Issue 47 | ChatGPT Canvas Changes How Executives Write -- Here Is the Full Playbook

ChatGPT Canvas Changes How Executives Write -- Here Is the Full Playbook

On October 3, ChatGPT launched Canvas, a document-level writing editor that works alongside the chat interface. Instead of copying AI output into a document and editing it separately, Canvas keeps the document open and makes inline changes in response to instructions. Version history means every iteration is preserved.

8 min read

Issue 47

Quick Scan

What matters today

On October 3, ChatGPT launched Canvas, a document-level writing editor that works alongside the chat interface.

Format Weekly newsletter issue

Audience Executives using AI at work

Time 8 min read

Topic ChatGPT Canvas Changes How Executives Write -- Here Is the Full Playbook

04 Prompt to try

Pro Tip

#2 OpenAI DevDay 2024 ships prompt caching

Prompt caching gives a 50% discount on repeated prompt prefixes. Fine-tuning for GPT-4o and vision fine-tuning also launched. Model distillation -- training smaller custom models from GPT-4o outputs -- announced.

Executive angle: Request a caching audit from your engineering team this week. Any AI application with a repeated long system prompt is leaving money on the table.

Read more ->

#4 Prompt caching economics

At scale, prompt caching reduces API costs by up to 50% for applications with repeated long system prompts. For an application processing 1,000 interactions per day with a 2,000-token system prompt, savings run approximately $900/month.

Executive angle: The ROI is immediate and the implementation cost is minimal. This is one of the clearest cost reduction opportunities available in AI budgets right now.

Read more ->

#5 Anthropic Claude instruction-following improvements

Claude models show measurably better performance on complex multi-step instructions and long documents in the latest update. No new model release, but the update is meaningful for enterprise deployments.

Executive angle: If you use Claude for document-heavy workflows or complex multi-step instructions, run a comparison against your existing prompts. Performance on long documents improved meaningfully.

Read more ->

02 / Top AI Updates | 3 Deep Dives

# 01 / 03

ChatGPT Canvas: The AI Writing Studio That Replaces Your Editing Workflow

Canvas opens as a split-screen interface with chat on the left and a persistent document on the right. ChatGPT can see exactly what is in the document, make inline changes, and track every version. Targeted rewrites let you ask the AI to adjust a specific paragraph without affecting the rest. For board memos and investor communications, this eliminates the copy-paste loop that currently defines AI-assisted writing.

Business Impact

Eliminates: The copy-paste loop between ChatGPT and Word that makes AI-assisted editing slow and disconnected

Time back: 60-90 minutes per complex executive document

Best for: Executives producing board memos, investor communications, and strategy documents that require multiple revision passes

Executive Action Steps

Open ChatGPT Plus or Team and click the Canvas icon at the bottom of the chat input.
Provide your source material and ask Canvas to produce a full document in your target format. For a board memo: structure with executive summary, options, recommendation, risks, and next steps.
Read the full draft as a board member would. Flag the three or four sections that need the most work.
Use targeted rewrites on each flagged section by highlighting the text and giving a specific instruction. Each rewrite creates a version checkpoint.

Why it matters: The copy-paste loop between ChatGPT and your document editor creates a fundamental disconnect: the AI cannot see your edits and cannot refine specific sections in context. Canvas resolves this by making the document persistent and the AI document-aware throughout the entire editing process.

Deep dive: ChatGPT Canvas ->

# 02 / 03

Three capabilities shipped at OpenAI DevDay on October 1: prompt caching (50% cost reduction on repeated prompt prefixes), GPT-4o fine-tuning (train GPT-4o on proprietary data for task-specific performance), and model distillation (compress GPT-4o capability into smaller, cheaper custom models). Each changes the economics of running AI at business scale.

Business Impact

Eliminates: The cost inefficiency of sending the same long system prompt with every API call

Time back: Up to $900/month per high-volume application from caching alone, before fine-tuning ROI

Best for: Operations and technology leaders overseeing AI application budgets

Executive Action Steps

Audit all AI applications for caching eligibility: identify any with repeated long system prompts. These are immediate savings opportunities.
Implement cache control flags on static prompt portions in the next engineering sprint. Implementation cost is minimal versus the ongoing savings.
List your top three highest-volume, most-repeated AI tasks. These are fine-tuning candidates for a Q1 project.
Request a cost projection from your engineering team: caching savings and fine-tuning breakeven based on actual task volume.

Why it matters: Most enterprises are overpaying for AI API calls by sending identical context with every request. Prompt caching is the fastest ROI opportunity in AI infrastructure right now. Fine-tuning and distillation are the capabilities that will define enterprise AI cost structures in 2025.

