Useful AI tips.
Every week.

Anthropic Just Plugged Claude Into Your Business Apps — What It Does and Who It's For

Anthropic launched Claude for Small Business on May 13. It's not a new model or a new website — it's a version of Claude that installs inside the tools small businesses already use. QuickBooks, HubSpot, Canva, DocuSign, Google Workspace, Microsoft 365, Slack. You connect your apps, assign a task, review what Claude prepares, and approve before it does anything. This issue covers what it actually does and whether it's worth looking at.

What it is (and what it isn't)

Claude for Small Business is powered by Anthropic's Cowork mode, which is their desktop tool for running AI across local files and connected services. The small business version adds 15 ready-to-use workflows — pre-built task sequences for things like payroll planning, invoice follow-up, contract review, and monthly reporting.

You don't build anything. You connect your apps, pick a workflow, and Claude does the preparatory work. The key phrase Anthropic keeps repeating is that Claude "prepares and drafts" — you approve before it sends, posts, or charges anything. It is not automated in the sense of running while you're not watching.

The connected apps list

At launch, the apps it works with are: QuickBooks, PayPal, HubSpot, Canva, DocuSign, Google Workspace (Docs, Sheets, Drive), Microsoft 365 (Word, Excel, Outlook), and Slack. That's a reasonable list for a business running on standard tools.

The 15 included workflows cover: payroll planning, month-end close reporting, business performance monitoring, marketing campaign drafts, cash-flow forecasting, invoice chasing letters, contract review summaries, lead triage from HubSpot, content strategy drafts, and a few others across finance, sales, and marketing.

A realistic example of how it works

Say you want to chase an overdue invoice. You open Claude, pick the "Invoice chasing" workflow, and tell it which client. It pulls the invoice data from QuickBooks, drafts a polite follow-up email, and shows it to you. You read it, edit if needed, and click to send via Outlook. The whole thing takes two minutes instead of ten.

The same applies to the contract review summary. You upload the PDF, pick the workflow, and Claude prepares a plain-English summary of the key terms, payment conditions, and anything that looks like an unusual clause. You review it before acting on it. This is a task that would normally take an hour or cost $300 in paralegal time.

Who it's for

This is designed for small businesses with 1–25 staff who currently use a mix of the tools above. It's not for enterprise teams (that's a separate product) and it's not aimed at solo contractors who don't use accounting software. The sweet spot is a 5–20 person business where the owner or office manager is doing a bit of everything and wasting time on repetitive admin.

Pricing wasn't published at launch — Anthropic directed interested businesses to a waitlist at claude.ai/small-business. Given the Cowork pricing model it's likely in the A$50–A$120 per month range, though that's not confirmed.

The bigger picture

What's notable about this launch isn't the technology — it's the packaging. Anthropic built pre-wired workflows instead of asking small business owners to figure out prompts themselves. That's the right approach for this audience. Expect Google and Microsoft to respond with their own versions before the end of the year.

ChatGPT Is Now Inside Excel and Google Sheets — Here's How to Use It

OpenAI has quietly shipped one of the most useful things it has ever built for office workers: a ChatGPT sidebar that lives inside Microsoft Excel and Google Sheets. No switching tabs. No copying and pasting. You describe what you want in plain English and it builds, edits, or explains the spreadsheet right there. This issue covers how to install it and what it's actually good for.

How to get it in Google Sheets

Open any Google Sheet. Go to Extensions → Add-ons → Get add-ons. Search for ChatGPT and install the official OpenAI add-on. Sign in with your OpenAI account. The ChatGPT panel will appear on the right side of your sheet. That's it.

It's available on Free, Plus, and Pro plans. Business and Enterprise users get a free preview through June 2026, after which it's included in the plan.

How to get it in Excel

Open Excel and go to Insert → Add-ins → Get Add-ins. Search for ChatGPT and install it from the Office Add-ins store. Sign in with the same account you use at chatgpt.com. The sidebar opens from the ribbon.

What it's actually useful for

This is not a formula explainer. It can build and edit the spreadsheet directly. Here are the tasks where it saves the most time:

Building a tracker from scratch. Describe what you want to track — budget categories, project statuses, weekly task completions — and it generates a fully formatted sheet with headers, formula columns, and data validation where relevant. Tell it "add a column that shows percentage of budget used" and it adds it.

Cleaning up messy data. Paste in a sheet full of inconsistently formatted entries — names in all-caps, dates in three different formats, phone numbers with and without country codes — and ask it to standardise everything. It rewrites the data in place.

