On 13 August a federal judge let key parts of the class action against Otter go ahead, over claims its notetaker recorded people without their consent and used those recordings to train its models. In July a similar case was filed against Granola, alleging meeting content was used for AI training by default unless you switched it off (ToolDirectory).

The uncomfortable line in that coverage: many US states require every participant's consent to record, the obligation generally sits with whoever runs the meeting, and no product obtains it for you.

The fix is a notetaker that never leaves your computer. No bot in the call, nothing uploaded, no subscription. Ten minutes to try.

What it replaces: Otter, Fireflies, Granola, Fathom, and the paid transcription, dictation and captioning tools you use for everything else. Granola's business plan alone is $14 per user per month.

The 20-second version: Local transcription apps now run speech recognition on your own machine, capture calls without a bot joining, and work with the wifi off. Your recordings stay on your computer as files you own.

Time: ten minutes. You need: a Mac or Windows computer with 8GB of RAM or more. Payoff: one less subscription, and recordings nobody else can read or train on.

Most people who dropped Otter switched to a tool with no bot and assumed the problem was solved. It was not.

Granola, Fathom, Fellow and Krisp all skip the visible bot, but your recording still uploads to their cloud. Only a handful of tools have no bot and are fully local, meaning nothing leaves your machine (Humla).

A bot you can see is a consent problem. An upload you cannot see is a data problem. Local solves both.

Start with a recording you already have: a voice memo, a downloaded Zoom file, anything.

On a Mac: MacWhisper has a no-cost version that transcribes on your device, and once its model is downloaded it works offline (App Store).

On Windows or Mac: Meetily's Community Edition is open source and runs fully offline (Meetily).

What you get: a full transcript produced on your own machine, with no internet connection at all. The same thing your notetaker does, with nothing sent anywhere.

Once it is installed, the same engine handles any audio: voice memos you want turned into documents, videos that need captions, interviews you need quotes from, a screen recording you want turned into a how-to guide. Most people are paying for two or three tools that this one replaces.

Transcribing a file is the easy half. Capturing live calls without a bot, getting speaker labels, producing useful summaries, and pointing it at those other jobs all need a little setup. That is the Pro section.

The template library has more prompts sorted by job, and the resource library has every guide from past issues.

Pro gets the full setup: which tool for your machine this month, live call capture with no bot, local summaries, the consent line, and seven jobs the same engine replaces, each with a prompt.

Also in your library: the Connector Playbook, Google's AI Tools guide, the Skills Bank, and every workflow pack from past issues.

Launched this month. Give it your website and it identifies your business, your products, your buyer personas and up to 10 competitors, then shows you how AI search engines describe you and them.

If you ran the AI Visibility Audit from our resource library, this is the faster, automated version of the first half: what AI says about you, next to what it says about the businesses it recommends instead.

It is brand new, so check the pricing and what is included before you rely on it. Use it to see where you stand, then do the fixing yourself.

🧰 Meta took its Muse agent into business. Muse for Small Business connects the agent to Asana, Zoom, Intuit, Box, Canva and Slack, at the same pricing as the consumer app (CNBC).

🛍️ TikTok launched a shopping agent with one-click checkout. Announced on 5 October, it remembers user context, guides discovery, sizing and availability, and completes purchases with partners including Shopify and Stripe (AI Agent Store).

⚡ Anthropic released a much cheaper small model. Claude Haiku 5.5 arrived on 7 October as a fast, low-cost model for high-volume work like classification, extraction and routing, with a 1M token context window (AI Agent Store).

💼 McKinsey says 11 million Americans will need a new occupation by 2035. AI and automation are expected to cut demand for about 36 million US jobs while growth elsewhere creates about 41 million. Only one in seven displaced workers has a direct path into a growing job (Fortune).

Quick version: 48% of US employees say they have exaggerated their AI use or expertise to coworkers or bosses. At the same time, employers stopped paying for the thing being faked. Demand for basic AI skills fell 25% this year, while demand for people who can build tools and automate workflows rose 51%.

A survey of 1,000 full-time US employees, run in part by Visier, found 48% had exaggerated how much they use AI, or how good they are at it, to coworkers or bosses. The researchers call it "performative AI." The main reason given: people want to look experienced, but get little training or guidance (TechSpot).

The management side is not much better. Only 28% of employees say their employer understands how AI is actually used at work, and 33% think their bosses understand how much training they need.

Now the twist. Workday's October workforce report shows employers have already moved on from the skill people are faking. Demand for basic AI skills fell 25% this year. Demand for hands-on skills like building AI tools, automating workflows and AI engineering rose 51% between September 2025 and July 2026. Companies want people who can build with AI, not just use it.

And mentions of training skills in job postings fell 13% over the same period (Workday). So the skill that matters is rising, the training to get there is shrinking, and half the workforce is covering the gap by pretending.

If you run a business, you probably do not know what your team actually does with AI. If you work for one, "good at ChatGPT" stopped being a differentiator this year.

🔒 The Pro breakdown: the 15-minute way to find out what your team really does with AI, the before-and-after log that proves real skill, and the three build skills employers are now paying for.

Workflow Library Updates:

NEW Resources:

Each resource lives permanently in your Pro account. Use them whenever you need them.

A notetaker is one of the most useful AI tools there is. It is also the one most likely to be listening to a client who never agreed to it, on a server you have never seen. The same engine on your own computer does that job and six others.

Try the ten-minute version this week. Wifi off, one recording, and see what your own computer can do.

👇 What’s behind the paywall:

Pro members deploy AI in their work an average of 5-8x more often than free readers (based on reply data from past issues). The difference is having the exact setup, not the concept.