Deep dive: OpenAI DevDay 2024 ->

# 03 / 03

Deep dive

OpenAI DevDay 2024: Prompt Caching and Fine-Tuning for Business AI

The economics and capability ceiling of business AI deployments changed on October 1.

Copy-ready guide

Power prompt

Using ChatGPT Canvas to Refine Board Presentations and Executive Memos

A five-step Canvas workflow that turns a two-hour editing session into 30 minutes.

Copy-ready guide

05 Reuse this

Productivity Gem

#3 Microsoft Copilot Wave 2 continues rollout

Excel now runs Python code via Copilot (describe an analysis in English, Copilot writes and runs it). PowerPoint Narrative Builder creates full decks from a document. Outlook prioritization improves.

Executive angle: If you have a Microsoft 365 Copilot license, try the Narrative Builder on your next board report. Input your Word strategy doc, get a structured slide deck in under 60 seconds.

Read more ->

Two Wave 2 features close the gap between what executives know and what they can produce in Microsoft Office without specialist help. Python in Excel via Copilot lets anyone run sophisticated data analysis by describing it in English. The PowerPoint Narrative Builder creates a full slide deck from a document or brief. Both are available to Microsoft 365 Copilot subscribers now.

Business Impact

Eliminates: The data analyst bottleneck for Excel analysis and the blank-slide bottleneck for presentation creation

Time back: 2-4 hours per complex data analysis or presentation first draft

Best for: Executives who own a question but currently need to delegate the technical execution to get an answer

Executive Action Steps

For Excel: open a spreadsheet with Copilot, describe your analysis in plain English, and let Copilot write and run the Python code in a cloud sandbox.
For PowerPoint: open a blank presentation, click 'Create presentation from file,' and select your source Word document or PDF.
Review the Narrative Builder output as a structural starting point. Edit slide by slide or ask Copilot to restructure the narrative arc.
For both tools: check your Microsoft 365 admin center for Copilot availability if features do not appear in your applications.

Why it matters: The bottleneck for executive data analysis was never the data -- it was the code. Copilot removes that bottleneck by translating natural language into Python that runs immediately. The same pattern applies to presentations: the bottleneck was blank-slide paralysis, not lack of content.

Deep dive: Microsoft Copilot Wave 2 ->

03 / Pro Tip

Using ChatGPT Canvas to Refine Board Presentations

Tools: ChatGPT Plus (Canvas mode) | Time: 30 min | Level: Beginner

The five-step Canvas workflow separates substance revisions from tone polish, uses targeted rewrites on flagged sections, applies a global pass at the end, and uses version history as a safety net. The result: a board-ready document in 30 minutes versus 90-120 minutes with the traditional copy-paste approach.

The Prompt

Why it works: Canvas makes each revision pass discrete and reversible. Targeted rewrites let you improve one section without disrupting the rest of the document. Version history removes the risk of losing a good draft during experimentation. The combination produces faster iteration with less anxiety about breaking what is already working.

Best for: Board memos, investor communications, strategy documents, policy documents -- any written communication that requires multiple revision passes to reach the right level of directness and clarity.

Deep dive: Using ChatGPT Canvas to Refine Board Presentations ->

04 / Productivity Gem

Microsoft Copilot in Outlook: The Inbox Prioritization Workflow

Tools: Microsoft 365 Copilot, Outlook | Saved: 15-20 min/day

Three Copilot features work together to replace 20 minutes of morning inbox triage with a 5-minute review: Prioritize My Inbox surfaces action items, Thread Summarization condenses long threads to one paragraph, and Suggested Replies drafts three response options for every email that needs one.

Open Outlook with the Copilot sidebar active. Provide role context: 'I am the [role] at [company]. My highest-priority areas are [list 3-4].'
Ask Copilot to surface action items from the last 24 hours using the prompt below. Review the resulting list -- this takes 2 minutes versus 20.
For each action item: use Suggested Replies for straightforward responses, delegate with a one-line note for items going to your EA, and use Thread Summarization for any thread needing full context before you respond.

The Prompt

Time savings: Saves 15-20 minutes per day. At 5 days/week, that is 75-100 minutes of recaptured executive time each week.

Deep dive: Microsoft Copilot in Outlook: The Inbox Prioritization Workflow ->

Two major platforms shipped document-level AI editing this fortnight: ChatGPT with Canvas and Microsoft with Copilot's Narrative Builder. The convergence is not coincidental. The productivity gain from having AI edit within a document rather than adjacent to it is large enough that every major platform is now racing to deliver it.

Start with Canvas on your next board memo or executive communication. The version history alone -- knowing you can revert to any prior state -- removes the hesitation that slows down AI-assisted editing. Once you have used it for one document cycle, you will not return to the copy-paste workflow.

Forward this to one executive you respect. They will thank you next week.

Premium Upgrade

Deep dive