AI Replaces Tasks, Not Jobs: Rethinking Work with Ryan Roslansky and Aneesh Raman

Had one of the best times of my life with Microsoft last week at their Microsoft Research Cambrdige where I got a chance to meet creators from 9 different countries, listen to leaders who are shaping AI for Science, Infrastructure and Responsible AI and also toured across their labs in CambrIdge. I also got a chance to sit with Doug Burger who is the MD of Microsoft Research and has been the one who established Microsoft Research Labs in 10+ countries including India, we talked about: - Microsoft Research Labs and what work they do - Careers in Research changing with AI - Research and Product Development speed - his Hardware AI paper that he wrote in 2000. - research choices one should make and more. Here is an exclusive sneak peak to some behind the scenes and all the fun we did and also my discussion with Doug. Thanks to Microsoft for the opportunity and the invitation. #microsoft #ai #tech #cambridge #research

A lot of companies are trying to figure out why AI still feels underwhelming. • They bought the tools. • They ran the training. • They told people to start using it. But Microsoft’s 2026 Work Trend Index points to something different. They looked at 20,000 workers across 10 countries, plus trillions of Microsoft 365 productivity signals. The big takeaway? The companies getting results are not just better at prompting. They have better structures around the work. Manager behavior. Culture. Talent practices. Clear expectations. Room to experiment. Same tools. Different environment. That is the piece of this puzzle a lot of companies are missing. I broke down what Microsoft calls Frontier Firms, and what it means for Indiana mid-market companies. *Link in the first comment.

#Microsoft just banned its own engineers from using #AI. The tool was costing MORE than the humans it was supposed to replace. For two years, the story has been: AI increases efficiency. AI reduces headcount. AI cuts costs. But inside companies actually using it at scale… A different reality is emerging. Microsoft gave thousands of engineers access to Claude Code and encouraged adoption. #Engineers loved it. Usage exploded. Then the invoices hit. Token-based pricing means every query, code review, and debugging session costs money. At scale across 100,000 engineers, costs surged so fast Microsoft moved to cancel most Claude licenses and shift teams to a cheaper internal tool. Let that sink in: They invested $5B in #Anthropic… Then told their own engineers to stop using it because it was too expensive. Uber saw the same pattern. By March, 84% of their 5,000 engineers were using AI, generating 70% of committed code. Heavy users spent $500–$2,000/month. Their CTO spent $1,200 in a two-hour demo. They even gamified usage with internal leaderboards. Then the budget was gone—by April. Now consider this from Nvidia’s VP of Applied Deep Learning: “For my team, the cost of compute is far beyond the costs of the employees.” That’s the company selling the chips. And even they’re saying AI can cost more than humans. Meanwhile, the market narrative hasn’t changed: Cut workers → stock goes up Announce AI → stock goes up But at scale, the math breaks: The MORE AI you use… The HIGHER the bill. Goldman Sachs projects a 24x increase in token usage by 2030. Gartner says even if token prices drop 90%, total costs will still rise due to agent-driven demand. #Meta tracks usage internally (“Claudeonomics”). Amazon pushes engineers to “tokenmaxx.” And Big Tech is spending ~$725B on AI infrastructure this year. Yet the companies closest to deployment are already pulling back. Because the economics don’t work—yet. That’s the gap no one on Wall Street is pricing in. I find this very interesting. What do you think? Source: @Ric_RTP via X

Microsoft reportedly began discontinuing internal Claude Code subscriptions after AI usage costs exceeded expectations. What initially appeared to be a productivity breakthrough rapidly evolved into a significant budgetary concern. Reports indicate that token-based AI coding tools consumed substantial portions of enterprise AI budgets within a matter of months. (ETTelecom.com) Microsoft is not alone in facing these challenges. Across the technology sector, organizations are reassessing the pace and scale of AI deployment as the economic implications become increasingly difficult to justify. Several companies have announced layoffs, restructuring initiatives, or revisions to their AI strategies while contending with escalating infrastructure and operational expenses. (The Times) This reflects a reality that many previously overlooked: AI is powerful. AI is efficient. AI is valuable. However, AI is not without cost. Nor is AI a substitute for human capability. Behind every successful AI system are human engineers, product strategists, designers, support professionals, domain specialists, and decision-makers. While AI can generate code, it cannot assume ownership or responsibility. It can automate processes, but it cannot replace accountability, empathy, creativity, leadership, or sound judgment grounded in real-world experience. For years, the prevailing narrative has been: “AI will replace employees.” Now, the discussion is evolving: “Can organizations realistically afford to replace employees with AI at scale?” The answer is becoming increasingly evident. The future is not AI versus humans. The future is AI working alongside humans. Organizations that eliminate human involvement entirely may achieve short-term efficiencies, but those that succeed in the long term will be the ones that empower skilled professionals through AI rather than replace them. Technology should enhance human potential, not diminish human value. Ultimately, innovation continues to begin with people. #AI #microsoft #chatgpt #claude #trending #axisbank #copilot #dataengineer #IT

