Get tomorrow's brief in your inbox
June 9, 2026
Today: Apple taught iPhones to build workflows from plain-language descriptions, cutting automation from minutes to seconds. VCs are buying legacy companies outright to retrofit them with AI, turning service businesses into software margins. n8n shipped OpenTelemetry tracing so workflows show up in the same observability stack as the rest of production.
Apple's WWDC 2026 just changed how non-technical people automate their work. The redesigned Shortcuts app in iOS 27 lets you describe what you want in plain language, and Apple Intelligence builds the automation for you. Instead of manually stitching together app actions and variables, you type something like "notify my partner when I leave work with my ETA," and Shortcuts assembles the workflow automatically, pulling from geofencing, Maps routing, and Messages.
This matters because automation has always been powerful but rarely accessible. Until now, Shortcuts required you to understand how apps connect, what actions exist, and how variables flow between steps. If you got stuck, you either gave up or spent an hour in forums. Apple just removed that barrier entirely. You describe the goal, the system figures out the steps, and you can edit afterward by describing changes instead of rebuilding from scratch.
For small business owners, this turns your phone into a custom automation platform without hiring a developer. Customer check-ins trigger CRM updates. Invoice approvals route to accounting. Appointment confirmations pull calendar details and send formatted messages. The workflows you've been paying SaaS tools to handle can now run locally on your device, controlled entirely by you, with zero monthly fees and complete privacy.
Apple will let you build workflows using AI in its new Shortcuts app
Apple just taught your iPhone to finish your sentences, your photos, and your workflows
NotebookLM gets Gemini 3.5 and a cloud computer. Google upgraded NotebookLM to Gemini 3.5, adding the ability to start research by asking questions instead of importing sources. The app now uses Google Search to find relevant materials, and each notebook runs on a secure cloud computer powered by Google's Antigravity coding platform. This means NotebookLM can write and execute code during your research, then output results as PDFs, data visualizations, Excel files, or CSV sheets. Available now for Google AI Ultra and Workspace customers, with broader rollout coming.
NotebookLM's Gemini 3.5 upgrade adds a cloud computer and help finding sources
n8n ships OpenTelemetry tracing for production workflows. Every n8n workflow execution now emits as a trace via OpenTelemetry, so your automation platform shows up in the same observability stack as your other services. This matters for teams running workflows in production that process payments, route tickets, sync CRM data, or drive AI agents. When something breaks, you see the full path across systems instead of context-switching between tools. The feature supports trace context propagation, meaning inbound requests to n8n continue the upstream trace, and outbound HTTP calls propagate context downstream. If you run n8n in queue mode across multiple workers, traces correctly parent back to the main process. Point n8n at any OTLP-compatible backend like Datadog, Honeycomb, Grafana Cloud, or Tempo, and start seeing what your workflows do at runtime.
Trace n8n workflow and node executions with OpenTelemetry
Cursor vs. Copilot: Which AI coding assistant wins in 2026. Cursor maintains an edge in agent capabilities and autocomplete accuracy, while GitHub Copilot offers broader editor compatibility. Cursor is a standalone IDE (a VS Code fork) and uses a proprietary fast model for tab autocomplete that pulls from your indexed codebase, making suggestions more contextually relevant in as little as 200ms. Copilot installs as an extension in VS Code, Visual Studio, JetBrains, and others, but doesn't use codebase context for autocomplete, only in its agent chat. Both tools index your entire codebase for RAG-based accuracy, and both propagate that index across team accounts. If you're willing to switch editors, Cursor offers faster, smarter autocomplete. If you want to stay in your current setup, Copilot delivers solid AI assistance without disruption.
Cursor vs. Copilot: Which AI coding tool is right for you?
Calendly now works with AI chatbots via MCP. Calendly shipped Model Context Protocol (MCP) support, meaning you can use Claude, ChatGPT, or Cursor to find available time slots, book meetings, and update availability without opening the app. This is most useful if you already live in an AI chat interface and want scheduling to happen inline without tab-switching. If you're not using AI agents daily, the standard Calendly workflow is still faster.
Calendly vs. Acuity: Which scheduling app should you use?
How to use your CRM for smarter email campaigns. CRMs improve email marketing in three ways: better segmentation using real contact data and behavior history, stronger personalization via dynamic tokens and lifecycle-based messaging, and more effective automation with trigger logic tied to sales outcomes. The CRM acts as the system of record that determines who should receive a message and why, while email software handles delivery and optimization. HubSpot's guide walks through setting up centralized contact data, building dynamic audiences, automating workflows based on CRM triggers, and measuring email performance against pipeline activity instead of just open rates. If you're still using static email lists or disconnected tools, this is the framework for upgrading to a data-driven approach.
