👋 Hello hello,
Claude users, you got a win this week. If you've ever had to split a single project into three prompts just to get through it, that's mostly behind you now.
Manus also rebuilt how scheduled tasks work, and it's a bigger deal than a changelog entry. Plus, 87% of law firm GCs are already using generative AI — and Anthropic just published the playbook they're running from.
Quick heads up before we dive in.
Starting today, we're adding something new: PAI Labs.
PAI Labs was born from conversations we kept having inside the community and with our own team. Founders, solopreneurs, operators — everyone asking the same thing: where are the workflows that actually work, not just look good in a demo?
So that's what this is. Tried and tested systems we've built and run in-house, now shared with you. The kind of stuff that genuinely moves the needle on your time, your costs, and your sanity. 🙂
A few times a week, we'll drop an exclusive PAI Labs playbook inside this newsletter. Paid subscribers get full access. If you're on the free plan, you'll get a glimpse of what's inside — and can decide if it's worth unlocking. We think it will be.
Let's get into it.
🔥🔥🔥 Three Curated AI Updates
Tokens are the unit of work Claude processes per session — more of them means longer documents, more complex projects, and conversations that stop hitting a ceiling halfway through.
For Claude Design users specifically — the built-in feature for generating visual outputs like infographics, slides, and branded assets — this means richer briefs and more context to build from in a single session. For everyone else, it's fewer interruptions on bigger work.
Available now across all plans.
Manus is an AI agent platform that lets you build and run autonomous tasks on a schedule. Think of it as AI doing things on your behalf, repeatedly, without you having to be there. With Scheduled Tasks 2.0, they've rethought how that works from the ground up.
Before, each scheduled run started completely fresh. Now, tasks can continue inside the same thread — so context carries over between runs. Scheduled tasks can also power background actions inside web apps you build on Manus. And there's a cleaner dashboard showing what's run, what's coming up, and what failed.
Here's a number worth pausing on: 87% of general counsel now use generative AI in their teams. The legal industry moved fast, quietly, and with very little press coverage.
Anthropic published a practical guide for legal teams this week covering how Claude actually fits into legal work — contract review and redlining, M&A diligence, regulatory monitoring, litigation prep, privacy assessments, and more.
Worth noting: Anthropic's own legal team uses Claude to speed up contract redlining and marketing content reviews. So this isn't a theory. They're running it themselves.
🧪 PAI Labs
The average person spends close to an hour every morning just clearing their inbox. Reading, thinking, drafting, second-guessing, sending. Repeat 27 times before 10am.
We built a workflow that killed that. Using Claude or Codex, you can set up an automation that goes through your unread emails every morning, drafts replies in your actual voice, and has them ready before you've finished your coffee. The whole inbox clear-out goes from an hour to under ten minutes.
In this guide we share the exact workflow, step by step. 👉 Full playbook here
🔥🔥 Two Things the AI Content Machine Gets Wrong
1. 🧠 Some Workflows Are Actually Worse With AI
There is a lot of content out there showing AI doing everything — video editing, carousels, infographic design. And technically, yes, an AI like Claude can do those things. But should Claude be doing it? Simply, no.
Because in this instance, Canva templates plus a quick text edit will get you to a better output faster. For video editing — silence cutting, captions, motion graphics — native AI features inside dedicated tools like Veed are far more refined. These tools were built for those jobs. Claude was not.
Knowing the difference is what actually saves you time, and the honest answer is: sometimes the old way is still the best way.
2. 🏛️ Hot Take: The AI Agent You Train Should Be Your IP
If you build an AI agent on your own expertise — your process, your judgment, your way of working — who owns it?
The argument is that you do. If a company wants to use an agent that runs on your knowledge, they should license it or pay for access the same way they pay for your time. This matters more as the idea of fractional work grows: one person running roles across multiple companies, with their agent doing a lot of the execution in each.
The decisions made now about ownership and compensation will define how that world plays out. Worth thinking about before someone else decides for you.
🔥 Things You Didn’t Know You Could do With AI
Here's one of the cleaner two-tool workflows doing the rounds right now. You need ChatGPT and Replit — both free to start.
Quick context: Replit is a browser-based builder where you can create apps, tools, and presentations just by describing what you want in plain English. It has a dedicated Slides mode that builds React-based decks — fully editable, exportable to PPTX or PDF, and shareable as a live link.
Open ChatGPT and use the image generation tool. Describe a visual style — color palette, layout feel, overall vibe. Try something like: "clean dark-mode slide design with neon green accents and minimal type."
Screenshot the generated image. This becomes your style reference.
Go to replit.com, open a new project, and select Slides from the app type dropdown.
Upload your screenshot and type: "Build a slide deck on [your topic] in this visual style." Replit generates a full structured deck matching the aesthetic.
Refine through Replit's chat — "make the fonts larger," "add a competitive landscape slide" — then export as PPTX, PDF, or share as a live URL.
For quick internal decks, client briefs, or content repurposing, this is fast. And the visual consistency beats anything you'd get from a blank generic template.
💬 Quick poll: What's an AI workflow you tried that ended up being more hassle than it was worth?
Don't forget to rate today's post
Until next time,
Team @PracticalyAI
