NextAI+ Praxis--:----UTCA close, weekly reading of how large enterprises actually wire generative AI into the business — not press releases, but architecture. We track real budgets, real workloads, and the slow shape of demand for infrastructure, governance, and the operators who have to live with the rollouts.
A useful tool is not automatically production-ready. Scenario risk — not tool effectiveness — should decide whether AI runs on public cloud, enterprise SaaS, private or hybrid infrastructure, or sovereign AI, as UiPath moves agentic automation on-premises and SoftBank and Oracle turn sovereign AI into a packaged model-plus-cloud offering.
Alibaba.com’s Accio Work builds a seller operating agent and Made-in-China.com’s SourcingAI a buyer procurement agent. Read as deployment chains — data integration, context construction, orchestration, permission control, observability — they show platform budgets shifting from connection efficiency to the capabilities buyers and sellers cannot complete on their own.
Microsoft and Salesforce break AI pricing down to credits, actions and conversations; Databricks and OpenAI ship spend controls after customers burned tens of millions in a month. As agentic workflows lengthen, the measure shifts from usage and token volume to workflow return — and to who decides which task nodes AI belongs in.
Internal agents can already handle tasks, but the moment work crosses the enterprise boundary the external interface is still built for humans. Cloudflare, Visa and Mastercard sketch the missing layer — discoverability, identity, authorization, execution and status feedback — that cross-enterprise AI workflows now depend on.
A new delivery form sits between big-tech platforms and bespoke projects: agent workspaces that connect existing SaaS, turn knowledge into context, and let business teams build their own agents. Assembled, Clay and Vanta report 90%+ adoption — but the form suits digitally-ready mid-sized firms, not end-to-end custom delivery.
Two hiring signals — Unilever’s process-focused artwork role and Marriott’s employee-enablement role — reveal the next question after deployment: whether an organisation can absorb AI. Process-intensive firms shore up workflows; service-intensive firms build people capability, and budgets follow both.
Before AI, Chow Tai Fook rebuilt its foundation — a data layer on SAP and a process layer on Alibaba Cloud — then extended AI into customer customisation, employee workflows with Microsoft, and retail planning with o9. How a traditional jeweller sequences the groundwork that full AI deployment depends on.
Over two years The Home Depot moved AI along the customer journey in three stages — Magic Apron for product knowledge, project decision support, then a voice agent that cut store phone-support time to a quarter. A staged playbook where each step maps to a measurable business result.
Chow Tai Fook runs 400+ AI agents while only 31% of surveyed CFOs report satisfaction with results. Once AI is everywhere in production, the next move is to inventory it: Lanai’s AI @ Work Operating System reframes scattered workflows as operating assets to measure, prioritise, and trade off.
Oracle, Trintech, and MYOB embed AI agents directly into finance, supply chain, and customer experience software, as AI penetrates enterprises at the suite, process, and product level. Multiple concurrent entry points raise new identity and coordination challenges.
Volkswagen, Gradient Labs, Transcend, Oracle and ISG, read as one week: the first wins are standardised workflows, and the next bottleneck is not models or compute — it is governance capacity, cross-function coordination, and multi-model orchestration.
The rules, in plain language. EU, UK, Singapore, US state-level — what changed, who has to act, by when. Filed every Wednesday.
The reports worth an afternoon, taken seriously at length. One document per entry — we read the appendices so the argument holds.
If you want both columns delivered together, four times a year, in one quiet email — leave an address. Otherwise just bookmark this page.