Google and Notion unveil major new AI Agents for product teams
Plus: Figma's new edit powers, How AI is reshaping product orgs, Anthropic Economic Index Report...
This week’s briefing is supported by Enterpret - the new way to transform your user feedback into crisp, actionable customer intelligence. Used by product teams at Perplexity, Notion, Figma and more.
PMs can use Enterpret to automatically transform user feedback into Slack-ready narratives with charts that communicate customer feedback without meetings.
Hi product people 👋,
Coming up this week, a major new release from Notion that helps product teams build their own custom AI agents, new analysis from Anthropic and OpenAI reveals how Claude and ChatGPT is used in the real world and a cautionary tale from the engineers that are being hired to clean up the mess caused by vibe coders.
Happy Friday and have a great weekend!
Rich
Watch on YouTube | Follow on Substack Notes
New MCP connectors and Agents from Perplexity and Notion’s AI agents - the “biggest in history”?
Perplexity has launched a new set of MCP connectors which allows you to connect to third party apps like Calendar, Notion, Linear and GitHub. Some of the use cases it suggests you might want to include getting a summary of your upcoming meetings and converting documents into Linear projects with prioritized tasks and subtasks. These issues can then be automatically assigned to the relevant product teams members depending on their role.
Notion has launched a series of new AI agents as part of its major “Notion 3.0” release. The agents include personal agents that are focused on one task or a whole team of agents who work together to achieve an outcome together. A Notion exec described the launch as the company’s “biggest in history” and for product teams, these custom agents could transform many of the common, everyday workflows. Some of the suggested agents include a “bug basher” agent and an agent that helps you prioritize your product roadmap. Here’s an example of a bug basher in action.
Google releases a major new payments protocol - could it drive stablecoin adoption further?
Google has launched a new protocol for AI agent purchases. It’s called the Agent Payments protocol (AP2) and it’s designed to be interoperable between different platforms. For product teams, this could be an important addition to the payment UX stack. You can read the full AP2 spec on the official GitHub project page. But here’s how it works: In a typical transaction, the agent builds a cart, the merchant signs it, the user approves and signs the mandate, and the transaction is executed with a clear audit trail for accountability.
Coinbase is one of the early adopters of the protocol and they’ve built a proof of concept that you can watch using stablecoins here. This week, Visa confirmed that stablecoin use is up 4x on year, suggesting that stablecoin adoption could be the next big shift in payment methods.
Figma’s prompt to edit
Figma has unveiled a new feature called “prompt to edit”. As you might expect, it lets designers use a conversational prompt to make edits in design files. If it does indeed work as well as the demo suggests then this could potentially make editing files much easier for product teams, lowering the barrier to entry. And for designers, this may have some unintended consequences where, as with vibe coding, folks from other disciplines take it upon themselves to make quick edits without running it by the person who is typically responsible for design changes.
In other news…
YouTube has launched a series of new AI features including the ability for creators to generate Shorts using Google’s Veo 3 models. AI generated videos will be labelled but of course, the arrival of Veo 3 does risk the sloppification of video content. Other announcements include a brand new AI assistant called “Ask Studio”, targeted to creators which allows them to ask questions about the performance of their content.
Key reads and resources for product teams
New from the Department of Product
Knowledge Series - Agentic Workflows Explained: an in-depth but simple guide for Product Teams
Google’s Gemini product leader recently announced a list of what he believes are the essential skills for AI product people in 2025. Unlike traditional software, he argues that LLMs present infinite use cases and infinite failure modes. As a result, it takes a new set of skills to balance the versatility of LLMs with the focus required to build high-quality products.one of the core skills he mentions is “an understanding of agentic flows”. And that’s exactly what this Knowledge Series is all about.
New prompts added to the AI Prompt and Mini App Library
The DoP AI prompts library includes a series of AI prompts you can use at work to develop your skills. Here’s some new prompts added this week:
Work alongside a panel of mental models expert to help you to tackle specific problems using each of the different frameworks
Ask your future self how to solve a specific problem - useful for pre-mortems and product strategy / discovery work
Productivity - The AI-Native Office suite
The next generation of AI-native office tools compiled by VCs at Andreessen Horowitz. (A16Z)
UX - The Anatomy of Good Icons
Consistent stroke thickness is crucial and ensures icons are visible and usable in different contexts. Corners, whether exterior or interior, should be carefully shaped—small changes in their roundness or sharpness can greatly affect how easily an icon is recognized. In this guide, NNGroup explains how to design good icons. (NNGroup)
Report - How AI is reshaping product orgs
Miro surveyed over 2000 product teams on how AI is reshaping their business. Here are the results (Miro)
Strategy - 9 ways growth is different in AI products
Traditional growth levers like activation are less relevant; in AI, the core product experience (often a single prompt box) is the main driver of user activation and value. (Elena Verna)
Tools you can use
Enterpret - Turn user feedback chaos into actionable customer intelligence. Product teams at companies like Canva, Notion and Perplexity use Enterpret to manage spikes, stack-rank work, and track sentiment after launches.*
AIApply - lets you use AI to apply for jobs on your behalf.
Oboe - a new AI learning app that you can use to teach yourself pretty much anything. Created by the founder of Anchor, previously sold to Spotify.
Whisper - a desktop AI that sees your screen and delivers what you need proactively.
📈 Product data and trends to stay informed
Anthropic has published its latest Economic Index report which explores Claude AI usage patterns and their societal impact. It’s a fascinating read if you’re interested in understanding how users are interacting with AI right now.
Here’s some takeaways from the report:
In the US, 40% of employees now use AI at work, doubling in two years. This is much faster than previous technologies like electricity, PCs, or the internet.
Businesses use Claude mainly for coding and administrative tasks, with 77% of API usage showing automation. Coding still dominates with 36% of all tasks but education rose from 9.3% to 12.4%.
The share of directive conversations has increased from 27% to 39%. In this context, directive refers to situations where users give Claude (the AI assistant) a clear, specific instruction or request, and Claude completes the task independently, without needing much further input, clarification, or collaboration from the user. In other words, more people are using Claude to automate - rather than collaborate - than ever before.
OpenAI published its own equivalent report on how people use ChatGPT. And what’s most striking is that just 4% of users of ChatGPT are using it for coding tasks.
Gemini has officially overtaken ChatGPT in the mobile app store but its Search dominance may be starting to slip. New data shows that the DAU / MAU ratio (the percentage of monthly active users who are active on a daily basis) is falling for Google, but rising for ChatGPT. What this data doesn’t show, though, is Gemini usage - and since the Gemini app is performing exceptionally well, this suggests that plenty of Google’s users are substituting traditional Search with Gemini and AI mode.
Almost 70% of top executives believe that the introduction of AI agents requires new management approaches. A new piece in the Wall Street Journal reflects this too with one COO saying “we’re going to have to rewire the whole company”. One of the emerging new trends is the increase in the sizes of teams with flatter org charts that include a mix of business people alongside traditional product development roles like engineering, product management and design.
Thanks to the rise of vibe coding, a new type of engineer is now in demand: the vibe coding cleanup specialists, hired to clean up the mess made by vibe coders. “Most of these vibe coders, either they are product managers or they are sales guys” said one engineers.
Paid subscribers get the full DoP Substack including: The Knowledge Series for sharpening your tech skills, AI tutorials for putting AI into practice at work and DoP Deep dive reports to learn from the world’s top tech companies.