🔵 OpenAI and Google reveal new Workspace Agents and Shopify's CEO teases "the start of something really big"

Plus: How Anthropic uses Claude in product engineering, a new email app from ex-Pinterest teams, Adobe continues the trend towards outcome-base pricing

Hi product people 👋,

This week, OpenAI and Google revealed new workplace Agents - with some use cases that could be particularly helpful for product teams.

Plus, a beautiful new email app from former Pinterest employees reimagines how email might work in the future, Shopify’s CEO teases what he calls the “start of something really big” and new rumors and data paints a muddled picture of what’s really going on inside Google’s AI coding adoption efforts.

Happy Friday and have a great weekend!

Rich

Watch on YouTube | Follow on Substack Notes

---

Key reads and resources for product teams

Process - How Anthropic uses Claude in Product Engineering
Software engineer Chuma Kabaghe shows how she uses Claude Code to onboard unfamiliar codebases in minutes, run autonomous testing loops, and manage parallel coding sessions, reducing context switching and shipping faster. (Anthropic)

Tool you can use - A beautiful new Email App from former Pinterest employees
Extra is a new Gmail-only app from ex-Pinterest leaders that replaces traditional inboxes with a “Today” view and smart tabs. it surfaces action items, events, shopping, and news, auto-sorts everything else, and makes bulk unsubscribing and cleanup fast. (Extra)

Knowledge Series - Claude Design Explained
How it works, 5 practical examples and a gallery of 25+ creations from others you can use as inspiration. We’ll test its core workflows hands-on, using Claude’s native design system capabilities to create things like mobile onboarding journeys, landing page prototypes, stakeholder decks and more.

New in the AI Library

• Transition to outcome-based pricing - Use this prompt to design a financial model that forecasts the impact of transitioning to outcome-based pricing

• Identify product feature gaps using your API logs - discover which of your APIs are most popular and translate this into feature gaps you can use to inform your product roadmap

(Department of Product)

Case study - How Microsoft built an AI Assistant that cut sprint planning from 1 hour to 3 minutes
Microsoft’s new AI agents in Azure DevOps are cutting engineering work down to minutes instead of hours. Instead of spending a full sprint planning session manually breaking down work items and writing queries, teams now use AI assistants that handle these repetitive tasks in seconds. (Microsoft)

Resource - Compare every design tool in 2026
Browse and compare over 82 different design tools, all neatly categorized to make it easy to navigate. Includes new entrants like Lovale, v0, Antigravity and Paper as well as traditional classic tools. Handy if you’re looking to augment your product design process with new tools but aren’t sure where to start. (Design Tools fyi)

Interview - Stripe’s head of design on how they reimagined their homepage after 6 years - Stripe’s head of design Katie Dill reveals how they completely reimagined their homepage after six years. She walks through the painstaking design decisions - from wave iterations to bento layouts - and explains how AI is changing her team’s process. And she’s adamant that AI doesn’t replace craft or taste. (YouTube / Y Combinator)

New technologies - Design.MD explained by Google’s David East
Stitch’s DESIGN.md format just went open-source - meaning your design rules can now travel between platforms and projects without starting from scratch. AI agents can understand your brand’s visual language at a deeper level, making smarter choices about color, layout, and accessibility. See how it works in this video, then start building your own DESIGN.md files. (Google)

Strategy - The end of one size fits all software
The software bargain that governed enterprise IT for 30 years is breaking. Generative AI has made custom-built systems fast and cheap to create, meaning companies no longer need to reshape their workflows around standardized tools. Deep Nishar and Nitin Nohria explore how this shifts the real question: not which software to buy, but which workflows you should actually own. (Harvard Business Review)

Opinion - Linear’s CEO on why “Output isn’t design”
New AI tools promise to generate interfaces instantly, but they’re solving the wrong problem, argues Karrr Saarinen. (LinkedIn)

---

New product features and innovation this week

This week, Google held its annual Cloud Next ‘26 event in Las Vegas. During the event, Google unveiled Workspace Intelligence, an AI system that turns scattered Workspace data into unified context for agentic work. Rather than manually jumping between apps to gather information, the new system automatically understands semantic relationships across your Docs, Slides, Gmail, Chat, and other Workspace apps - plus external tools like Asana, Jira, and Salesforce.

Some of the specific features announced that might be of interest to product teams include:

• A new Sheets canvas that lets you create and share custom interactive visualizations (dashboards, heat maps, kanban‑style boards, mini‑apps) directly on top of your data, rather than static charts

• New infographic capabilities that let you embed data visualizations inside Google Docs

• The ability to create automations with custom-built skills - a packaged workflow that can pull from your Workspace data (Gmail, Drive, Docs, etc.) and execute multi-step tasks on a recurring basis. You build them once, and they run whenever triggered.

In theory, Google now has a structural advantage in work-context retrieval that external assistants can only partially replicate through connectors and integrations. Every user who asks Gemini a question that requires knowing what’s in their inbox is a user who can’t ask Claude the same question with the same quality of answer. But this is only the case if it actually works reliably as intended.

