ChatGPT Generates Images Inline, Copilot Finance Agents Hit Excel, and Claude Cowork Lands on Windows

#2 Microsoft Finance Agents go GA in Excel

Variance analysis, forecast scenarios, and budget rollups now run inside Excel via Copilot Chat against live workbook data with enterprise data protection. A controller can brief the CFO in 90 seconds instead of 40 minutes.

Executive angle: Cut close-week pivot table work in half -- enable the Finance Agent toggle in the M365 admin center this week.

Read more ->

#3 Google Gemini drafts in Docs, Sheets, and Slides

A native side panel writes prose, fills formulas, and assembles decks without copy-paste. Available across the Google Workspace suite for eligible accounts.

Executive angle: Stop tab-switching during weekly deliverable production -- use the Gemini side panel to draft and fill in context without leaving your document.

Read more ->

#4 Perplexity Personal Computer hits GA on Mac

An always-on background agent with file and app access on macOS. Queue tasks before you log off and review results at 5 pm.

Executive angle: Queue tasks before you log off and review at 5 pm -- test the inbox and research triage use cases first.

Read more ->

#5 xAI rolls out Grok Voice with extended Think Fast

Real-time voice plus deeper reasoning is now available on Plus and Premium tiers. Think Fast enables faster multi-step reasoning in voice sessions.

Executive angle: Solve a strategy question on the commute and get clean notes back -- try Grok Voice on your next back-to-back travel day.

Read more ->

02 / Top AI Updates | 3 Deep Dives

# 01 / 03

ChatGPT Image Studio: DALL-E 4 Inline Generation and Editing

OpenAI shipped inline image generation and editing inside ChatGPT, powered by DALL-E 4. Plus and Pro users mask elements with natural language, iterate on the same subject, and export at 4K. The round-trip from chat to a separate design tool disappears for first-pass marketing visuals -- saving 4 to 6 hours per week on design iteration across the team.

Business Impact

Eliminates: The contractor wait and Figma round-trip for first-pass marketing visuals -- hero images, carousel cards, and product tiles now generate in 2 to 3 minutes.

Time back: 4 to 6 hours per week on design iteration, per team member who previously used Midjourney or Photoshop for first-pass assets.

Best for: Marketing and product leads who need fast asset iteration for campaigns, social carousels, and launch tiles without a design contractor in the loop.

Executive Action Steps

• Open ChatGPT Plus or Pro and type a detailed image prompt: specify context, composition, color palette, mood, and technical spec (e.g., 4K, no text).
• Review the first output and use natural-language masking to edit specific elements -- change a color, adjust a pose, or modify the background.
• Iterate once or twice to reach acceptable output. Aim for 2 to 3 minutes total from first prompt to final version.
• Request a 4K export, download the file, and share with your team or drop it directly into the campaign asset folder.

Why it matters: DALL-E 4 interprets multi-part prompts in a single pass, handling context, composition, palette, and technical spec together. Subject consistency across iterations means you can generate the same person or product across a LinkedIn ad, website banner, and case study cover without regenerating from scratch. For decades, design capacity was the gating factor in marketing velocity. That constraint is now optional.

Deep dive: ChatGPT Image Studio ->

# 02 / 03

Anthropic brought Claude Cowork to general availability on Windows on April 23, 2026, alongside Enterprise Role-Based Access Controls, per-team data scoping, and audit logs. Until this release, desktop automation was a macOS-only story -- meaning most Fortune 500 operations teams were locked out. The Windows release, combined with Enterprise RBAC, makes Cowork the first agent platform suitable for SOC 2 and HIPAA-regulated teams.

Business Impact

Eliminates: The manual queue for email triage, calendar management, and research aggregation -- all three workflows now run async while the executive focuses on higher-priority work.

Time back: 8 to 14 hours per executive per month across inbox triage (4.5 hours/week), calendar prep (3 to 4 hours/week), and weekly research scan (6 to 8 hours/week).

Best for: COOs, Chiefs of Staff, and operations leads at Windows-based organizations who manage high-volume inboxes, complex calendars, and recurring research tasks.

Executive Action Steps

• Download Claude Desktop on Windows (or macOS) and sign in with a Pro, Max, or Enterprise account. Verify the SHA-256 checksum from release notes.
• Open Cowork and grant access to one app -- start with the browser or email client to keep the first task scoped and reviewable.
• Write your first task in natural language: 'Read all email flagged URGENT or from board members, summarize in Slack format, tag for my review before any reply is sent.'
• Set an execution schedule (e.g., daily at 7 AM), run one manual execution to test the output, then enable audit logging in settings before expanding to additional tasks or team members.

Why it matters: Cowork's persistent memory across sessions eliminates the context loss that kills traditional automation scripts -- the agent remembers prior weeks and can spot escalation patterns across time. Enterprise RBAC and audit logs mean every action is tracked, compliant, and reversible. Compared to Copilot Agents (cloud-native, slow to set up) and ChatGPT Tasks (web interface, no audit logs), Cowork combines desktop speed with enterprise controls and a reasoning model optimized for multi-step tasks.

