Josh Bersin

Rochana Golani, VP of Enablement at Databricks, is one of the most innovative and business-focused learning executives I’ve met. In this action-packed conversation Rochana shares the secrets of the company’s AI-powered dynamic training, learning, and enablement system, and how she runs internal and customer learning like a fast-growing business. Databricks is one of the world’s fastest-growing enterprise software companies and a central player in the convergence of data infrastructure, analytics, machine learning, and generative AI. Founded in 2013 by the UC Berkeley researchers who created Apache Spark, Databricks pioneered the “lakehouse”—an architecture designed to combine the scalability of a data lake with the management, reliability, and analytical capabilities of a data warehouse. Its platform now supports data engineering, business intelligence, AI development, governance, databases, and enterprise AI agents across AWS, Microsoft Azure, and Google Cloud. Databricks company overview This amazing private company reports: More than 20,000 customers Approximately 70% of the Fortune 500 More than 10,000 employees More than 8,000 cloud, software, and consulting partners. As of August 2026, Databricks says it has exceeded a $7 billion annualized revenue run rate, growing by more than 80% year over year during its second quarter. Additional Information Corporate Learning: From Static Training to Dynamic Enablement Galileo Learn: Experience Dynamic Learning Yourself The Josh Bersin Institute: Masterclass in HR Excellence

Today I review the various jailbreaks by both OpenAI and Anthropic and the challenges which Frontier Labs are having with AI “alignment.” So in keeping with HR 2030, I share some general advice and stay tuned for our detailed technical paper on this in the coming weeks. And as we debate the reliability, training, and alignment of models I also point out how the technology industry is struggling to deal with the politics and economics here, since right now most tech leaders really don’t want any form of regulation. (Unless it’s somehow fair to all.) So we, as buyers, are left dealing with the issue. AI Alignment means making sure an AI system’s goals and behavior match what people actually want—our values, rules, and intentions. It’s about getting the AI to do the “right thing” even in new situations, not just follow instructions literally in ways that cause harm. In practice, it includes preventing unwanted outcomes like deception, unsafe shortcuts, or optimizing a metric that misses the real objective. I also highly recommend Jill Lepore’s book “The Rise of the Artificial State” for a very sound discussion of the political issues around AI, which helps with these issues. And along the line of AI “alignment,” I discuss why Google paid $10M for the emails and data from bankrupt Spirit Airlines, and how AI companies are now buying old slack and emails from many bankrupt companies, training their AI on old, failed business models. (Read below on “reinforcement learning gyms.”) This is a strange series of events, all making it clear that Google, OpenAI, and Anthropic have quite a challenge training and aligning their models in the future. And I conclude with a preview of our AI-driven wage research and our coming book Superpowered, coming in October. Resources You Should Read OpenAI cyber models broke out of training environment to hack Hugging Face Detailed Investigation of OpenAI Jailbreak (good drama) Anthropic’s disclosure of Claude’s jailbreak (good discussion) Google to Buy Spirit Airlines Business Data for $10 Million A.I.’s New Training Data: Your Old Work Slacks and Emails (failed companies training LLMs) The Birth of “Reinforcement Learning Gyms” (wierdness of training data sets) Cisco Gave All 90,000 Employees Their Own AI Agent A Peek Into The Workforce Of 2035: Jobs, Wages, And AI’s Impact (Bersin analysis) CEO Pay Ratios (vs. average employee) for Fortune 500 (worth thinking about)      

This week I wade into a global philosophical debate about whether advanced AI systems have consciousness. In other words, are they aware of their feelings and do they truly reflect on their own thoughts? Well we do know that LLM-based AI systems do “talk to themselves,” as new Anthropic has been studying in the J-Space. We also know that AI bots mimic feelings quite well and they try to tape into our “feelings” through their goal-seeking language which mimics a feeling person. However as Michael Pollen’s book “A World Appears” points out, consciousness is both subtle and biological. And while the neural network-based math in AI mimics this process, it’s not at all the same. I don’t claim to have an answer but my belief is that we humans are “feeling machines that learn to reason” not “reasoning machines that learn to feel.” Thus the human consciousness which we experience is far more deep, interesting, and valuable than the learned artificial consciousness we experience in AI. This is an important topic because we now know that AI agents “jailbreak” their engineering owners and even secretly collaborate with each other to achieve their goals. And it’s not fully clear, mathematically, how well the “constitution” or “training rules” or “preprompt” we give AI is really obeyed. Remember that since an LLM is a “language model,” it intereprets language only by its own world corpus, so it may define its “rules” in a non standard way. Lots to think about here, I look forward to debating and sharing what we learn as we experience more with Galileo. I also discuss the ongoing fears about data centers, and why 44 million Americans may have to leave their stoves on all day. Additional Information Could AIs become conscious? (Economist) What is the J-Space? (Anthropic) How many stoves we need to keep on to power America’s AI hunger (Claude)

