|
The so-what: this week wasn’t about new models. It was about AI finally showing up on the balance sheet — in planning hours, forecast errors, and one record-breaking IPO. AI AgentsFeatured: McKinsey Puts a Number on Agentic AI — 83% Fewer Planning Hours, ~50% Fewer Forecast ErrorsWhat happened: A new McKinsey report details a life-sciences company whose monthly forecasting once demanded nearly 100 meetings and 8–10 rounds of manual revisions. After redesigning the process around AI agents — which collect data, investigate anomalies, test assumptions, explain forecast changes, and execute approved decisions — forecast errors fell roughly 50% and people hours dropped 83%. Why it matters: This is the first headline-grabbing, third-party number showing agentic AI moving a real operational metric, not a benchmark. The kicker is the inventory math: each one-point improvement in forecast accuracy correlates with roughly 2.7 points less finished-goods inventory. In one chemicals-sector phase, forecast accuracy improved 15–20 points while on-hand inventory dropped 20–30%. That’s cash freed up, not a demo. What’s next: Note the caveat buried in the report — “fully agentic” sales-and-operations planning is still early, and most deployments still require a human to approve recommendations. The line worth remembering: “The scarce capability is no longer producing a number. It is supervising an AI-enabled decision system and knowing when that system is wrong.” That’s the skill your team is about to need. Source: ET EnterpriseAI / McKinsey Quick HitsMicrosoft brings model-routing to the enterprise. At Ignite, Microsoft unveiled Copilot modes plus Microsoft-Decision-1.1, which routes each part of an agent request to the cheapest model that can handle it — across cloud, on-premises, and edge. The pitch is quieter agents that burn fewer tokens, a direct answer to the “agents are too expensive to scale” objection. Seeking Alpha BlackLine adds governed agents to the CFO’s office. BlackLine — serving 4,300+ organizations — is embedding OpenAI frontier models into its Agentic Financial Operations Platform, with an explicit emphasis on human-led governance and auditability. The message: in finance, the agent is only as good as the audit trail behind it. FF News AI NewsAnthropic’s record IPO cements the AI trade. The Claude maker debuted on Nasdaq under ticker ANT at a record $81.3 billion raise — topping SpaceX — after upsizing to 980 million shares and pricing at $83, above its range. It’s the first frontier AI lab to go public, and the deal alone helped snap a three-day European stock slide. The signal: institutional capital now treats AI labs as core holdings, not speculative bets. LiveMint OpenAI is building a phone — with Foxconn. Supply-chain reports say Foxconn subsidiary FIH Mobile has landed the assembly order for OpenAI’s smartphone project, with the Jony Ive-designed device reportedly targeting a 2026–2027 launch. Specs and volume are unconfirmed, but the move would put OpenAI directly against Apple and Google over AI-first user experiences — a bet that the interface of the future is a device, not an app. Benzinga South Korea tops $1 trillion in exports on the AI chip boom. Annual exports crossed $1 trillion for the first time, powered by surging memory-chip demand from AI data centers — with semiconductor shipments up roughly 70% year over year. But more than 40% of exports now sit in a single category, and analysts warn the concentration is a vulnerability even as the memory supercycle looks set to run through next year. Yonhap
|
