Get tomorrow's brief in your inbox
Today: Google shipped Gemini 3.5 Flash, a cost-efficient model built for long-horizon automation that cuts project time in half. Meta's WhatsApp Business Agent went global, letting small businesses automate customer support for the first time. Uber capped employee AI spending at $1,500 per tool monthly after blowing through its 2026 budget in four months.
Google released Gemini 3.5 Flash, and this one matters if you run a business that burns time on repetitive, multi-step work. Unlike chatbots that handle one-off questions, 3.5 Flash is designed for what Google calls "long-horizon agentic tasks," meaning projects that span days or weeks and involve planning, execution, and iteration. Think building an application, preparing financial documentation, or maintaining complex software systems.
The pitch is simple: what used to take a developer days or an auditor weeks, 3.5 Flash can help complete in a fraction of the time at less than half the cost of competing frontier models. For small business owners, this translates to real savings. If you currently pay contractors or employees to handle time-intensive projects like app development or compliance documentation, this model could reduce both timeline and cost. The model can "rapidly plan, build and iterate," which means you can prototype ideas faster and adapt to market changes without waiting weeks for results.
Google pairs this with a system called Antigravity, which lets you deploy collaborative subagents for more demanding workflows. When supervised correctly, this setup promises better reliability on multi-step tasks. The catch: you'll need to invest time upfront to learn how to deploy these tools effectively, and your team will need comfort working alongside AI systems. But if you're already stretched thin on operational work, 3.5 Flash offers a credible path to doing more without hiring more people.
Meta Business Agent Goes Global
Meta's AI customer support bot is now available worldwide in WhatsApp, Instagram DMs, and Messenger. The agent can answer customer questions, recommend products, book appointments, qualify sales leads, and hand off to a human when needed. If you run a small business that gets repetitive customer inquiries, this is worth testing. Meta is building features like daily briefings of overnight chats, market research, and calendar management. Larger businesses can create custom agents that connect to Shopify, Zendesk, and Shopee. Pricing is included in WhatsApp Business Premium tiers; large businesses pay based on token usage. This is Meta's play to turn WhatsApp into workflow software, not just a messaging app.
Zapier's Agent Skills Library
Zapier released a public library of pre-built agent skills for marketing, sales, and customer success teams. These are structured, reusable instructions that let coding agents like Codex, Claude Code, or Cursor pull from your CRM data, meeting notes, and approval workflows to produce reviewable work. Examples: Mega Campaign Generator (turns scattered campaign inputs into launch-ready packages), Sales Sequence Builder (creates outbound sequences from CRM and account context), and Customer Deck Builder (generates presentations in Gamma from account data). The skills work with Zapier MCP and SDK, so agents can connect to your existing systems without exposing credentials. If you use coding agents but struggle to keep them grounded in your business context, this library solves that problem.
Google Dreambeans: AI Lifestyle Suggestions
Google Labs launched Dreambeans, an iOS and Android app that uses your Gmail, Calendar, Photos, YouTube, and Search History to generate 10-14 curated "stories" daily. These are lifestyle suggestions like nearby coffee shops, event reminders, or puppy care tips based on your calendar. It works while you sleep, processing all your connected Google data overnight and handing you a "concentrated drop of inspiration" in the morning. Privacy controls let you choose which services to connect and delete data anytime. Only your stories are visible to you. Available to U.S.-based Google AI Ultra subscribers, with a waitlist for personal Google accounts. If you're drowning in data across Google services and want a single morning digest, this could help. If you're already skeptical of Google's data collection, skip it.
Microsoft's MAI-Thinking-1 Reasoning Model
Microsoft released MAI-Thinking-1, its first reasoning model, built from scratch for math, coding, and enterprise deployment. It's cheaper than OpenAI equivalents on some tasks, which matters in the current AI cost crunch. Microsoft is years behind OpenAI and Anthropic here, but the company emphasized that the model was developed without distillation (not trained using another company's AI model). If you're an enterprise buyer evaluating reasoning models and care about pricing, this gives you a non-OpenAI option. Microsoft also announced MDASH, a cybersecurity tool that uses 100 AI agents to find exploitable bugs. Both products are aimed at government and enterprise markets where Microsoft competes directly with Anthropic and OpenAI.