How to use your CRM for smarter email marketing campaigns
Keyword research for answer engines in 2026. Traditional SEO keyword research focused on search volume and keyword difficulty. Answer engine optimization (AEO) shifts to intent, relevance, and answerability because users now type paragraph-length queries into AI search tools, and results are synthesized inside the interface instead of blue links. You can't track search volume for AEO queries the way you could in Google, so the strategy becomes creating content that AI crawlers can easily parse and synthesize. HubSpot's guide covers how to map entities, write for conversational phrasing, structure content for direct answers, and measure visibility via citations in AI-generated responses instead of clicks. If your traffic is stagnating and you haven't adapted your keyword strategy for AI search, this is the playbook.
Keyword research for AEO: A guide for winning answer engine traffic in 2026
Building reliable multi-agent AI systems. Most teams hit a complexity cliff when they move beyond single-agent workflows. The fix isn't avoiding complexity, it's building with clear boundaries, explicit interfaces, and isolated failure domains. n8n's production AI playbook covers how to decompose tasks into discrete steps, decide which need an LLM versus deterministic logic, and use the AI Agent Tool to create orchestrator patterns where one agent delegates to others. The key insight: treat this as an architecture challenge, not a prompting challenge. Before you connect nodes, map out every sub-task, identify what needs reasoning versus what's a database lookup, and build a blueprint that keeps the system manageable as it grows.
Production AI Playbook: Complex Agent Patterns
How to monitor what your AI agents are actually doing. Uptime dashboards tell you if an agent is running, but not if it's confidently giving wrong answers or retaining session data it shouldn't. Behavioral monitoring fills the gap by logging agent responses, decision reasoning, tool usage, guardrail triggers, and memory state. Store this in a queryable destination so you can audit what the agent said, why it said it, and whether the behavior matches expectations. The EU AI Act (Article 12) already requires high-risk AI systems to log activity throughout their lifetime. Even if your agent doesn't fall into that category, structured logs turn "why did the agent do that?" questions into debuggable trails instead of guesswork.
How to monitor usage and performance of AI steps
What LLM observability actually measures. LLM observability captures what the model saw, why it decided, and what it returned, turning silent failures into debuggable decision trails. Unlike traditional monitoring that tracks symptoms like error rates, observability provides context to explain why a seemingly healthy model is failing. Instrument traces and spans to isolate latency and logic errors, track metrics across system performance, resource cost, output quality, and integration health. The challenge is that LLMs are non-deterministic, so a prompt that worked yesterday might hallucinate today. You can't rely on unit tests; you need live data streams and structured logs that turn subjective qualities like groundedness and relevancy into measurable indicators.
LLM Observability: What To Instrument and How To Act on It
Apple waives AI infrastructure costs for small developers. Developers with fewer than 2 million first-time App Store downloads can now use Apple's Foundation Models running in Private Cloud Compute with no cloud API cost. This removes the infrastructure barrier for indie developers experimenting with AI features, similar to how Apple's Small Business Program offers lower commission rates to smaller developers. The Foundation Models framework is expanding this year to support image input and server models, with an API that integrates with any cloud model provider. If you're building an iOS app and want to test AI features without burning cash on API calls, this is your opening.
Apple bets cheaper AI will woo small developers
Tray.ai pricing: What the enterprise iPaaS actually costs. Tray doesn't publish pricing, but here's the breakdown. Pro gives you basic workflow triggers, 700+ connectors, and three workspaces (good for solopreneurs and small teams). Team adds 20 workspaces, collaboration tools, and optional HIPAA compliance. Enterprise includes unlimited workspaces, SSO, log streaming, and a dedicated account manager. Two notable add-ons: Agent Gateway for MCP (enterprise control layer for AI agents connecting to external tools, claims 10x token reduction via composite tools) and Merlin Agent Builder (AI agent orchestration layer). Pricing depends on your package, monthly task volume, and add-ons. Contact sales for a quote, and expect negotiation room.
UiPath RPA pricing decoded. UiPath runs on two products: Automation Cloud (core RPA platform) and Test Cloud (agentic testing). Automation Cloud has three tiers: Basic starts at $25/month for individuals, Standard and Enterprise are custom quotes with enhanced governance and self-healing automation. Test Cloud has two tiers, both custom-priced. You can license either product under Unified pricing (single currency of platform units for all workloads) or Flex pricing (pay for what you use as you use it). If your automation roadmap shifts mid-year, Unified lets you redirect platform units without renegotiating. If usage is unpredictable, Flex avoids paying for capacity you won't use. Contact sales for quotes; expect significant negotiation leverage on multi-year deals.
