Copyright International Business Times
In an exclusive interview, An Business operations & Analytics leader challenges the replacement narrative and makes the case for workforce development over cost-cutting. IBT: Goldman Sachs warns 300 million jobs could vanish to AI¹. Wall Street banks plan to eliminate 200,000 roles in the next few years¹. Stanford research shows early-career workers already facing a 13% employment decline in AI-exposed fields². Yet you're arguing the "push for cost savings and human replacement is overblown." What makes you so confident? Aayush Agarwal: The headlines miss the nuance. Goldman Sachs' projection emphasizes jobs that "could" be lost—it's conditional, not certain⁴. The World Economic Forum's forecast of 92 million jobs displaced also projects 170 million jobs created—a net gain of 78 million positions¹,⁵. Brookings research from September 2025 found "stability, not disruption" in aggregate labor markets⁷. PwC's 2025 AI Jobs Barometer shows AI linked to fourfold productivity growth and a 56% wage premium in AI-exposed occupations⁸. The story isn't mass unemployment—it's workforce transformation. IBT: But companies are already cutting. What about the workers experiencing real displacement right now? Agarwal: The transition will be difficult, especially for early-career workers. Stanford's 13% employment decline in AI-exposed fields is real and concerning²,³. But it's concentrated among junior workers, not the entire workforce. Older workers are growing in the same firms, pointing to a shift in skill requirements³. These workers need to focus on building AI-fluency as a basic skill—like knowing how to type when typewriters were adopted. Educational institutions and companies must prioritize training to ease this transition. IBT: You've seen both sides of this. At Philips Healthcare a decade ago, you could have automated away service engineers. Why didn't you? Agarwal: We could have pushed for near-complete automation and elimination of most service engineers. But that would have cut a strong trust channel between hospitals, lab technicians, and Philips itself. Instead, we pioneered machine learning-driven IoT analytics for predictive maintenance of medical devices—MRIs, CT scanners, nuclear medicine equipment. The results: $9M in annual savings, 25% increase in medical scans per day, 12-hour reduction in average service time. Service engineers transitioned from reactive firefighting to proactive consultation. The preserved trust channel proved far more valuable than labor cost savings—it cemented long-term contracts and customer loyalty that automation would have destroyed. IBT: Meanwhile, companies like Klarna are learning this lesson the hard way. Agarwal: Exactly. Klarna claimed its AI chatbot could do the work of 700 customer service agents. By May 2025, they began rehiring after quality slipped and customer complaints spiked⁹. Fifty-five percent of companies that executed AI-driven layoffs now regret it⁹. Failed AI initiatives cost Fortune 500 companies an average of $15.7 million each in 2024¹⁰. Ninety-five percent of organizations investing in generative AI are getting zero return¹¹. The replacement strategy is failing in real-time. IBT: What are the augmentation leaders doing differently? Agarwal: They're elevating human work. Brightview Senior Living uses AI-based fall detection that augments caregivers, reducing response time from 40 minutes to under 3 minutes¹³. Capital One built generative AI tools that help call agents access information faster while improving empathy¹³. Camden Property Trust uses speech analytics so agents can focus on the human aspects of service¹³. These companies understand AI isn't about eliminating tasks—it's about elevating what humans can accomplish. IBT: At Meta, you developed LLM-powered Digital Twins that achieved 93% human review consistency. Were you replacing reviewers or augmenting them? Agarwal: Augmenting, absolutely. I manage operations for 14,000+ content reviewers. The Digital Twins enable comprehensive policy evaluation and improved accuracy by 11% for AI-assisted human review. We're making reviewers more effective, not eliminating them. The same principle applies across industries. IBT: How do you prove to boards that augmentation delivers better returns than replacement? Agarwal: You need rigorous metrics. I recommend tracking three things: decision velocity—how quickly you move from insight to action—quality-gated by decision reversal rate and financial forecast accuracy. Recent research identifies