YPO Technology Network AI Brief

On Tuesday, Kumar Mangalam Birla, chairman of the Aditya Birla Group, said his group will go more gradually on AI than it planned a year ago, and will not spend out of fear of missing out. Stephen agrees that slow is an option, but argues the real word is deliberate, and that the most overlooked benefit of AI has little to do with AI. In this episode, Stephen Forte covers: What Birla said. A shift from an AI arms race to a hard look at returns, "no one cap fits all," and the best description of executive AI anxiety of the year: a monster sitting in the last row. What he did not say. He did not call AI overhyped or say his group is stopping. On the same stage, Infosys said AI services are already about 8 percent of its revenue. Slow is fine. Deliberate is better. Nail down the approach and what success looks like before you spend. Throwing AI at the same old process is a jet engine on a shopping cart. The Bill of Work callback. AI bolted onto a process nobody has questioned since the fax machine gives you a faster fax machine. Show the P&L along the way. Go at a pace that makes sense and show real profit and loss impact every quarter, instead of boiling the ocean. The overlooked benefit. Explaining a process to an AI forces you to re-examine it. Many fixes are not AI at all: automated routines, or a vendor upgrade that slipped in years ago and nobody noticed. This quarter. Pick one process everyone complains about, write down every step, cross out what nobody can justify, fix what you can with what you own, and only then bring in AI with a number and a date. Sources: The Economic Times (PTI), 13 October 2026: https://economictimes.indiatimes.com/ai/ai-insights/kumar-mangalam-birla-advocates-for-roi-focused-approach-to-ai-investments-amidst-growing-technological-landscape/articleshow/134919103.cms The Indian Express, 13 October 2026: https://indianexpress.com/article/business/infosys-tech-ceos-ai-demand-growth-energy-bottleneck-bloomberg-forum-10920080/ The AI Brief from the YPO Technology Network is a daily executive briefing on the AI developments that matter to business leaders. Hosted by Stephen Forte.

On Monday, HCLTech, India's third-largest IT services firm, announced a business unit dedicated entirely to Anthropic, adding to its Gemini business unit and its OpenAI partnership. Stephen thinks organizing around the AI lab is the wrong way to slice it: the model will matter less and less, and the plumbing around it will matter more. In this episode, Stephen Forte covers: What HCLTech announced. A Global Premier Partnership with Anthropic, a dedicated Anthropic Business Unit, forward-deployed engineers, and Claude inside its AI Force platform. Accenture formed an Anthropic business group last December. Fair to HCLTech. Customers ask for lab-specific help, the labs fund it, the release talks about guardrails and human oversight, and HCLTech's advanced AI revenue rose 78 percent on a year ago. The model is not the moat. A different lab leads every few months. Building your company around one lab's model is building your house around this year's refrigerator. Agnostic plumbing in practice. Keep the credentials your agents need in an external secrets store such as Infisical or 1Password, then bring your Python files and local programs. Stephen's company moved from Claude to Grok Bot this way in about twenty minutes. The Microsoft exception. For Microsoft shops, deep platform expertise makes sense, because AI that runs inside your own tenant matters most in regulated industries such as health care under HIPAA and financial services. Price, briefly. HCLTech's CEO says IT services are in the early phases of deflation. Agnostic plumbing lets you shop for it every quarter. The test. Ask how long it would take to switch your AI to a different lab next Monday. A day means plumbing. A project plan means a landlord. Sources: HCLTech and Anthropic, PR Newswire, 12 October 2026: https://www.prnewswire.com/news-releases/hcltech-and-anthropic-sign-global-premier-partnership-to-scale-governed-agentic-ai-across-global-enterprises-302905626.html CNA (Reuters), 12 October 2026: https://www.channelnewsasia.com/business/indias-hcltech-signs-pact-anthropic-set-up-business-unit-6452836 HCLTech Q2 FY27 results, PR Newswire, 12 October 2026: https://www.prnewswire.com/news-releases/hcltech-delivers-strong-q2-led-by-record-deal-bookings-of-3-3-billion-302906028.html DQ India, 12 October 2026: https://www.dqindia.com/news/hcltechs-ai-revenue-jumps-78-but-ai-led-deflation-is-just-starting-12670107 HCLTech Gemini Enterprise Business Unit, 22 April 2026: https://www.hcltech.com/press-releases/hcltech-launches-gemini-enterprise-business-unit-accelerate-agentic-ai-adoption HCLTech named OpenAI Advanced Partner, 6 August 2026: https://www.hcltech.com/press-releases/hcltech-named-openai-advanced-partner The AI Brief from the YPO Technology Network is a daily executive briefing on the AI developments that matter to business leaders. Hosted by Stephen Forte.