Writing formulas in plain English. Instead of Googling "how to write a SUMIF with two conditions," just type what you want: "Sum column C only where column A is 'Marketing' and column B says 'Q2'." It writes the formula and explains it.

Understanding a spreadsheet someone else built. If you've inherited a complex workbook full of formulas you don't understand, you can ask: "What does this formula do?" or "Walk me through how this sheet is structured." It reads the sheet and explains it in plain English.

What it won't do (worth knowing)

It can't connect to live data sources or pull from other systems. It works with what's in the spreadsheet in front of it. If your data is in a separate database or another file, you'll need to paste the relevant section in first. It also won't execute macros or run VBA — this is for building and editing spreadsheet content, not automating click sequences.

One thing to check first

If your organisation manages your Google or Microsoft account, the add-on may be blocked by your IT team. If the install button is greyed out, that's why. Check with your IT department before raising a support ticket — some organisations have already approved it and others have blocked it by policy.

Stop Re-Explaining Yourself to ChatGPT: The Memory and Projects Setup

Every time you open a new ChatGPT chat, it knows nothing about you. Your job title, your company's tone, the project you've been working on for three months — gone. This issue covers two features that fix the problem. Most people have them switched off by default and have never touched them.

Memory: what it is and how to turn it on

ChatGPT's memory feature lets it store specific facts about you and recall them in future conversations. Not summaries of past chats — actual notes that persist across every session.

Go to Settings → Personalisation → Memory and make sure it's switched on. Once it is, you can start building a useful memory bank by telling ChatGPT things like:

"Remember that I'm an operations manager at a construction company. My audience for most writing is non-technical project owners. I prefer short paragraphs and plain English."

Or: "Remember that I send weekly status updates every Friday. The format is: summary, three bullet points, one risk item."

ChatGPT will save these notes and reference them automatically. To see exactly what it's stored, type "What do you remember about me?" at any time. To edit or delete individual memories, go to Settings → Personalisation → Memory → Manage memories.

Memory is not unlimited — it stores roughly 1,200 to 1,400 words in total. Keep entries short and specific. Once it fills up, ChatGPT stops adding new memories until you clear some space.

Projects: the better option for recurring work

Memory works across all your chats. Projects work at a task level — they give you a dedicated workspace for a specific type of work, with its own files and instructions, separate from everything else.

To create one, click Projects in the left sidebar and then New Project. Give it a descriptive name — "Monthly Board Report," "Client Emails — Henderson Account," or "HR Policy Documents."

Inside the project, you can upload files (templates, style guides, examples) and write a set of custom instructions that apply to every chat in that project. The instructions are the most important part. Here's a starter you can adapt:

Which one should you set up first?

Use Memory for general preferences that apply across everything — your job title, your audience, your tone preferences, recurring formats you use.

Use Projects for a specific recurring task that has its own files, rules, or context — weekly reports, a specific client account, a type of document with its own template.

Together, they mean you spend the first minute of each chat giving context instead of spending it every time.

One thing to know about Projects

Each ChatGPT Project has its own memory that's separate from your general memory. If you want the project to know things stored in your general memory, you may need to add them to the project instructions as well. It's a small extra step, but it means your project stays focused rather than pulling in unrelated context from other work.

Why AI Is Making You Work More (Not Less) — And the Simple Fix

A 2026 study tracking over 10,000 workers found something uncomfortable: after adopting AI tools, time spent on most job tasks went up — in some cases by more than 300%. If AI is supposed to save you time, why does your workload feel bigger? The research has an answer, and so does this issue.

The pattern the data shows

ActivTrak tracked 10,584 workers for six months before and after they started using AI tools. The result wasn't what anyone expected: time spent across every job category increased, not decreased. A separate study from Fortune found that time spent emailing specifically had doubled for many AI users, and uninterrupted, focused work sessions fell by 9%.

The researchers' explanation is straightforward. AI makes it easier to produce things — emails, reports, documents, summaries. So people produce more of them. Managers expect more output from the same number of staff. The bar shifts upward. The tool saves you 20 minutes on a task, and your employer fills that 20 minutes with something new.

That's the macro problem, and you can't individually fix your company's expectations. But there's a second problem that you can fix, and it's the one costing most individuals the most time.