𝐒𝐨𝐦𝐞𝐭𝐡𝐢𝐧𝐠 𝐪𝐮𝐢𝐞𝐭𝐥𝐲 𝐟𝐥𝐢𝐩𝐩𝐞𝐝 𝐢𝐧 2026 𝐭𝐡𝐚𝐭 𝐧𝐨𝐛𝐨𝐝𝐲'𝐬 𝐫𝐞𝐚𝐥𝐥𝐲 𝐭𝐚𝐥𝐤𝐢𝐧𝐠 𝐚𝐛𝐨𝐮𝐭. The whole bet was simple — replace expensive humans with AI. Cut costs. Move faster. Build leaner. But the last few weeks of news made me stop and re-read. 👉 Uber blew through its entire AI budget months ahead of schedule. 👉And here's the kicker — NVIDIA's own leadership admitted that compute is now costing them more than their people. 𝐑𝐞𝐚𝐝 𝐭𝐡𝐚𝐭 𝐚𝐠𝐚𝐢𝐧. The company selling AI infrastructure is saying AI is more expensive than humans. 👉And just this week, Microsoft — which dropped $80 billion on AI data centers — told its engineers to stop using AI coding tools. Why? The bills got too big. 👉And IBM just announced they're tripling their human hiring next year. 𝐓𝐡𝐚𝐭'𝐬 𝐧𝐨𝐭 𝐨𝐧𝐞 𝐬𝐭𝐨𝐫𝐲. 𝐓𝐡𝐚𝐭'𝐬 𝐚 𝐩𝐚𝐭𝐭𝐞𝐫𝐧. And it got me thinking — 𝐀𝐈 𝐝𝐨𝐞𝐬𝐧'𝐭 𝐦𝐚𝐠𝐢𝐜𝐚𝐥𝐥𝐲 𝐫𝐞𝐦𝐨𝐯𝐞 𝐜𝐨𝐬𝐭. 𝐈𝐭 𝐣𝐮𝐬𝐭 𝐬𝐡𝐢𝐟𝐭𝐬 𝐢𝐭. From salaries to compute. From teams to infrastructure. From office floors to GPU clusters. The invoice changed names. The weight of it? Pretty much the same. Sometimes heavier. We didn't solve the cost problem. We just repackaged it. Nobody's saying this out loud. Companies are just quietly course-correcting. But the lesson sitting underneath is hard to miss — 𝐖𝐞 𝐰𝐞𝐫𝐞 𝐚𝐬𝐤𝐢𝐧𝐠 𝐭𝐡𝐞 𝐰𝐫𝐨𝐧𝐠 𝐪𝐮𝐞𝐬𝐭𝐢𝐨𝐧. We kept asking — "how do we replace humans?" When the smarter question was probably — "what does this actually need?" Because the moment you try to automate everything, you realize you can't automate the decision itself. Someone has to say "use this tool here, don't use it there." Someone has to catch what it gets confidently wrong. Someone has to make the call when the numbers don't tell you anything. 𝐓𝐡𝐚𝐭 𝐬𝐨𝐦𝐞𝐨𝐧𝐞 𝐢𝐬 𝐬𝐭𝐢𝐥𝐥 𝐡𝐮𝐦𝐚𝐧. And I wonder if that's actually the whole point. Not that AI is bad. Not that humans are irreplaceable. Just that some decisions require a person, and pretending they don't is expensive. The ones who figure this out first won't be talking about how much AI they use. They'll just be quietly better at knowing what to use it for. 𝐒𝐨 𝐭𝐞𝐥𝐥 𝐦𝐞 — when you read this kind of news, does it land different for you too? #AIStrategy #FutureOfWork #ThoughtLeadership #HumanAndAI #MarketingMindset