Deep dive: Claude Cowork Windows GA ->

# 03 / 03

Microsoft moved its Finance Agents to general availability inside Excel on April 22, 2026. Copilot Chat now runs variance analysis, forecast adjustments, and budget rollups against live workbook data with enterprise data protection. A controller can ask 'compare actual to budget by department, flag variances over 5 percent' and receive a ranked variance report with cited rows in 3 seconds -- replacing a 40-minute manual pivot workflow.

Business Impact

Eliminates: The close-week pivot table rebuild cycle -- no more extracting from SAP, cross-referencing the GL, and manually refreshing formulas. Finance Agents answer variance questions in natural language directly inside the workbook.

Time back: Close-week effort reduces by 40 percent. Teams delivering the same package in 6 hours instead of 40, freeing Thursday night and shifting Monday from rework to analysis.

Best for: CFOs, Controllers, and FP&A leads at M365 Copilot-licensed organizations running monthly close cycles, scenario modeling, or budget rollups across departments.

Executive Action Steps

• Confirm your M365 Copilot license is active. Open the M365 admin center, navigate to Settings - Org Settings - M365 Copilot, and enable the Finance Agents toggle under Excel and Finance Features. Allow up to 4 hours for tenant propagation.
• Open the budget workbook in Excel and click Copilot in the ribbon, then choose Finance Agent from the options panel.
• Type your variance query in plain language: 'Compare actual to budget by department; flag variances over 5 percent.' The agent scans the workbook and returns a ranked table with department, line item, budgeted amount, actual, and variance percentage.
• Follow up with scenario prompts such as 'If Q4 ad spend increases by 15 percent and we hold headcount flat, what happens to operating margin?' to run forecast adjustments without rebuilding formulas.

Why it matters: Finance Agents treat Excel as a first-class AI workspace rather than a calculator. Analysts ask questions in English, the system answers in seconds, and stakeholders act on insights the same day they request them. The close cycle compresses because variance hunting becomes a conversation instead of a weekend project. For executives managing remote FP&A teams or lean finance departments, this is a structural multiplier -- before deploying broadly, audit source data first to ensure GL reconciliation and remove duplicate transactions.

Deep dive: Copilot Finance Agents ->

03 / Pro Tip

The ChatGPT Project Stand-Up That Writes Itself

Tools: ChatGPT Projects | Time: 20 min | Level: Beginner

ChatGPT Teams Projects now connect to Linear and Notion live. Every Monday morning, a product lead previously spent 45 minutes pulling status from Linear, scanning Notion notes, and checking Slack before drafting a stand-up half from memory. With live connectors, a single prompt replaces all of that. Setup takes 20 minutes once; the stand-up generates in 2 to 3 minutes every Monday after that, saving 30 minutes per workstream per week.

The Prompt

Why it works: Before April 2026, ChatGPT Projects used snapshots -- a single file upload that went stale. Live connectors now pull current data every time you run the prompt. The Project remembers context from previous stand-ups, so it can spot escalation patterns across weeks. ChatGPT does the cross-tool reading so the lead does not have to context-switch between Linear, Notion, and Slack before writing a single word.

Best for: Product leads, engineering managers, and account team leads running recurring workstreams with tickets in Linear and notes in Notion -- especially those managing three or more sprints or launches simultaneously.

Deep dive: The ChatGPT Project Stand-Up That Writes Itself ->

04 / Productivity Gem

The Claude Cowork Friday Recap: A One-Page Weekly Summary While You Sign Off

Tools: Claude Cowork | Saved: 2 hrs/week

Manually stitching email threads, calendar summaries, and project-tool activity into a coherent weekly narrative takes 60 to 90 minutes on Friday -- mental energy executives do not have at end-of-week. Set up Claude Cowork once in 20 minutes, and a one-page recap lands in your inbox before 5 pm every Friday, replacing Sunday-night anxiety with clarity about what shipped, what is open, and what decisions drive next week.

• Launch Claude Desktop, select 'Create New Session' from the sidebar, and name it 'Weekly Recap Agent' so it persists across weeks. Grant access to email, calendar, and your project tool of choice (Linear, Notion, or Asana).
• Paste the reusable prompt into the task definition, set delivery to 'Email to [your address]', set the schedule to Friday at 4 pm, save, and confirm. Your first recap arrives the following Friday.

The Prompt

Time savings: 60 to 90 minutes of manual Friday review per week, plus reduction in Sunday-night rumination. Executives who review a structured weekly recap on Friday do not enter the weekend looping on what they accomplished -- the decision thread is clear, the wins are visible, and the open items are explicit.

Deep dive: The Claude Cowork Friday Recap: A One-Page Weekly Summary While You Sign Off ->

05 / Health Tip