🧠 Food for thought

Implications, context, and why it matters.

Block’s AI plan to cut managers and rethink roadmaps

• Jack Dorsey wants two AI “world models,” one pulls together internal Block data such as code, workflows, decisions, and performance metrics, and another tracks customer and merchant behavior using transaction data tied to Cash App and Square 1.
• Block says these models power an “intelligence layer” that can put together financial products from building blocks like payments, lending, card issuance, and payroll based on demand 1. The example describes spotting a merchant with a seasonal cash flow gap, then assembling a fix from existing capabilities rather than defaulting to a loan 1.
• The plan swaps a traditional roadmap for a “system-generated backlog,” where missing capabilities drive what gets built next 1. A separate report says Block targets $12.20 billion in gross profit in 2026 2.

Fewer managers can muddy accountability

• In regulated areas such as banking and money transfers, AI model “opacity” can leave teams unable to fully explain how the system reached its conclusions 3.
• IMD (a Swiss business school) quotes a decades-old IBM training manual principle that reads “A computer can never be held accountable, therefore a computer must never make a management decision” 4.
• Humans can turn into rubber stamps, as in a ProPublica-reported Cigna case where one doctor denied over 60,000 claims in a month and physicians averaged 1.2 seconds per case 4. CFA Institute Research and Policy Center (the research arm of the CFA Institute, a global investment professional body) cites a survey where firms with structured human-in-the-loop processes cut model-related incidents by nearly 40% compared with fully automated setups 5.