Most AI tools move on an 18-month product cycle. Big announcements, long betas, slow rollouts. You’ve barely learned the last version before the next one is another year away.
Anthropic’s pace right now is completely different — and if you’re running a business, that pace is the story.
In this post I’ll break down everything Anthropic shipped in March 2026, the real numbers on what businesses are saving, and the simplest possible framework for where to start if you haven’t yet.
Everything Anthropic Shipped in March 2026
In roughly three weeks, Anthropic shipped:
Computer Use + Dispatch — Claude can now control your Mac directly. Open apps, fill forms, navigate browsers, handle files. And Dispatch means you can brief Claude from your iPhone while you’re on the move — by the time you’re back at your desk, the work is done. For business owners, this means tasks that previously took 45 minutes of admin can run in the background without you touching them.
Auto Mode — Claude now makes low-risk decisions without pausing to ask for confirmation. If you’ve used Claude Code, you know that the permission prompts can slow longer tasks down. Auto Mode removes that friction for decisions that don’t need a human sign-off, which makes longer, more complex workflows significantly faster.
Voice Mode — Push-to-talk inside Claude Code, across 20 languages. Hands-free direction, useful if you think better out loud, useful if you’re working across devices.
/loop — Recurring Tasks — Set Claude to check something on a schedule. Every 30 minutes, every hour, daily. No Zapier workflow, no automation builder. Just: “check this and report back.” Monitoring, reporting, and alerting without building a pipeline.
1 Million Token Context Window (Sonnet 4.6) — A million tokens is roughly 750,000 words. That’s your last three years of emails. A full software codebase for a mid-size product. An entire customer database export. You’re no longer feeding Claude summaries or excerpts — you can give it the whole thing.
Enterprise Self-Serve — You can now sign up for enterprise-grade Claude without a sales call. No minimum seat negotiations. SSO, audit logs, spend caps, compliance tools — available like any SaaS subscription. Mid-market businesses no longer have to go through enterprise procurement to access enterprise capability.
Claude Code Included in Team Plan — Claude Code used to be a separate add-on. It’s now included in every Team seat. Lower barrier, full capability.
These aren’t isolated features. They’re a system coming together — each one removing a specific friction point that was slowing down real business use.
What This Costs You To Ignore
The case for using Claude agents isn’t speculative anymore. The numbers are in.
80% of enterprise technical leaders report measurable ROI from Claude agents. eSentire, a cybersecurity firm, compressed threat analysis from 5 hours to 7 minutes. Doctolib’s engineering team is shipping features 40% faster. E-commerce merchants are saving 15–25 hours per week on product tagging and order analysis.
Anthropic’s annualised revenue hit a $14B run-rate — Claude Code alone is driving $2.5B of that. This isn’t a research project. It’s a commercial product with production-scale adoption.
Here’s the framing that matters for smaller businesses: the same leverage — 5 hours of work done in 7 minutes — is available to a one-person consultancy, a 5-person agency, a 20-person brand. The tools aren’t scaled to enterprise size. They’re available right now on a Team subscription.
The gap between businesses using this and businesses not using it is widening. Not in a vague “AI is the future” sense — in a real, measurable, weekly-hours-saved sense. The pace of shipping means that gap is growing faster than it was 6 months ago.
Where to Actually Start
This is where most AI content fails you: it tells you the tools exist and then leaves you to figure out the rest.
If you’re a business owner:
- Identify your highest-friction repetitive task — the one you dread, the one that eats a predictable chunk of every week
- Ask: could Claude do this if I gave it the right context and access?
- Start there — not with a grand AI strategy, with one task
The compound logic here matters. If Claude saves you 2 hours a week, that’s 100 hours a year. At even a modest hourly rate, that’s a real return. Then you find the next task. The goal isn’t to automate your whole business in one go.
If you’re a service provider: Computer Use is the most visual of these updates — and visual sells. Build a demo with a use case relevant to your client’s niche. Show the before (manual, takes an hour) and the after (Claude handles it, takes four minutes). The time-saved metric is the one that lands in a client meeting.
If you want to see how the workflow side of this looks in practice — the actual agent builds, not just the theory — the Claude Cowork setup video is a good place to start. It covers how to structure a workspace so Claude has the context it needs to work autonomously. The full Cowork walkthrough goes deeper into plugins, skills, and how to build this into a working environment rather than a chat window.
Final Thoughts
The features Anthropic shipped this month aren’t interesting because they’re new — they’re interesting because they compound. Computer Use plus Auto Mode plus 1M context plus recurring tasks is a qualitatively different thing from any one of those alone.
The businesses that figure out how to extract value from this will have a structural advantage. The ones that wait until it “settles down” will find that it never does — this is the pace now.
Pick one task. Start there.
What’s the task you most want Claude to handle for your business? Drop it in the comments — I’ll be covering the practical implementations of these updates as they mature.