The Week AI Moved Into the Work Itself
GPT-5.5 became better at decisions, Claude joined Slack, Grok moved into Word, and Perplexity made research branch like a workflow.
What matters today
GPT-5.5 became better at decisions, Claude joined Slack, Grok moved into Word, and Perplexity made research branch like a workflow.
Key points
- GPT-5.5 Instant improves decision work
- Claude Tag makes Slack delegation visible
- Grok moved into Microsoft Word
- Turn Perplexity into branching research
- Run the GPT-5.5 decision memo test
Welcome byte
This week was not one giant model launch. It was something more useful: AI moved into the places work already happens. GPT-5.5 Instant became better at decisions, advice, planning, research, and vendor-style comparisons inside the default ChatGPT flow. Anthropic launched Claude Tag so teams can bring Claude into Slack channels and delegate work in the open. xAI shipped Grok for Word, which means the draft, research, rewrite, and formatting loop can happen inside the document instead of across three tabs. The Pro Tip gives you the decision memo prompt to test GPT-5.5 today. The Productivity Gem turns Perplexity Computer into a branching research desk. Below those: a health-data privacy audit and a no-code image classifier project for kids.
Quick Hits
ChatGPT gets better app controls and project organization.
OpenAI's June 18 app updates add clearer app permission controls, project pinning, sidebar organization, easier sharing, note creation from responses, faster photo uploads, and Android one-off model selection for paid plans. Read more
Long ChatGPT pastes now become attachments.
On June 22, OpenAI expanded long-paste handling so text over 10,000 characters automatically becomes an attachment for Free and Go users, with Plus, Pro, and Business already covered. Read more
Codex Record & Replay turns workflows into skills.
ChatGPT Business users on eligible macOS setups can demonstrate a workflow once and turn it into a reusable Codex skill using Computer Use. Executive Angle: Record one repeatable publishing, reporting, or admin workflow before building another checklist. Read more
Grok links to Interactive Brokers.
xAI now lets Grok users connect Interactive Brokers accounts for portfolio analysis, scenario modeling, market research, and order-instruction generation. Read more
Google ships a short AI bootcamp for marketing teams.
AI Boost Bootcamp 201 for Marketing is a role-based training path under 45 minutes, covering NotebookLM, Gemini Deep Research, Canvas, Google Vids, Workspace Studio, `=AI()` in Sheets, and Gemini Enterprise agents. Read more
Top Updates
GPT-5.5 Instant improves decision work
OpenAI updated GPT-5.5 Instant on June 24 to improve conversational quality, especially for decisions, advice, planning, research, and shopping. That matters because Instant is where many executives do the everyday work: compare vendors, choose between tradeoffs, pressure-test a plan, or ask what to do next. The update is not a new button. It is a better default surface for judgment work.
Read the full analysis GPT-5.5 Instant Is Now a Better Decision Partner
Claude Tag makes Slack delegation visible
Anthropic launched Claude Tag on June 23 in beta for Claude Team and Enterprise customers. Teams can grant Claude access to selected Slack channels, tools, data, and codebases, then tag `@Claude` inside a channel to delegate work. The important change is shared context: the team can see what Claude is doing, pick up the thread, and review the output together.
Read the full analysis Claude Tag Turns Slack Into an AI Delegation Surface
Grok moved into Microsoft Word
xAI launched a free Grok add-in for Microsoft 365 on June 18. Grok can transform rough notes, research across the web and X, rewrite, fix grammar, change tone, and format inside Word. For anyone who lives in proposals, board updates, policy drafts, or client memos, the main value is removing the copy-paste loop between chat and the document.
Read the full analysis Grok for Word Removes the Copy-Paste Memo Loop
Pro Tip
Run the GPT-5.5 decision memo test
Tool: ChatGPT | Time: 12 minutes | Level: Beginner
GPT-5.5 Instant is specifically improved for decisions, advice, planning, and research. Test that claim on a real decision, not a toy prompt. The key is to force the model to define options, expose tradeoffs, and give one recommendation.
