TLDR IT 2026-09-10

Meet Salesforce in Claude, your AI CRO: it understands your entire revenue cycle and acts on it, from prospect to closed deal. When you stream theMain Keynote at Dreamforce next week, you'll see how Salesforce and Anthropic are bringing Claude's reasoning together with the trusted data, workflows, permissions, and governance businesses already rely on.

Then choose from 400+ expert-led sessions to see the biggest launches and hear the boldest ideas. Register for free to stream it live on Salesforce+, September 15–17.

HP is working with Red Hat and Nvidia on an enterprise AI platform that can run production inference across devices, data centers, and hybrid environments. The system combines HP hardware powered by Nvidia Blackwell with Red Hat AI Factory for deployment and orchestration.

Claude for Microsoft 365 integrates Claude into Excel, PowerPoint, Word, and Outlook with shared context across apps. Claude for Excel, PowerPoint, and Word is now generally available on all paid Claude plans, while Claude for Outlook is in beta. The offering emphasizes reviewable edits, use of customer templates, and deployment through existing cloud and compliance frameworks.

Apple unveiled the $1,999 iPhone Duo, its first foldable phone, alongside the iPhone 18 Pro lineup powered by its new A20 Pro chip. The launch also pushes Apple further into on-device AI, with expanded Siri capabilities and a continued focus on privacy.

Enterprise AI evals fail when vendors can see POC test questions. Teams need a permanently held-out eval that vendors never see. Google DeepMind's double-blind frontier-model evaluation shows the model: neither side saw the other's evaluation assets. Cleaner lab benchmarks alone won't fix enterprise adoption failures.

AI will end the era of identical software for every customer by making per-customer variants cheap to create and maintain. Vendors, not open-source templates, will own these divergences, reviving SAP-style customization without its historic migration cost. The result will be the end of the common UI, where each customer uses a recognizably different version of the same product.

Google was named a Leader in Gartner's inaugural Magic Quadrant for Enterprise AI Assistants. Google says the category is shifting from basic copilots toward agents that connect enterprise systems and automate multi-step workflows.

AI infrastructure needs stronger reproducibility and control across hardware, drivers, models, and data than Kubernetes alone provides. Governance and bandwidth constraints are expected to push more AI workloads toward hybrid and distributed environments.

They're building their own tools to run their operations. OverAI enables teams of any size and skill set to build and share AI-powered tools, with data and user management that stay with IT. Start free with $200 in credit.

Start free → | See how it works →

Google Cloud and Accenture have launched the Accenture Gemini Enterprise Business Group, a joint unit to embed engineers in enterprises to accelerate adoption of Google's AI tools. Google will train up to 1,000 Accenture forward-deployed engineers to build custom applications on the Gemini Enterprise platform. The move is aimed at closing an enterprise deployment gap as Google holds about 6% of US enterprise AI spending, far behind Anthropic and OpenAI.

Red Hat and Rafay published a reference architecture that lets cloud providers turn distributed GPU infrastructure into multi-tenant sovereign AI services. The design covers self-service provisioning, governance, metering, and isolated GPU environments.

Writer launched Enterprise Brain, a context layer that gives AI agents access to shared organizational knowledge, business logic, and live data. It also introduced team-level Agent Memory and integrations for Slack, Teams, Zoom, and Google Meet.

TeleGeography's seventh WAN Manager Survey finds enterprise networks shifting from MPLS to Dedicated Internet Access as the primary underlay, with SD-WAN mature and SASE/Zero Trust expanding. AI has minimal current impact on bandwidth, with cloud and SaaS remaining the main drivers, while satellite use stays niche and Network as a Service adoption remains flat.

Lightbits unveiled Inferra, a KV-cache orchestration engine designed to improve GPU utilization for long-context and multi-session inference workloads. The company says the storage layer can help AI infrastructure providers serve larger workloads without adding GPUs at the same rate.

Alteryx added conversational analytics, agent-building tools, an MCP server, and ChatGPT integration so companies can expose governed datasets and workflows directly to AI agents.

Enterprise AI adoption requires tamper-evident audit trails that automatically capture prompts, outputs, model versions, guardrails, and human approvals to prove what AI did and who authorized it.