UiPath pricing: Exploring RPA pricing models
VCs are buying legacy companies to rebuild them with AI. Instead of selling AI tools to enterprises, venture firms are buying legacy companies outright and retrofitting them with AI from the inside. General Catalyst calls it "service as software," applying SaaS growth economics to service businesses. The playbook: acquire companies in industries where software adoption lagged (healthcare, accounting, insurance, construction), embed engineers on-site for months, and deploy proprietary AI tuned for those workflows. Long Lake Management acquired 30+ businesses across HOA management, construction, and corporate travel, runs a proprietary AI platform called Nexus that it claims performs 5x better than general-purpose models, and just took American Express Global Business Travel private at a $6.3 billion valuation. Traditional private equity made the opposite bet, buying enterprise SaaS at peak multiples in the early 2020s. Those companies are now the ones most exposed to AI disruption.
Silicon Valley's new buyout playbook is hitting Wall Street
n8n targets 200 UK employees by 2029. Germany's most valuable AI startup (valued at $5.2 billion after a strategic partnership with SAP) announced plans to grow its UK team to 200 by 2029. Over 35% of the Fortune 500 have adopted n8n, with UK customers like Vodafone saving 5,000 person-days and avoiding $2.2 million in costs by automating threat intelligence workflows. The company's fair-code license lets organizations self-host with full access to source code, and its visual interface puts AI workflow construction in the hands of the teams running them, not just engineering. The UK expansion reflects growing enterprise adoption in financial services, tech, and government sectors.
n8n targets a 200 person UK team by 2029
Amazon pays Corning billions for AI data center fiber. Amazon locked in a multi-year deal with Corning for optical fiber to power and connect its expanding U.S. data centers, creating 1,000 jobs at Corning's North Carolina factories. Corning's fiber optic cables are essential for AI infrastructure because they enable fast connections between data centers and the racks and chips inside them. Corning's shares have more than doubled in 2026 and are up sixfold since the end of 2023, driven by AI demand. Nvidia committed $3.2 billion to Corning in May, and Meta pledged $6 billion in January. Optical communications is now Corning's largest and fastest-growing business, and hyperscalers will be its biggest customers by next year.
Corning shares jump 4% after company strikes deal to power Amazon AI data centers
Informatica and Microsoft enhance AI with trusted data integration. Informatica (under Salesforce) announced new integrations with Microsoft at Informatica World 2026. Key capabilities: Model Context Protocol (MCP) servers now available in Microsoft Foundry, letting businesses access data governance and management services directly in Azure. Informatica's Intelligent Data Management Cloud expanded to Microsoft Fabric for mass ingestion of data (billions of rows from 300+ enterprise sources) into Fabric Data Warehouse, cutting compute time and cost. The partnership addresses the 89% of data and analytics leaders who say interoperability is crucial for business success. For small businesses, this means enterprise-grade data management is becoming more accessible for AI adoption.
Informatica and Microsoft Team Up to Enhance AI with Trusted Data Integration
Microsoft's open source packages compromised twice in five weeks. 73 cryptographically verified Microsoft packages on GitHub were compromised to add credential-stealing malware triggered when developers opened them in AI coding agents. The malware (tracked as Miasma) harvests credentials from AWS, Azure, GCP, Kubernetes, password managers, and 90+ developer tool configurations, then spreads laterally through cloud infrastructures. This is the second supply-chain attack in five weeks to breach Microsoft's repository account. The malware steals legitimate Microsoft OIDC tokens used in SLSA provenance attestation, bypassing integrity checks. If you used Microsoft's durabletask Python SDK or any of the 73 flagged packages in AI coding agents, assume compromise and rotate credentials immediately.
For the 2nd time in weeks, Microsoft packages laced with credential stealer
Integration patterns cheat sheet. When connecting systems, choose ETL (extract, transform, load) for analytics and compliance workloads, ELT (extract, load, transform) when the target is a cloud warehouse that handles transformation efficiently, broadcast for one-to-many real-time updates (like customer sign-ups hitting CRM, billing, and analytics), bidirectional sync for two-way updates between systems of record, and CDC (change data capture) for streaming only what changed since the last sync without full table scans. If you're designing a data pipeline and not sure which pattern fits, this is your decision tree.
Integration Patterns: How To Choose for Your Architecture
AI is moving from external tools you integrate to native capabilities you embed. Apple's Shortcuts redesign, NotebookLM's cloud computer, and n8n's OpenTelemetry tracing all reflect the same shift: AI workflows are becoming first-class parts of production infrastructure, not experimental side projects. Meanwhile, VCs buying legacy companies to retrofit them with AI signals that enterprise software adoption has been slower than the industry expected, and ownership is the fastest path to deployment. Watch for more take-privates in service-heavy industries, more observability tooling built specifically for AI agents, and more platforms waiving infrastructure costs to attract developers before the next wave of consolidation hits.