You're investing tens of millions in a programme designed to optimise tasks you already do. Unit economics improve. Competitors match. Improvement passes through to customers as price. You optimised your way to competitive convergence.

The faster you automate, the more valuable the unautomated becomes. Most enterprises are investing in the wrong side of that trade.

This is the concept of talent dehumanisation in my book, and is completely the wrong path to take. It's a tool for humans, not a replacement. Effort gets shifted, not removed entirely

Steam boilers exploded for fifty years. Every one was called a freak accident, until an insurer priced inspection in 1866. AI agents are in year two of the same curve, and every benchmark still scores completion, not damage. Who's the insurer? thehumanityparadox.substack.…

Last week, eight of our engineers spent two days in Salesforce’s Sydney and Melbourne offices. The most useful thing they came back with wasn’t a technique. It was a better question. Most Agentforce conversations still start with “How do we get an agent to do the job a person does today?” During Forward Deployed Engineering training, Ashish Jain reframed that to: “What can an agent inherently do better than a human?” — filtered on faster, cheaper, or more accurate. If a use case clears none of those bars, it doesn’t need an agent. That single filter kills more bad pilots than any amount of governance. It matters because 95% of GenAI pilots are failing, and most of those are adoption failures rather than technical ones. The gap between deployed and adopted is where enterprise value leaks out. FDE is @salesforce’s answer to that gap: engineers building alongside the client against a measurable outcome, not a POC. This is completely aligned with @Cognizant and @imravikumars vision of our new Frontier Engineers that support our AI Builder strategy. The piece I’d point any CXO to is the Agent Care Plan. You define success before the agent ships, with named guardrail KPIs and a zero-violation target, then monitor it after go-live. It turns “we launched an agent” into “we can prove what it’s worth.” We’ve built accelerators to generate these for any Agentforce use case, which is what lets us deliver against outcomes rather than effort. Thanks to Justin Plant, Aman Jain, Montu Pradhan and Gajendra Sisodia for running it. And thank you to my amazing team for taking part! If your agent programme can’t state its success metric before build starts, you’re just experimenting on your customers. @trailhead @partnerforce @SFResearch @agentforce

This will be a game changer and signal for what’s to come

Honestly blown away by the performance of these lads. Incredible under the pressure. Looking forward to a 7am Sunday watch from down under

Just in: Oil found in Belgium, US investigating 😂