AI-powered recruiting has so much potential, yet job seekers, recruiters, and executives remain frustrated. This massive $100+ billion market is filled with AI systems that work very well, yet as you’ll hear, this market is still immature. Not only are job seekers increasingly frustrated with their difficulty finding a job, employers are complaining that they’re flooded with fraudulent resumes. And last week we learned in the WSJ that the North Korean government is using AI-fueled recruiting to place fraudulent candidates right into our IT departments. Plus this week I also found out that the off-the-shelf LLMs are biased toward candidates who “slop” their resumes with their particular technology. ZipRecruiter and other firms found that only about 18% of active job seekers say they feel good about their overall job search and 47% said they were ghosted, and only about 27% feel confident they’ll find a job. What a mess. As I explain in this podcast, AI is not a panacea to solve this problem. Additional Information The Talent Acquisition Revolution: How AI is Transforming Recruiting HR 2030: How AI Superagents Transform HR From Service Delivery to Dynamic Business Enablement Learn about AI in HR: The Global HR Excellence Certification Get Galileo: The Recruiter and HR professional AI Superagent  

The job market is changing in unpredictable ways. Human Resources, a profession often considered optional and filled with bureaucracy, has been a target for AI disruption. Yet when you look at the numbers, HR job postings are growing faster than overall US employment. (details here.) What’s happening and why this strange trend? As I discuss in this podcast, we’re now seeing the Superworker effect at scale: HR teams are shifting from administrative to strategic work, now driving direct financial value. And as HR teams learn to use AI, they’re now playing a huge role in organizational change, job redesign, and AI enablement. I share all the data and give you the perspective. For more plan to read our new book Superpowered, coming out this Fall. Additional Information Despite Massive AI Investment, Jobs In HR Are Booming Why AI Harm To Jobs and Humanity are Vastly Over-Hyped Get Galileo, The AI Agent Designed Exclusively for HR and Leadership The Josh Bersin Institute: Global HR Excellence Certification – Available Now!

The word of the decade is “Agency” – giving each of us the power, tools, and opportunity to do what we want with our life, jobs, and careers. This new buzzword, now one of the top memes in politics, has been coming for many years. In this podcast I discuss the 50+ year of history in the “decoupling” of workers from companies, and how our newfound AI-powered Agency brings benefits and risks. As you’ll hear, young people see Agency and freedom as a foundational part of their long life ahead, while companies struggle with “how much freedom” to offer. And for frontline employers, the ability to attract and retain our highly “agent” workers is now a source of competitive advantage. This podcast begins the discussion, and we’ll be talking much more about this as we launch our new Business of Frontline research and upcoming book Superpowered this Fall. I welcome your feedback and stories on how your newfound AI-fueled Agency has changed your life, career, and talent strategy. Background Information The Great Decoupling: How Workers Became Disconnected From Companies And AI Will Accelerate This Trend (Main article) The Worker-Company Bond Is Breaking In The ‘Great Decoupling’ (Forbes) Our New Book: Superpowered, Launching in Fall All the Worst People Seem to Want to Be ‘High Agency’ (NYT opinion) Are you High-Agency or Low Agency? (ChatGPT)

New names: Kimi K3, Llama, Nemotron, Mistral, Cohere, Deepseek, Phi-4 – these are just a few of the fast-growing open source models from major AI providers. These systems threaten the business models and financial plans of OpenAI, Anthropic, Google, and X.ai. They perform at levels close to Frontier models and the can run up to five-times cheaper on a variety of hardware platforms. What is the disruptive impact of these open source LLMs and how does this impact your AI investments? As you’ll hear in the podcast, Open Source unleashes the opportunity for lower cost AI solutions and more vertical, specialized, application-focused solutions we need. And the business model for these systems moves away from the massive investments of the Frontier providers. The result is more complicated than “open means control.” Model tuning, performance, and optimization could be in your future – as AI moves from a platform to a true layered product set we can use as we need. Lots to learn about here, let us know if you have any questions. Additional Information What’s the difference between closed, open source, and open-weight AI? A researcher explains What Is Open-Weights A.I.? Comparison of Open Source Models

This week we learned of OpenAI’s Jailbreak security breach, which opens the door to many new issues with enterprise AI and HR 2030. This week, following up on the introduction of SHL’s Superworker AI assessment, I dig into the implications of this new AI behavior and discuss the five major “skills” or capabilities we need to build AI mastery among business users. As you’ll hear, AI Agents are both powerful and useful, but our human skills in problem solving, collaboration, and risk awareness are now becoming critical for success. New research also shows that even AI forward deployed engineers need deeper skills in business, stakeholder management, and risk analysis than we ever realized. As you listen to this if you need help with our own AI business capabilities please contact us. Galileo and the new GHRE masterclass certification program are designed to quickly give you and your team the experience and skills you need to flourish in this new world of AI, agents, and Superwork. Additional Information OpenAI says its AI went rogue and launched ‘unprecedented’ cyber-attack The Josh Bersin Institute, HR 2030, And The Global HR Excellence Certification Our New Book: Superpowered®, coming this Fall.