandate open access to foundational training data—not the models but the data beneath them,” Kashyap says.

He believes data ownership increasingly determines competitive outcomes.

“Today, a handful of firms control the raw material of intelligence, and whoever owns the data owns the future. That isn’t a market. It’s a moat.”

Opening access to training corpora while maintaining privacy protections would dramatically increase participation, he argues.

“Open the corpus, with privacy and provenance enforced, and you let a thousand competitors build where only five can build now.”

Kashyap also rejects the assumption that competition and safety are opposing goals.

“A field with many players is harder to capture, easier to audit and far less fragile than one ruled by a few,” he says. “Concentration is the real risk. Diversity is the real safety.”

Build a Shared Data Commons

Sathish Anumula, Enterprise and Business Architect at IBM Corporation, also sees data concentration as the central challenge.

“I would create a standardized, legally protected framework for AI training called the ‘Data Commons,'” Anumula says.

He argues that today’s AI market risks becoming dominated by “a handful of tech giants” with “structural monopolies on proprietary data and large user feedback loops.”

His solution combines legal clarity with broader access.

“Establishing clear statutory safe harbors for fair-use training data and public data trusts will significantly reduce the entry barrier for startups and researchers who are priced out by litigation risk and data hoarding.”

For Anumula, democratizing data access ultimately allows innovation to emerge from merit rather than scale.

“Market forces, not concentrated capital, should drive AI progress.”

“Organizations should be able to move their data, models, prompts and workflows across providers through standardized interfaces without sacrificing security or intellectual property protections.”

– Dileep Rai, Manager of Oracle Cloud Technology at Hachette Book Group (HBG)

SHARE IT

Create Standards for AI Portability

Dileep Rai, Manager of Oracle Cloud Technology at Hachette Book Group (HBG), argues that interoperability should extend beyond data alone.

“I would require interoperability and controlled data portability between major AI platforms,” Rai says.

Organizations increasingly build business processes around prompts, workflows and AI-enabled systems. Rai believes those assets should remain portable.

“Organizations should be able to move their data, models, prompts and workflows across providers through standardized interfaces without sacrificing security or intellectual property protections.”

Such standards would stimulate innovation while reducing dependence on dominant providers.

“This would lower switching costs, encourage innovation from startups and researchers and reduce dependency on a handful of dominant players.”

The benefits extend beyond competition.

“Diversity in AI providers reduces systemic risk, fosters transparency and prevents any single entity from exerting disproportionate influence over the future of AI.”

Protect Open Models and New Entrants

Lynn Comp, Head of AI Center of Excellence at Intel, focuses on preserving competitive opportunities for emerging innovators.

“Maintaining the rights to run open models and open weights on ‘owned’ infrastructure will be increasingly critical,” Comp says.

Without those protections, she warns, the market could consolidate around a handful of dominant providers.

“This way, the market continues to see innovation from new entrants rather than consolidating to four vendors.”

Comp also believes policymakers should pay closer attention to pricing strategies.

“Limitations on upfront loss-leader pricing for digital services may be needed.”

In her view, transparency around pricing practices could help ensure that competition remains fair and sustainable as AI markets mature.

Prevent Control of Infrastructure Bottlenecks

Venkata Kondepati, Manager of Data Architecture and Engineering at Ascentt, focuses on access to critical infrastructure.

“I would require fair, auditable access to essential AI infrastructure when a dominant platform controls it,” Kondepati says.

His definition of infrastructure extends beyond compute, however.

“Compute, cloud credits, model distribution channels and high-value data interfaces” all deserve scrutiny, he says.

Importantly, Kondepati is not advocating for weakened intellectual property protections.

“This should not force every model to be open or weaken IP.”

Instead, he wants to prevent gatekeepers from leveraging control over bottlenecks to suppress competition.

“AI policy should reward innovation, not ownership of the bottleneck.”

Slow Down Before AI Makes the Final Call

Goran Paun, Principal and Creative Director at ArtVersion, takes a different approach.

Rather than focusing primarily on data or infrastructure, he argues that competition currently over-rewards speed.

“I would build in a generative pause,” Paun says. “Not a pause on innovation, but a practical pause before AI outputs move directly into business, legal, creative or operational decisions.”

According to Paun, organizations increasingly equate faster automation with better outcomes.

“Too much of the current competition rewards speed: who can generate faster, automate faster and ship faster.”

His proposal introduces greater human oversight.

“A healthier rule would require clearer human checkpoints, provenance and accountability around AI-generated work.”

The result, he believes, would strengthen both safety and competition.

“The best companies would not be the ones that generate the most. They would be the ones that know what to trust, what to question and when a human decision still matters.”

10 Actions Leaders Can Take Today

• Prioritize data portability in every AI procurement decision. Ensure contracts preserve your ability to move data and avoid long-term lock-in.
• Measure AI success by customer outcomes, not deployment metrics. Efficiency gains matter only when they improve the customer experience.
• Reduce switching costs wherever possible. Open standards create leverage and encourage vendors to continually earn trust.
• Demand transparency from AI providers. Understanding training data sources and accountability structures strengthens governance.
• Treat access to data as a strategic issue. Competitive advantage increasingly depends on how organizations access and use information.
• Support shared innovation ecosystems. Data commons and public-interest datasets can expand participation and accelerate progress.
• Build portability beyond data alone. Consider prompts, workflows and models as strategic assets that should remain transferable.
• Preserve opportunities for open models. Open ecosystems help sustain competition and innovation.
• Watch for infrastructure bottlenecks. Access to compute and distribution channels can be as important as access to data.
• Maintain meaningful human oversight. The most effective AI deployments combine automation with accountable decision-making.

The Real Stakes in AI Competition

If there is one takeaway from these experts, it is that the future of AI competition is not really about AI at all. It is about access, accountability and trust. Whether they advocate for data portability, open training corpora, infrastructure access, transparency requirements or stronger human oversight, the members of the Senior Executive AI Think Tank are ultimately arguing for the same principle: Innovation should be earned through better outcomes, not protected through lock-in, opacity or control of critical resources.

The decisions being made today—by regulators, technology providers and enterprise leaders—will help determine whether AI evolves into a diverse ecosystem of innovators or a market dominated by a small number of gatekeepers. The organizations that thrive in the years ahead will likely be those that look beyond the capabilities of today’s models and focus instead on building AI strategies that are transparent, adaptable and trusted. As the technology continues to advance, the most important competitive advantage may not be having the biggest model—it may be creating an environment where the best ideas have the opportunity to compete.