You can read more about all 10 Workspace announcements made at Cloud Next ‘26 here.

OpenAI releases its own workspace agent

OpenAI has released workspace agents - AI assistants that handle complex, ongoing tasks for teams. Powered by Codex, these agents run in the cloud and can execute multi-step workflows across connected tools and systems.

Here’s how they work: You describe a recurring team workflow (like lead qualification or weekly reporting), and ChatGPT guides you through building an agent step-by-step. The agent learns what tools to access, what approvals to request, and how to remember context between tasks. It can operate on schedules, respond to requests in Slack, and integrate with dozens of business tools.

Agents take actual actions (writing code, updating CRMs, sending emails) rather than just providing information. For example, OpenAI’s sales team uses an agent that pulls call notes and account data, qualifies leads, and drafts emails - work that previously took reps 5-6 hours weekly.

Here’s an example of a Product Feedback Agent that reads product feedback from the web, groups it into recurring issues, posts a daily summary to Slack, and creates or updates Linear tickets automatically:

The release coincides with the launch of OpenAI’s newest model, GPT 5.5 which scores 58.6% on software engineering benchmark SWE-Bench Pro - significantly below Claude Opus 4.7’s 64.3% but above Gemini’s 54.2%.

Meanwhile…

Perplexity has released Personal Computer - an expansion of its super successful product Perplexity Computer. Unlike that product which runs in the cloud, Personal Computer runs locally on your machine where it can work across local files, native apps, connectors, and the web in one system. You can also speak to it directly and have it carry out actions on your Mac.

Figma has released a new way to view Figma Make creations in the mobile app. For mobile product teams, this makes testing new prototypes on mobile - with swipes, clicks and other mobile native gestures - easier.

Claude has expanded the number of Connectors you can use to over 200+. New connectors include AllTrails, Audible, Booking.com, Instacart, Intuit Credit Karma, Intuit TurboTax, Resy, Spotify, StubHub, Taskrabbit, Thumbtack, TripAdvisor, Uber, Uber Eats, and Viator, with more on the way. Anthropic says they’ve noticed that users connect multiple apps and use them together in a single conversation with Claude.

For example, a product manager will pull a query from Amplitude, turn it into a Canva deck, and drop the link into Asana for the team, all without leaving the conversation.

Shopify CEO shares a vision of the future web; UI generated on demand

Shopify’s CEO shared this as a glimpse of what the future of the web experience might look like, describing it as “the beginning of something really big”:

Flipbook is an infinite visual browser created by Zain Shah, Eddie Jiao, and Drew Carr that generates pages entirely as images on demand. When you click anything in an image, the system generates a new image exploring that topic in deeper detail. There’s no underlying HTML, code, or pre-built links - just AI-generated pixels that represent information retrieved from web searches and the model’s training knowledge.

For product teams, this could lead to entirely new experiences:

• Tutorial systems that visualize concepts differently based on user interactions

• Data exploration tools that generate custom visualizations on-the-fly

• Learning platforms adapting visual complexity based on topic requirements

It’s a bit slow / clunky at the moment but still a fascinating experimental look at what the future might feel like.

---

📈 Product data and trends to stay informed

Former Google engineer Steve Yegge claims Google’s internal AI adoption looks like a tractor company’s, not a frontier lab’s. He says the internal usage distribution looks like this: 20% power users, 60% on basic chat tools, 20% refusing entirely.

According to Yegge and his sources, DeepMind uses Claude daily but the rest of the company gets pushed onto internal Gemini products. Demis Hassabis wasn’t impressed and called the numbers “absolute nonsense” but Yegge has responded this week saying internal Googlers corroborated his original claims. Sundar Pichai also revealed that 75% of the company’s new code is now AI generated and so the true picture of what’s actually happening in Google is muddled.

44% of online buyers surveyed by Bain & Company mostly start their journey in an LLM or split their search between AI tools and traditional search engines. The study also found 50% of online shoppers in the survey trust generative AI for initial research and product comparisons.

Google’s latest ad safety report shows a surge in attempts at placing scam ads using AI.

The trend towards outcome-based product pricing continues. This week, Adobe confirmed that it is planning to introduce new pricing for its CX Enterprise AI suite that charges customers not just for usage, but for results - for example, tying fees to the number of ad campaigns or end-to-end “jobs” its AI agents successfully complete, rather than simply how many tokens they consume.

GitHub has paused new Copilot sign-ups and tightened token limits, saying agentic, long‑running AI coding workflows now cost more in compute than fixed monthly fees cover.

OpenAI’s Codex hit four million weekly active users this week, up from three million two weeks prior. A 33% jump in two weeks. OpenAI had two million weekly users just a month ago.

Music streaming service app Deezer says that AI-generated tracks now represent 44% of all daily uploads.

Occupations with more exposure to AI have grown faster than least-exposed ones in the UK.

---

Paid subscribers get the full DoP Substack including: The Knowledge Series for sharpening your tech / AI skills, the AI Prompt and Skills library and DoP Deep dive reports for in-depth analysis to learn lessons from the world’s top tech companies.