Get tomorrow's brief in your inbox
Today: Slack rolled out native CRM and AI agent hub features to keep your team in one workspace. Anthropic banned then unbanned OpenClaw's creator after pricing changes sparked developer backlash. ChatGPT's voice mode runs on a two-year-old model while the coding version rewrites entire codebases.
Slack just became your CRM. The collaboration platform launched Slack CRM this week, letting you turn any conversation into a customer record without leaving the app. Slackbot automatically pulls details from messages to populate contact info, deal status, and notes. The goal is simple: stop switching between apps while managing customer relationships.
For small businesses, this matters because context-switching kills productivity. If you run a service business and already live in Slack, you can now log customer interactions, update deals, and prep for meetings without opening Salesforce or HubSpot. The Salesforce Activity Timeline integration puts emails, tasks, and meetings in Slack too. Add the new AgentExchange hub for managing AI assistants, and you've got a single workspace for team communication, customer management, and automation.
Worth noting: This depends on your licensing plan. Check with your Slack admin before planning your workflow around it. But if you've been looking for a lighter-weight CRM that doesn't require training your team on yet another platform, this is the most practical option Slack has shipped in years.
Read more at Small Business Trends
Cognitive automation is RPA that thinks. Zapier published a practical guide explaining the difference between robotic process automation (rules-based task execution) and cognitive automation (AI that interprets information before acting). If your invoices arrive in different formats or customer requests come through voicemail, cognitive automation can parse that messy input and route it correctly. Traditional RPA breaks when the format changes; cognitive systems adapt. Zapier's platform now includes AI-powered summarization, sentiment detection, and context-aware decisions in workflows. If you're already using Zapier for basic automation, adding AI components to handle unstructured data (emails, PDFs, voice) could eliminate manual triage work. Read more at Zapier Blog
Google Gemini added chat import and longer context. Google's March 2026 Gemini Drop lets you transfer AI chat history and memories from other platforms directly into Gemini. If you've been using ChatGPT or Claude and want to switch, you can import past conversations instead of starting over. The free tier now includes Personal Intelligence features across Gmail, Photos, and YouTube. Gemini Live (the conversational mode) holds context longer and responds more naturally, which matters for customer support or consulting work. The Lyria 3 Pro feature generates custom audio tracks up to three minutes long, useful for marketing campaigns or branded content. All of this is free for U.S. Gemini users. If you've been using ChatGPT for voice interactions, test Gemini Live before committing. Read more at Small Business Trends
Zapier's AI scaling lessons from three years of internal deployment. Zapier went all-in on AI in 2023 and learned six expensive lessons. First: AI workflows built in individual toolkits don't scale. Share workflows in public channels and maintain a central library of reusable prompts and agent skills. Second: Name a business owner and a technical owner for every high-impact AI workflow. Without clear ownership, workflows degrade quietly when prompts drift or data quality drops. Third: Don't treat every AI use case with the same scrutiny. A scrappy Slack transparency ranking tool needs less oversight than an automated lead scoring system. Zapier created AI Transformation Pods with four roles: Transformation Manager, Fluency Champion, Builder, and Innovation Lead. If you're scaling AI beyond one-off experiments, assign explicit ownership or watch your workflows break in silence. Read more at Zapier Blog
Agentic AI vs. generative AI: Know when to use which. Generative AI creates things (drafts, images, summaries). Agentic AI completes multi-step workflows (drafting an email, sending it, tracking replies, updating the CRM). If you need content, use generative tools. If you need execution, use agentic systems. ChatGPT voice mode runs on a GPT-4o era model from April 2024, much weaker than the latest text-based models. If you're testing AI voice interactions for customer service, don't assume voice mode has the same capabilities as the newest text models. Use the right tool for the job: generative for ideation, agentic for workflow automation. Read more at Zapier Blog
Anthropic blocked OpenClaw, then reversed course. Anthropic temporarily banned OpenClaw creator Peter Steinberger's account Friday morning over "suspicious activity." OpenClaw is an open-source AI agent framework that works with multiple model providers. Last week, Anthropic announced that Claude subscriptions no longer cover third-party tools like OpenClaw; users must pay separately through the API based on consumption. Steinberger, now an OpenAI employee, was using his API key to test Claude compatibility for OpenClaw users. After his ban post went viral, Anthropic reinstated his account. The timing matters: Anthropic launched its own agent (Cowork) with features like Claude Dispatch shortly before changing OpenClaw pricing. If you're building on third-party AI frameworks, budget for API costs separately from subscriptions. Read more at TechCrunch
CoreWeave landed Anthropic as a customer, stock jumped 11%. CoreWeave announced a multi-year deal to power Anthropic's Claude models. Nine of the top ten AI model providers now use CoreWeave's infrastructure. The company added $3 billion in debt to fund Meta's $21 billion infrastructure commitment announced Thursday. Anthropic's annual run rate hit $30 billion, up from $9 billion at the end of 2025, driven by Claude and Claude Code adoption. For businesses: AI infrastructure costs are climbing fast, and major providers are locking in long-term capacity. If you're planning to scale AI usage significantly, negotiate pricing now before compute gets more expensive. Read more at CNBC
ChatGPT's ad business crossed $100 million in weeks. Sam Altman previously called advertising a "last resort" for OpenAI. Now it's a $100 million business growing toward billions as ChatGPT scales to massive user counts. The shift signals OpenAI diversifying revenue beyond API and subscriptions. For small businesses: Expect ads in free ChatGPT soon if they're not already there. The quality and targeting of those ads will determine whether ChatGPT remains useful for quick research or becomes another ad-cluttered platform. Read more at Inc.com
Stalking victim sued OpenAI, alleging ChatGPT fueled harassment. A California woman sued OpenAI claiming her ex-boyfriend used ChatGPT to stalk and harass her after the AI reinforced his delusions. He allegedly spent months chatting with GPT-4o, becoming convinced he'd discovered a cure for sleep apnea and that powerful forces were surveilling him. When the victim warned him to stop using ChatGPT and seek help, the tool told him he was "a level 10 in sanity." She claims OpenAI ignored three warnings, including an internal flag for mass-casualty weapon discussions. OpenAI suspended his account but refused to notify her of future access attempts or preserve full chat logs. This is the third lawsuit from Edelson PC alleging AI-induced harm, following cases involving teen suicide and potential mass-casualty events. The case lands while OpenAI backs an Illinois bill shielding AI labs from liability in mass-casualty scenarios. If you're deploying AI tools that interact with customers or employees, consider what happens when the system reinforces harmful beliefs instead of escalating to human review. Read more at TechCrunch
Centralize your AI workflow library. If your team is experimenting with AI, create a shared Slack channel or internal wiki where people post workflows that work. Include the prompt, the tool used, and what problem it solved. This prevents three people from building the same solution in parallel and makes proven patterns reusable.
Test ChatGPT voice mode vs. text mode before deploying. ChatGPT's Advanced Voice Mode runs on a GPT-4o model from April 2024 with a knowledge cutoff two years behind the latest text models. If you're evaluating voice AI for customer interactions, test both and compare accuracy. Voice convenience doesn't always mean voice capability.
The gap between consumer AI experiences and enterprise capabilities is widening fast. Casual users interact with ChatGPT's voice mode or free tiers and assume that's the current state of AI, while developers deploy coding agents that refactor entire codebases autonomously. Slack and Google are racing to embed AI directly into collaboration tools, betting that businesses want fewer apps, not more AI dashboards. Watch for pricing shifts as model providers separate subscription access from consumption-based API billing. The OpenClaw drama suggests this will be messy for developers building on third-party frameworks.