Endava Redesigns Software Delivery Around AI Agents
Endava, a global technology services company, made OpenAI its enterprise AI platform and embedded ChatGPT Enterprise and Codex across its 11,000-person workforce. The result: more than half of enterprise customers now use either Endava's AI agent (Olly) or their own AI models through command-line and agentic interfaces to investigate incidents and query operational data. The shift changed how teams work. Legal teams streamlined research and documentation. Project managers generate governance reports and summarize engineering progress in Codex. Commercial teams replaced spreadsheet-heavy planning with lightweight AI-generated apps. In one pricing discussion, employees built a single-page pricing app instead of passing around spreadsheets, which "changed the conversation completely."
The lesson for small businesses: AI adoption is a behavior change, not a software rollout. Leaders need to actively use AI to drive organization-wide adoption. Hands-on experience is the fastest way to overcome skepticism. Make AI part of everyday workflows, not a separate initiative. Endava's AI-native delivery methodology, DavaFlow, embeds OpenAI technology throughout the entire software delivery lifecycle, from meeting prep to deployment.
Wasmer Built a Node.js Runtime in Two Weeks with Codex
Wasmer, a small edge computing company, used Codex to build Edge.js, a JavaScript runtime that runs Node.js workloads in a WebAssembly sandbox. This effort would have taken one year without Codex; with Codex, it took two weeks. Wasmer engineers increased development speed by 10x to 20x. They used Codex from start to finish: building architectural blocks, finding bugs, identifying root causes, and polishing the final product. Codex mastered low-level debugging tools and spotted C++ subtleties the team didn't know existed because they weren't C++ experts. Before Codex, this project would have been impossible for a small team. Now, Wasmer is the first cloud host to provide full Node.js at the edge layer.
The takeaway: small companies can now achieve things that were only possible at big companies. If you have ambitious technical projects but lack the team size or expertise, coding agents like Codex can multiply your efforts. The key is learning to guide the agent toward the goal rather than writing every line of code yourself.
Uber Caps AI Tool Spending at $1,500 Per Employee Monthly
Uber blew through its 2026 AI budget in four months and responded by capping employee spending at $1,500 per month per AI coding tool (like Cursor or Claude Code). Spending on one tool doesn't affect the budget for another. At two actively used tools per engineer, that's a $36,000 annual cap per employee. For context, Uber's median software engineer compensation is $330,000, meaning the AI spending cap is about 11% of total comp. If you're a business considering AI tool budgets, this ratio provides a useful benchmark. Executives report that AI generates at least 10% of Uber's code, but the token-burning nature of coding agents makes cost management essential. The lesson: set limits upfront rather than chasing unlimited usage.
Lovable Signs Multiyear Deal with Google Cloud
Lovable, the Stockholm-based vibe-coding startup, expanded its Google Cloud usage by 5x under a new multiyear deal. The agreement gives Lovable expanded access to both Anthropic's Claude and Google's Gemini models. Lovable crossed $400 million in annualized revenue in February, adding $100 million in a single month with just 146 employees. More than half of Fortune 500 companies use its product. Lovable's agent will be available through Google Cloud's Gemini Enterprise Agent Gallery, simplifying enterprise procurement. The company will integrate with Wiz (Google's $32 billion security acquisition) to identify and remediate security problems in real time. If you're evaluating low-code or AI-powered development platforms, Lovable's rapid growth and Google partnership signal enterprise credibility.
Coralogix Raises $200M to Monitor AI Agents
Coralogix, a Boston-based software-monitoring startup, raised $200 million at a $1.6 billion valuation, betting that AI agents will drive demand for monitoring, troubleshooting, and management tools. The Series F round came just 11 months after its $115 million Series E, reflecting investor appetite for AI infrastructure. Coralogix collects and analyzes operational data (logs, metrics, traces) to help companies detect outages, investigate incidents, and optimize applications. More than half of enterprise customers now use either Coralogix's AI agent (Olly) or their own AI models through command-line and agentic interfaces to investigate incidents. CEO Ariel Assaraf says engineers are increasingly interacting with software through AI assistants rather than traditional dashboards. Coralogix grew revenue by more than 60% over the past year and counts about 30 customers spending more than $1 million annually.