The head of AI at Microsoft said something this week that I have not been able to stop thinking about. 18 months. That is how long he thinks it will take for AI to automate most white-collar work. Not 10 years. Not 5 years. 18 months. Now I will be honest — I do not think he means every job disappears in 18 months. What he means is far more specific and far more important: The way white-collar work gets done is going to look completely different by the end of 2027. Reporting, analysis, drafting, research, scheduling, summarising — the tasks that fill up most professionals' days — are already being handled by AI tools that cost less than a cup of coffee per day. And today — as I am writing this — Google I/O is happening live. The biggest AI event of the year. Announcements dropping every hour about new tools, new capabilities, new ways AI is embedding itself into the software every professional uses daily. This is not a slow shift. It is happening in real time. Today. Right now. Here is what keeps coming back to thought. The Microsoft AI chief did not say professionals are being replaced. He said the work is being automated. Those are two very different things. The professionals who will thrive are the ones who let AI handle the repetitive work — and spend their freed-up time on the things AI genuinely cannot do. Judgment. Relationships. Creative thinking. Reading a room. Knowing when the data is wrong, all where human understanding and judgement is required. The 18-month window is not a threat for people paying attention. It is a head start. Are you using that head start — or is 18 months going to arrive as a surprise? Drop your honest answer below 👇 and save this post 🔖 — this conversation is worth having right now. #AITools #FutureOfWork #CareerGrowth #GoogleIO #AILiteracy

$37,000,000,000. This is Microsoft’s annual AI revenue. It increased by 123% in just one year. Consider that for a moment. Last year, people were wondering: “Is AI profitable?” Microsoft has just given them an answer loud and clear. What does it all mean for business executives? AI is no longer an experiment. It is a source of income. It is being planned, budgeted, and presented to investors. Companies that are not developing an AI strategy today are losing ground. In 2026 alone, Microsoft will invest $190 billion in artificial intelligence infrastructure. This is not a gamble. It is a certainty. The financial aspect everyone ignores? Despite this, their stock price fell 5% after announcing quarterly results — as Wall Street expected more. The threshold is officially set to infinity. If the most valuable company in the world is investing heavily in AI and failing to meet expectations — what does that say about its future? Are you thinking of artificial intelligence as a cost-saving solution or as a revenue source in your business? 👇 #AI #Microsoft #Finance #Business #Investing #ArtificialIntelligence #FutureOfBusiness #USA

Microsoft didn't invent Cowork. They shipped it fast by pairing Anthropic's frontier models with Microsoft's distribution. 5 things IT and business leaders should pay attention to: 1. Distribution beats invention. Microsoft owns the seats, the contracts, and the ecosystem to scale Cowork. That alone moves more enterprise AI than any model release. 2. Model quality changed the story. Cowork is viable now because the underlying intelligence got good enough to trust with multi-step work. 3. Better tech doesn't fix bad habits. A stronger agent won't turn weak user habits into strong ones by itself. 4. Governance decides the ROI. Readiness, permissions, rollout, training. That's where companies either get value or quietly lose it. 5. Buying licenses is not transformation. The real work is change management, not procurement. The companies that win the next 24 months won't be the ones with the most AI licenses. They'll be the ones who get their teams to actually use the tools. Full breakdown: https://ow.ly/4r4O50YYjMk #MicrosoftCopilot #Cowork #Anthropic #AIAdoption

💡 Opportunity should not depend on where someone starts. It should be shaped by the skills they can build. That is the focus of the Future Skills Program, a global initiative from The Adecco Group and Microsoft, powered by the Spark AI team. The program gives Candidates and Associates access to free foundational AI training through curated learning paths from Microsoft Learn. It’s designed to help underrecognized talent gain practical skills for an AI-driven economy and move closer to meaningful employment. By connecting learning to employability, we can help create a workforce that’s more capable and ready for what comes next. Find out more 👇 #Akkodis #Microsoft #SparkAI

Microsoft AI CEO Mustafa Suleyman believes most white-collar tasks could be automated within the next 12–18 months. Speaking to the FT, he said AI is nearing “human-level performance” across professional work — especially roles done on computers, from legal and accounting to marketing and project management. He joins a growing group of leaders making similar calls. Anthropic CEO Dario Amodei has warned that up to half of entry-level white-collar jobs could disappear, while Ford CEO Jim Farley and Elon Musk have also highlighted significant disruption ahead. The driver: rapid gains in compute power and model capability. Suleyman also pointed to Microsoft’s ambition to build frontier AI systems and ultimately achieve “superintelligence.” Yet current adoption tells a slower story. AI is being used in targeted ways, with mixed productivity results and limited broad job displacement so far. Market expectations, however, are moving faster. Source: Fortune #AI #FutureOfWork #Automation #ArtificialIntelligence #Technology #Workforce #DigitalTransformation