About two weeks ago, OpenAI fired three of its own safety researchers. They published an open letter, OpenAI published a long reply, and the two sides now agree on almost everything except what actually happened. Stephen argues that the gap is where every company's real AI policy sits: what happens to the person who raises an AI concern matters more than the policy on the intranet. In this episode, Stephen Forte covers: OpenAI's account. OpenAI says an investigation found the researchers violated clear policies on handling sensitive information, describes a significant breach of trust beyond what they published, and keeps the details private. The researchers' account. Tomek Korbak, Jasmine Wang and Mikita Balesni deny leaking anything and say their work required talking to outside safety groups under rules that were still being written. Korbak says talking to the evaluation group METR was his job. Wang describes an email access she had asked IT to remove. What nobody outside can know. Two stories, no public details, and why Stephen won't pick a side on the facts. Fair to OpenAI. Its statement commits to outside safety assessors, says it is forgiving of good-faith mistakes, and agrees with the researchers' recommendations. It kept the recommendations, not the people. Where Stephen sits. His company uses all the major models and shifts between them. Lately Grok Bot has been faster and significantly cheaper, and in the company's own testing its output beat Claude's, so they are moving fully onto it and will report back. Stephen's read. When the person who raised the flag is gone, everyone else stops raising flags. Intent does not travel; only the story does. The document is the menu; what happened to the last person who complained is the food. One page this month. How someone raises an AI concern, who hears it, what protects them, a path around their own manager, and public thanks when it is used. Then a test: count the bad news. Sources: AP News, 9 October 2026: https://apnews.com/article/openai-chatgpt-ai-artificial-intelligence-safety-789d4f5293fba45a22fcb62ebfbc2a41 NPR, 9 October 2026: https://www.npr.org/2026/10/09/nx-s1-5996890/openai-safety-firings CNBC, 9 October 2026: https://www.cnbc.com/2026/10/09/openai-fired-researchers-ai-concerns.html TechCrunch, 8 October 2026: https://techcrunch.com/2026/10/08/fired-openai-safety-researchers-dispute-misconduct-claims-warn-of-chilling-effect/ The researchers' open letter: https://mikitabalesni.com/letter/letter.pdf The AI Brief from the YPO Technology Network is a daily executive briefing on the AI developments that matter to business leaders. Hosted by Stephen Forte.

