In this episode of The Bridgecast, host Scott Kinka welcomes Maya Sherman, an AI policy researcher, ethicist and technology diplomacy professional who advises international institutions and governments on AI governance. Maya unpacks the operational friction points facing organizations as autonomous models move from behind-the-scenes cybersecurity tools to client-facing products with increasing agency. This discussion provides business leaders with a clear roadmap for cutting through AI hype, safeguarding proprietary data and aligning tech investments with actual business needs.

Maya Sherman is an AI policy researcher, ethicist and technology diplomacy professional working at the intersection of AI governance, enterprise technology and the future of work. She serves as Israel's AI expert to the GPAI OECD Future of Work Committee and advises international institutions, startups and public bodies on building safe, commercially viable innovation ecosystems. Her career spans cybersecurity startup advisory across EMEA and Latin America, academic research in data ethics and tech policy work across Israel, the UK and India.

To find out how Bridgepointe Technologies helps businesses make IT decisions faster with world-class engineering support and ongoing guidance, head to https://bridgepointetechnologies.com/

[00:00:19] AI Is Not the Answer to Everything 

Maya Sherman opens with a direct challenge to the prevailing AI hype cycle, arguing that AI is not currently the solution to every business problem. This matters because many organizations are pressured to "match" their operations to AI usage rather than asking whether AI is even the right tool for the job. The core issue is that blind adoption, without evaluating fit, can cause more organizational damage than value. Her recommended approach is to twist the equation: start with the actual business problem and only introduce AI where it demonstrably helps. In practice, this means resisting the urge to deploy AI across every department just because the technology is available and trending. For leaders, this reframes AI strategy from a technology-first mindset to a problem-first mindset, reducing wasted investment.

[00:11:21] Rising AI "Agency" Is Now a Security and Governance Risk 

 Maya explains that AI models are gaining increased autonomy or "agency," meaning they can take independent actions with less human oversight than ever before. This is critical for businesses because models with greater agency can access, process, or leak sensitive data in ways that traditional security postures weren't built to handle. The challenge is that companies are racing to delegate more tasks to AI agents without fully understanding the downstream risks of reduced supervision. One key step is to treat AI deployment as a security initiative first, not just a productivity upgrade, budgeting specifically for the governance layer around agentic AI. She points to real-world examples like government-mandated restrictions on certain models due to national security concerns in countries like Australia and India. For business leaders, this means cybersecurity and AI strategy can no longer be treated as separate conversations.

[00:15:16] Regulation Is Shifting From Restriction to Enablement 

Maya describes a global shift in AI policy, moving away from strict, prevention-focused regulation (like the EU AI Act) toward more flexible frameworks designed to enable responsible AI use. This is important for business leaders because overly rigid regulation isn't the norm most governments are converging toward—adaptability and cultural context are becoming central to policymaking. The challenge regulators face is balancing innovation and national competitiveness against legitimate risks like election integrity and data misuse. Her recommended approach is for businesses to actively monitor regulatory evolution in the specific markets they operate in, since frameworks vary significantly between regions like the EU, India, Japan, and Israel. She cites Japan's focus on culturally adapted AI and India's back-and-forth on AI law as examples of this flexible approach in action. For global or scaling companies, this means compliance strategy must be regionally nuanced rather than one-size-fits-all.

[00:22:19] Most Companies Are Still in the AI Investment Phase, Not Profit Realization 

Maya agrees with the framing that most businesses remain in an investment and due-diligence phase of AI adoption rather than a profit-realization phase. This matters because boards and executives often expect immediate cost savings or efficiency gains, when in reality foundational work—security, data management, and process redesign—must come first. The problem this solves is the common mistake of rushing to scale AI without the governance infrastructure to support it responsibly, as seen in cases like Deloitte Australia being fined for AI-driven hallucinations in official reports. Her suggested method is to build robust internal processes and validate use cases before scaling spend on proprietary models or advanced tooling. She notes that sectors like recruitment and climate tech have shown strong ROI, while high-stakes fields like healthcare and law require far more caution. Leaders should calibrate expectations and investment timelines accordingly, rather than assuming AI spend will immediately translate into bottom-line results.

[00:28:08] Avoid the Token Trap by Testing Fast, Not Building Flashy 

Maya warns against the growing trend of companies building proprietary AI models or elaborate features simply because it looks technologically advanced, rather than because it solves a real business need. This is a critical insight for cost-conscious leaders since token consumption and infrastructure costs can spiral quickly when AI initiatives aren't grounded in actual use cases. The challenge is that AI hype pushes teams toward sophistication for its own sake, leading to wasted spend on capabilities that don't serve daily operations. Her recommended method is to use sandbox testing and rigorous evaluation cycles to determine what's actually required before committing resources, essentially "testing fast" rather than just "moving fast." She references how policy sandboxes have helped regulators and businesses alike assess real impact before full deployment. The payoff for leaders is a more economical, precise AI strategy that avoids the common pitfall of mistaking technological novelty for business value.

[00:38:50] Augmentation Over Automation Is the Sustainable Future of Work 

When pressed in the episode's rapid-fire segment, Maya firmly chose augmentation over automation as her vision for AI's role in the workforce. This distinction matters enormously for business leaders designing AI-driven transformation strategies, since framing AI as a replacement for workers versus a tool that enhances their capabilities leads to very different implementation approaches. The challenge many organizations face is defaulting to headcount reduction as the primary AI value proposition, which can erode morale and overlook AI's best use as a productivity multiplier. Her approach emphasizes designing workflows where AI handles supporting tasks while humans retain decision-making and judgment roles. This aligns with her broader prediction that AI will create new jobs and titles rather than eliminate the workforce entirely. For organizations building AI adoption roadmaps, this reinforces the value of involving employees in identifying use cases, since they're the ones best positioned to find where augmentation actually works.