Morgan Stanley Opens Wealth Management to AI Agents
Morgan Stanley will allow clients' autonomous agents to pull data and insights directly from its stock administration platforms (ShareWorks and Equity Edge), bypassing traditional software interfaces built for humans. The bank has granted a handful of clients early agentic access and plans to open it to its 3,400 administration clients by next year. This is one of the earliest instances of a major Wall Street bank opening its platforms to external AI tools. The move is based on the Model Context Protocol, an open-source standard that allows AI models to plug into data sources. Morgan Stanley executives attributed $1.2 trillion in assets gathered to its workplace strategy. For small business owners managing stock compensation plans, this signals a shift: future interactions with financial platforms will likely happen through AI agents, not web logins.
Alphabet Raises $85 Billion for AI Infrastructure
Alphabet's record-breaking $85 billion stock sale signals voracious investor appetite for AI. Google's parent company initially intended to sell $40 billion worth of equity instruments but raised $45 billion instead due to oversubscription. Berkshire Hathaway picked up $10 billion. Alphabet plans to sell another $40 billion next quarter. The money is earmarked for AI infrastructure and data centers. CEO Sundar Pichai said the company expects to spend between $180 billion and $190 billion on capital expenditures before the year is out. The timing matters beyond Alphabet: as Anthropic prepares to go public, this enormously successful stock sale indicates that public investors are ready to fund AI companies at scale. For small businesses, this means the AI infrastructure you rely on (Google Cloud, AI APIs, enterprise tools) will continue to expand rapidly.
Europe Unveils Tech Sovereignty Package
The European Commission proposed new rules to bolster homegrown chips, AI, and cloud services amid growing reliance on U.S. and Chinese tech. The Cloud and AI Development Act (CADA) sets sovereignty levels for cloud computing for sensitive workloads at public organizations. Commission President Ursula von der Leyen: "We cannot afford to depend on others for the technologies that keep our hospitals running, our energy grids stable and our services secure." The rules will make it difficult for U.S. companies to reach the highest sovereignty levels because of the U.S. Cloud Act, which allows law enforcement to request user data from American companies regardless of where the data is stored. Chips Act 2.0 aims to build capacity in cutting-edge semiconductor technologies that power AI, prioritizing a foundry for advanced semiconductor manufacturing within the EU. If you sell cloud or AI services to European clients, these rules will affect procurement requirements.
Set Up Meta Conversions API for Better Lead Quality
If you run Facebook lead gen ads, you're likely optimizing against the wrong outcome. Meta sees someone submitted the form but has no idea if they were a qualified lead, a tire-kicker, or a bot. The fix: Meta's Conversions API for CRM (CAPI), wired through Zapier. Set up three events: (1) New lead event when a lead lands in your CRM, (2) Funnel event when the lead progresses through a meaningful stage (appointment booked, demo completed, intro class attended), (3) Purchase event when revenue happens (contract signed, payment processed, membership starts). According to Meta, advertisers using CAPI for CRM see roughly 21% lower cost per quality lead on instant form campaigns and 9.5% lower on website form campaigns. The data Meta needs already exists in your CRM. Wire it up and let Meta optimize for actual customers, not just form-fillers.
Use Google Lens to Price Thrift Store Finds
If you resell products or run a secondhand business, Google Lens can tell you what items are worth before you buy them. Snap a photo, use Lens to find similar items, and browse what they sell for online. You can also see how many similar offers exist to determine if something is a rare gem or common. Use Circle to Search on Android to find similar items and prices while scrolling. If you're clearing out inventory, Lens can help you price items for resale. Ask "Could I resell this?" or "What kind of stores buy items like this?" to identify resale opportunities.
AI tools are shifting from individual task assistance to full workflow orchestration. Google's Gemini 3.5 Flash, Meta's Business Agent, and Morgan Stanley's agentic platform access all point to the same trend: AI is moving from answering questions to executing multi-step processes autonomously. For small businesses, this means the tools you evaluate today should handle end-to-end workflows, not just one-off tasks. Watch for cost management strategies as companies like Uber set spending caps to control token burn. The gap between early adopters who multiply their productivity and businesses that hesitate is widening. If you're not testing AI agents on repetitive workflows this quarter, you're falling behind competitors who are.