The Prompt
I need to make a business decision. Decision: [STATE DECISION]. Context: [PASTE FACTS]. Constraints: [BUDGET, TIME, PEOPLE, RISK, CUSTOMER IMPACT]. Success criteria: [WHAT WINNING MEANS]. Options I am considering: [OPTION A, OPTION B, OPTION C]. Analyze the options in a table. Then identify hidden assumptions, second-order risks, and what information would change the recommendation. End with one forced recommendation, a 3-sentence rationale, and the first action to take in the next 24 hours.
Why it works: the prompt turns a vague advice request into a bounded decision memo with constraints, tradeoffs, assumptions, and an immediate next action.
The GPT-5.5 Decision Memo Prompt
A 12-minute prompt that turns ChatGPT's updated default model into a sharper decision partner.
Copy-ready guideProductivity Gem
Turn Perplexity into branching research
Tool: Perplexity Computer | Setup: 15 minutes | Saved: 2 to 4 hours per research cycle
Perplexity's June 19 update added Deep Research inside Computer, a command panel, forking, inline actions, and enterprise controls. The workflow: run one broad research question, fork the best branches, and turn each branch into a different deliverable.
- Start with a market, account, or competitor question.
- Use Deep Research to create the base brief.
- Fork one branch for risks, one for opportunities, and one for next actions.
- Convert the strongest branch into a report, spreadsheet, deck, or dashboard.
- Save the structure as your recurring weekly research template.
Perplexity Computer Branching Research Workflow
Turn one Deep Research session into a competitor brief, account brief, risk memo, and action plan.
Put it to workHealth Tip
Audit health-data sharing before AI
Tools: iPhone Health + ChatGPT Data Controls | Time: 10 minutes | Result: cleaner privacy hygiene
The goal is not to avoid AI. The goal is to avoid pasting sensitive health details into the wrong place with the wrong settings. Apple lets you review which apps and people can read or write Health data. OpenAI lets ChatGPT users turn off model training and use Temporary Chats for conversations that should not create history or memory.
- In iPhone Health, open Sharing and review people, providers, apps, and devices with access.
- Remove anything stale.
- In ChatGPT settings, open Data Controls and confirm whether model training is on or off for your account.
- Use a Temporary Chat for sensitive wellness summaries.
- Strip names, exact dates, and unnecessary identifiers before pasting.
This is not medical advice. Do not use AI output as diagnosis or treatment guidance. Consult a qualified healthcare professional before making health decisions.
The Health Data Sharing Audit Before AI
Before asking AI to summarize wellness data, check who can access it and what your AI settings do with it.
Step-by-step guideKids Tip
Build a Teachable Machine classifier
Tool: Google Teachable Machine | Ages: 8 to 16 | Time: 45 to 75 minutes
Teachable Machine lets kids train a computer to recognize images, sounds, or poses without writing code. The lesson is not that AI is magic. The lesson is that examples, labels, lighting, backgrounds, and test data shape what a model learns.
- Pick three safe household categories, such as mugs, books, and toy cars.
- Collect 20 to 30 image examples for each class.
- Train the model.
- Test it with new objects and different lighting.
- Keep a scorecard of when the model gets confused and why.
Deep dive: Build a Family Image Classifier With Teachable Machine
Build a Family Image Classifier With Google Teachable Machine
A no-code weekend project that teaches examples, labels, testing, and model mistakes.
Try the activityWrap Up
That is the June 25 brief.
The week's dominant theme is AI leaving the separate chat window and entering the work surface. ChatGPT is better at decisions. Claude can be tagged in Slack. Grok is inside Word. Perplexity can branch research into deliverables. The practical question is not which launch sounds biggest. It is which surface already contains your work.
Pick one surface this week: ChatGPT for a decision, Slack for a stalled thread, Word for a messy memo, or Perplexity for a recurring research loop. Give it one real task before next Thursday.
Pierre Bradshaw Founder, PromptHacker.ai
About the author
Pierre has spent 25+ years building growth systems across fintech, real estate, lending, campaigns, and AI workflows, with $1.5B+ in client value delivered.
If you have any questions or comments about The Week AI Moved Into the Work Itself feel free to reach out. I'd love to hear from you.
Contact PierreThree deep dives. Four useful moves. One email worth opening.
PromptHacker turns the AI firehose into practical next steps for work, health, family, and everything time keeps trying to steal.