The real time sink: fixing AI output

A separate 2026 survey found that 66% of workers spend six or more hours a week correcting AI mistakes. The term researchers are using is "workslop" — AI-generated content that looks polished but contains errors that require checking and reworking.

The problem isn't AI; it's how people are using it. A vague prompt produces a vague output. A vague output requires significant editing. Editing AI is slower than editing your own writing, because you have to read it carefully for errors rather than just polishing something you already understand.

The fix: be more specific, not more productive

The people who are actually saving time with AI have narrowed what they use it for. They don't use it for everything — they use it for the 3-4 specific tasks where a precise prompt produces an output they can use with minimal editing.

The rule is: the more specific your prompt, the less editing you do. Compare these two approaches:

❌ "Write a summary of this project for my manager."

✅ "Write a 150-word summary of this project for my manager Sarah. Include: what we completed this month, the current budget status (on track), and the next milestone (June 30 testing phase). Tone: professional, direct. No jargon. No filler phrases."

The first one produces something generic you'll spend 10 minutes fixing. The second produces something close to ready.

What actually works

The desk workers who report the biggest time savings use AI for three things only: first drafts of structured documents (with very specific prompts), summarising long content they need to read anyway, and turning rough notes into formatted outputs. Everything else, they do themselves. Less AI, used more precisely, saves more time than AI applied to everything.

Stop Re-Explaining Yourself: Set Up a Claude Project in 5 Minutes

Every time you start a new Claude chat, you lose all context. Your tone preferences, your company's style guide, that briefing doc you uploaded last Tuesday — gone. You end up pasting the same background info over and over. Claude Projects fix this by giving you a permanent workspace where the context stays put.

What a Project actually is

A Project is a folder inside Claude that holds three things: your uploaded files (the "knowledge base"), a set of custom instructions, and all the chats you have inside that project. Every conversation in the project can see the files and follows your instructions automatically. You set it up once. Then every chat starts with Claude already knowing what you need.

Projects are free for all users — you get up to five. Paid plans (Pro, Team, Enterprise) give you more capacity and the ability to share projects with colleagues.

How to set one up

Go to claude.ai/projects and click + New Project. Give it a clear name — "Q3 Board Report" or "Weekly Client Emails," not "Stuff." The name is for you; Claude doesn't read it.

Next, add files to the knowledge base. Click the + button on the right side of your project page. Upload your style guide, a brief, a template, sample outputs — whatever Claude needs to do the job well. Clean your files before uploading. Remove headers, footers, and irrelevant pages. The less noise, the better the answers.

Then write your custom instructions. This is the most important step. Tell Claude what the project is for, what role it should play, and how you want replies formatted. Be specific. Here's a starter prompt you can copy and edit:

That's it. Start a new chat inside the project and Claude already knows the brief, the tone, and the rules. No re-explaining.

Three projects worth setting up first

Email drafter. Upload your company's tone guide and a few example emails you've sent. Set instructions for length, sign-off style, and formality level. Every email you draft inside this project sounds like you wrote it.

Meeting notes processor. Upload your team's action-item template. Tell Claude to pull out decisions, owners, and deadlines from pasted transcripts. Same format every time, no extra prompting.

Document reviewer. Upload your organisation's report template and style rules. Paste drafts in and ask Claude to check structure, flag missing sections, and tighten the language.

One thing to watch

Each project is isolated. Your "Email" project can't see files from your "Reports" project. This is actually useful — it keeps Claude focused — but it means you'll need to upload shared files (like a style guide) to each project that needs them.

The 5-Minute AI Audit: Is Your Workflow Actually Saving You Time?

Most people start using AI tools with enthusiasm, then quietly fall back to old habits. This week: a fast self-audit to check whether AI is genuinely working for you — or just adding friction.

Why the drift happens

Three months into using AI tools, most professionals have 2-3 workflows that genuinely save time and another 5 they tried once, got mediocre results, and quietly abandoned. The problem is they never figured out why it didn't work — they just stopped.

The fix is a quick audit. Not a big project. Five minutes. Right now, if you like.

The 5-minute AI audit

Open a blank doc or note and answer these four questions honestly:

1. What do you use AI for right now? List every task, even if it's just occasional. Don't filter.

2. Which of those actually saves you time vs creates more work? Be honest. Some AI outputs take longer to fix than doing it yourself.

3. What work task do you dread every week? That's probably your best AI opportunity — one you haven't explored yet.

4. What's stopping you from using AI for that dreaded task? Usually it's not knowing how to prompt it, or fear the output will be wrong.