decision velocity as the new metric for enterprise AI success¹⁴,¹⁵. Companies using AI to make faster decisions systematically outperform competitors¹⁶. Over longer horizons, track the growth rate of revenue per employee, not just the static number. This captures whether performance is sustained. IBT: What does that look like in practice? Agarwal: Apple leads at $2.38M per employee, Meta at $2.19M, Nvidia at $2.06M¹⁷,¹⁸. But trajectory matters more than snapshots. Nvidia scaled from 13,500 to 29,600 employees while revenue grew 126%—sustained augmentation¹⁷. Amazon's low $410K per employee reflects the replacement trap¹⁷. Product-led growth companies that augment sales hit $350K median revenue per employee versus $284K for traditional firms¹⁹. The data validates augmentation's superiority. IBT: What about the risk dimension? AI making contracts, processing agreements—where's the human role? Agarwal: Risk management will drive augmentation over replacement. Imagine contracts created, evaluated, and processed entirely by AI. Any mistake could have massive impacts. AI contract review tools can reduce review time by 40%²⁶, but legal experts warn about inaccuracies, omitted clauses, and outdated legal knowledge²⁷,²⁸. Questions about whether AI can legally bind companies remain unclear²⁸. High-stakes decisions—contracts, diagnoses, financial forecasting—will require human-AI collaboration, not just for ethics but for legal liability and risk mitigation. We'll see regulations, even if they lag. IBT: Universities are scrambling to respond. Is the education system ready? Agarwal: It's getting there. Ninety percent of U.S. college students now use AI academically²⁰. MIT is launching Universal AI Fluency in 2026, Ohio State committed to ensuring every undergraduate graduates with AI fluency by 2029²⁰,²¹. The federal government mandated AI literacy for federal employees through Executive Order 14179 and OMB Memorandum M-25-21²². The U.S. Department of Education emphasizes hands-on experimentation, not just theory²³. Companies investing heavily in development retain employees 5.4 years versus 2.9 years for low-investment companies²⁴. Seventy percent of workers would leave for organizations known to invest in development²⁴. Training isn't optional—it's competitive advantage. IBT: You're betting on competitive pressure to force laggards to adapt. How long will that take? Agarwal: I acknowledge a 3-5 year learning period of hiring freezes, failed experiments, and disruption⁴. Not all companies will choose augmentation initially. But those that do, and learn from others' failures, will systematically outperform. Competitive pressure will force laggards to adapt or fail. They'll suffer costs from failed projects, rehiring expenses—$30,000-$45,000 per $60K role³⁴—and lost competitive ground. Companies that cut too aggressively face replacement costs up to 2x annual salary³⁵,³⁶. Meanwhile, AI-augmented workers show 66% productivity gains when properly supported³⁷. IBT: What's your message to C-suite executives reading this? Agarwal: Develop your employee pool with AI-fluency and augmentation. Tech leaders must share lessons publicly, advocate for industry standards—ISO/IEC 42001, NIST's AI Risk Management Framework²⁹,³¹,³². Internally, redesign systems for human-AI collaboration, invest in workforce development, establish ethical governance. McKinsey found 92% of companies plan to increase AI investments, yet only 1% consider themselves mature in deployment³³. The gap between aspiration and execution is massive. And remember this: it's better to train your employees even if they leave, than to not train them and have them stay. That's not just ethics—it's strategic advantage. Richard Branson said it best: "Train people well enough so they can leave. Treat them well enough so they don't want to."³⁸,³⁹ Companies making that choice today are building the competitive moats of tomorrow. AI in business is an aid, not a replacement. Our collective responsibility is gaining fluency on this technology—especially for people entering the workforce, and for businesses wanting to improve their market position through value creation rather than cost reduction. Aayush Agarwal is Senior Program Manager for AI Operations at Meta Platforms, where he manages AI systems protecting Meta's ad revenue and oversees operations for nearly 15,000 content reviewers. He previously led AI/ML implementations at Philips Healthcare and Deloitte, and holds an MBA from the University of Michigan Ross School of Business. -----SOURCES