Stephen Forte puts a name on the record: the Bill of Work™. Every manufacturer knows its bill of materials: every part, what it costs, the order it goes in. Almost no company has the same document for its work. "Your payroll is your people cost," Stephen says, "but your Bill of Work is really what it costs to get the work done." Drawn from his recent talk at the CEO Global Leadership Forum in New York, this episode is about the difference between an incremental improvement from AI and reinventing your P&L. In this episode, Stephen Forte covers: Why the big AI bets are waiting on business. Most companies haven't reassessed their standard general and administrative processes in years. Everyone knows their seat mile or bed night. Almost nobody knows what one request costs. The line shaft. Why a 1903 factory put its machines where the belts were, and why the org chart grew the same way around where capability lived. The motor on the old line shaft. Bolting a new motor onto the old arrangement is the sensible first move, and it is where most companies are with AI today. In 1899, William Tapley at the Government Printing Office gave his presses their own motors, then found room for 40 more presses. The Bill of Work. Payroll tells you what your people cost. The Bill of Work tells the task. It is also an audit of complexity in an organization. Taking the work apart. Follow one request task by task, the way a chef knows every step of a dish, and sort the tasks by what they are, not which department does them. Why the cost of one task now matters. Vendors are starting to sell work by the outcome. If you don't know what a task costs you, you can't judge their price. The question for your next leadership meeting. Pick one thing your company does every day and ask the room what one of those costs. Earlier on the show: Stephen first used the name on air in S1E158, "Your Company Has No Bill Of Work," September 26, 2026: https://rss.com/podcasts/ypo-technology-network-ai-brief/3180260/ The book: Bill of Work, Stephen's book on taking the work apart, is coming. Follow the show to hear first when it is available. Feedback? Write to Stephen at [email protected]. Sources: Stephen Forte, recent talk at the CEO Global Leadership Forum, New York William Tapley, Journal of the Franklin Institute, 1899, on electric drive at the Government Printing Office S1E158, "Your Company Has No Bill Of Work," September 26, 2026: https://rss.com/podcasts/ypo-technology-network-ai-brief/3180260/ The AI Brief from the YPO Technology Network is a daily executive briefing on the AI developments that matter to business leaders. Hosted by Stephen Forte.

On Friday, Sierra and Meta published Draft 0.1 of the Personal Agent Protocol, nicknamed "Poppy": the first draft of the rules for what your customers' AI assistants may do inside your business. More than forty companies have signed up to help design it, including Bank of America, Visa, United Airlines, Target and OpenAI. Stephen thinks this draft matters more to most companies than any AI model launched this year, and that every company should write its guest list now. In this episode, Stephen Forte covers: What shipped. Sierra announced it last Tuesday with Meta, Shopify, Stripe and Walmart, then on Friday published the draft and 35 more design partners. These are design partners, not adopters. Project lead Bret Taylor co-founded Sierra and chairs OpenAI's board. How it works, in four ideas. A house-rules file on your website. Agents arrive as guests and must say who they are. Customers choose whether an assistant may only look or may also change things, and either side can revoke access. For anything that costs money, the customer approves the exact terms, and pressing the button twice buys it once. The sentence for your board. The draft says the customer's approval "works on trust." It doesn't prove the customer saw the terms, and it doesn't settle who pays when an agent gets it wrong. What's left out. Payments, notifications and attachments are open topics. Any part of the draft can change. Stephen's calls. Customer service: if your agent can't do the job, the customer's agent will sit on hold at your call center for free. Pricing: a bot never feels guilty leaving the cancellation flow. Brand: your website is about to have the same audience as the terms and conditions. Who's not on the list. Google, Apple, Amazon, Microsoft, Salesforce and Anthropic, as of Friday. Disclosure: Stephen's company runs largely on Claude. Fair to Sierra. Sierra sells the customer-service agents that one of the protocol's doors favors. But before Friday critics said there was no draft and no license, and Friday delivered both under Apache 2.0. A neutral governing body is still missing. The guest list. One page, four columns: what a guest agent may see, what a signed-in agent may see, what it may change that you can undo, and what moves money. Allow identified guests, limit changes to named permissions, require approval on your own page for money, and never give an agent full access. No code yet; policy first. Sources: Sierra, "Sharing a draft of Personal Agent Protocol," 9 October 2026: https://sierra.ai/blog/poppy Sierra, "Introducing Personal Agent Protocol," 6 October 2026: https://sierra.ai/blog/introducing-personal-agent-protocol Personal Agent Protocol, Draft 0.1 specification: https://personalagentprotocol.org/docs/spec CNBC, 6 October 2026: https://www.cnbc.com/2026/10/06/meta-joins-companies-to-tame-chaos-of-doing-business-with-ai-bots.html The AI Brief from the YPO Technology Network is a daily executive briefing on the AI developments that matter to business leaders. Hosted by Stephen Forte.