AI in Q3 2026: the evidence needed to expand
AI developments from July to 28 September 2026, with a decision memo for expanding, holding or stopping a business deployment.
Insights
Evidence and practical judgement for the decisions institutions face when operating with AI.
Start with the operating decision
A useful pilot still needs ownership, support and evidence for the work people will run. This guide connects the demonstration to the next funded commitment.
From experimentation to institution: what deployment requires21 items
Search the libraryRead
The AI business questions worth investigating in 2027
A practical research agenda covering realised value, workflow reliability, regional rollout and the evidence business leaders need to invest.
Ten decisions boards need to make about AI in 2027
Ten board questions linking AI strategy, delegated authority, economics and resilience to evidence, accountable owners and concrete decisions for 2027.
The 2027 AI delivery agenda: deciding what to fund and what the team can run
A portfolio planning framework for balancing AI operations, new investment, shared delivery capacity and decisions to expand or stop.
When AI agents control money
A practical approach to delegated spending: identify the principal, bound the mandate, enforce limits and preserve evidence when an agent initiates a payment.
Stablecoins as settlement infrastructure for the agent economy
How to compare stablecoins with cards, bank payments and tokenised deposits for agent transactions, including redemption, liquidity, custody and regulatory status.
Does AI-assisted content improve commerce economics? How to test it
An experiment design for measuring AI-assisted content through contribution, production cost, review effort and incremental demand.
From supplier data to reporting evidence
How to use AI in sustainability data preparation while preserving source records, methodology, accountable review and reporting boundaries.
Sovereign AI does not necessarily mean sovereign models
A practical way to decide which AI capabilities a country or institution must control, and where partnerships can meet the requirement.
How to test AI-assisted reconciliation before relying on it
A reproducible evaluation protocol comparing manual work, rules and AI assistance, with clear error measures and an illustrative workload calculation.
Human oversight in an agentic organisation
How to make human oversight work through enforced authority, action-bound approvals, useful evidence and rehearsed recovery.
AI in Q2 2026: what the purchase commits you to
What April–June 2026 AI developments meant for enterprise buyers, with a worksheet for comparing configuration, specialist software and custom delivery.
The WCAI Briefing
A monthly perspective on the AI developments that change institutional decisions.
Explore the Briefing