The rule that changes everything

AI should make you faster at tasks you already do well — not replace your judgment. If you're spending more time editing AI output than it would take to write it yourself, the prompt needs work, not you.

Stop Writing Emails From Scratch. Here's the 3-Step AI Email System

The average professional writes 40+ emails a day. That's 3-4 hours of your week on correspondence alone. This issue: the exact 3-step system to cut that in half using any AI chatbot — no special tools required.

Why most people get bad AI emails

They ask AI to "write an email about X" with no context. The result is generic, wordy, and sounds nothing like them. They spend more time editing than writing. They give up.

The issue isn't the AI — it's the prompt. Good email prompts have three parts.

The 3-part email prompt formula

Part 1 — Context: Who are you, who is this to, what's the relationship, what's the situation?

Part 2 — Outcome: What do you need to happen after they read this email?

Part 3 — Constraints: Tone (formal/casual), length, anything to avoid.

The tighten pass

Once you have a draft you like, run this second prompt: "Shorten this by 30% and remove any filler phrases. Keep the core message and tone."

This single step turns a decent AI email into a great one. Most AI first drafts are 20% too long.

AI at Work: What's Safe to Type, What Isn't

One of the biggest blockers for professionals using AI at work isn't skill — it's uncertainty about what they're allowed to share. This issue covers the privacy basics every desk worker needs to know.

The core principle

Public AI tools (ChatGPT, Claude, Gemini on free/personal plans) process your input on external servers. Most providers say they don't train on your data by default, but your organisation's data governance policy may have stricter rules.

Before you paste anything into an AI tool at work: would you be comfortable if your manager read exactly what you typed? If yes, proceed. If not, pause.

Generally safe to share

Generic task descriptions ("summarise this type of document"), your own writing that doesn't contain confidential information, publicly available information you're researching, draft ideas that don't reference specific clients or deals.

Generally not safe to share

Client names, personal data, financial figures, HR information, legal matters, contracts, anything marked confidential, passwords or API keys, medical information, or information covered by NDA.

What to do at your organisation

Check whether your employer has an AI usage policy. Many larger organisations now have approved tools (often Microsoft Copilot within M365) where data stays inside the company's environment. If your company has one — use it.

Meeting Notes in 90 Seconds: The AI Workflow That Changed How I Work

Meetings are time thieves twice over — once when you're in them, and again when you spend 20 minutes writing up notes afterwards. This week's workflow eliminates the second theft entirely.

The old way

You leave a meeting, open a doc, stare at your scattered notes, try to remember what was actually decided, write up action items that are missing key context, send an email recap that half the team ignores. Sound familiar?

The AI meeting notes workflow

Step 1 — During the meeting: Open your notes app and type key points as bullet fragments. Don't worry about grammar or completeness. Just capture: decisions made, who said what matters, actions mentioned, anything unclear.

Step 2 — After the meeting (takes 2 minutes): Paste your messy notes into your AI tool with this prompt:

Step 3 — Review and send. Spend 60 seconds checking it's accurate, then send. Done.

What changes when you do this consistently

People stop saying "wait, what was decided in that meeting?" Your action items actually get done because they're clearly documented. And you leave every meeting feeling 10% less exhausted.

Welcome. Here's the One Thing That Actually Matters About AI at Work

First issue. No fluff. Just the one insight that separates people who get real value from AI from those who don't — and it's not about which tool you use.

The mistake everyone makes

Most people approach AI tools as search engines. They type a question, get an answer, feel mildly impressed or mildly disappointed, and move on. They're using AI as a lookup tool, not a thinking partner.

The ones who save hours a week treat AI differently. They give it context.

Context is everything

Compare these two prompts:

❌ "Write a project update email"

✅ "Write a project update email to my manager. The project is a website redesign. We're on track for the June deadline but the design phase took longer than expected. Tone should be professional but confident. Keep it under 150 words."

Same AI tool. Completely different output. The second one you can send. The first one you'll spend 10 minutes editing.

Your first action this week

Pick one repetitive writing task you do every week — a status update, a client email, a report section — and build a template prompt for it using the formula above. Test it twice. Adjust. Save the prompt somewhere you'll find it.

That single saved prompt will pay back the 10 minutes it takes to create it every week for the rest of your career.

What's coming

Each week we'll cover one practical use case: meeting notes, research, email, decisions, reports, automation. Always with a real prompt you can copy and use immediately. No theory. Just what works.