🔵 Gmail enters its Gemini Era and what developers really think about working with AI Agents

Plus: ChatGPT's product roadmap, How PMs can use Antigravity, Copilot adopts the Agentic Commerce Protocol and more…

Hi product people 👋,

Welcome back and Happy New Year!

Coming up in this week’s Briefing, Google picks up where it left off last year by doubling down on its Gemini-everywhere strategy, with Gmail becoming the latest product to roll out a bunch of new AI features. Find out what Blake Barnes, Google’s VP of Gmail, has to say about that.

Plus, OpenAI reveals a new health product, practical ways you can use Claude’s new Chrome extension and a new study reveals what engineers really think about working with AI agents.

As always, if you have any comments about any of the updates this week, just drop a comment below!

Rich

Follow on Substack Notes | Upgrade to unlock paid editions and support the newsletter

---

Key reads and resources for product teams

New from the Department of Product this week:

Deep - The UX of Search explored

The ability to search and find exactly what you need inside a product can play a critical role in engagement and retention; a user who can’t find what they need is ultimately a user who churns. For many leading companies, AI has transformed in-product search from a simple find-and-retrieve function into something more ambitious: visual, conversational, and capable of taking actions. 20+ real-world examples of how search is being reimagined across B2C, SaaS, and ecommerce - with downloadable UI examples you can keep and use as a reference.

New prompt in the library - Apply Second Order Thinking Models to problems you’re facing or decisions that need to be made

This new prompt guides you through a structured decision-making process that goes beyond surface-level thinking by forcing you to consider not just what happens immediately when you take an action, but what happens because of those immediate effects. It’s designed to help PMs slow down enough to ask questions like: what behavior am I incentivizing, what precedent am I setting with my team or stakeholders, what doors am I closing by walking through this one, and what will the state of the product look like six months from now if I make this choice. (Department of Product)

Industry leader perspectives - OpenAI’s CEO of Applications shares her roadmap for the year ahead

OpenAI’s CEO of Applications Fidji Simo outlined her product roadmap 2026 in a new piece published this week. The roadmap mentions the phrase “capability overhang” twice and given that Fidi Simo is the CEO of Applications, it’s also interesting to note that she doesn’t mention the consumer App Store once. Another nod towards this year becoming the year they attempt to double down on enterprise. (Fidji Simo)

Skills - How product managers can use Google Antigravity to build prototypes

Google’s product design lead for Antigravity, Andy Zhang, explains how product managers can use Google’s new Antigravity tool to test ideas, build prototypes and add new features. (X)

Personal development - Test your UX knowledge with the 2025 UX Quiz

This is a short, 10‑question multiple‑choice quiz from Nielsen Norman Group that tests core user‑experience knowledge based on articles they published in 2025. It covers practical UX topics like contextual menus, research methods, AI‑simulated users, smart‑home motivations, workshops, and interface design details (e.g., button styles vs. states). A neat way to get up to speed with UX trends quickly. (NN Group)

Product process - The modern Product Manager in the Age of AI Agents

Google’s PM Shubham Saboo shares his views on the future of the product manager role, arguing that AI coding agents are compressing the traditional PM “translation” role, making the spec effectively the product.

How to prepare for AI in 2026

The NotebookLM co-creator shares her perspective on what to do to prepare for AI in 2026:

---

New product features and announcements

ChatGPT has kicked off 2026 with their first major launch of the year, by launching a new sub-product called “ChatGPT Health”. From the UI shared by OpenAI, Health is a separate section within ChatGPT that you access from the sidebar menu. It has its own conversation history, its own memories, and its own connected apps, all walled off from regular ChatGPT chats. In the launch notes, they mention the “HealthBench” framework, the clinical safety emphasis, and it looks as though this is groundwork for selling ChatGPT to healthcare providers.

OpenAI exec Fidji Simo shared her own experience with using the app for health and OpenAI says that 230 million users are currently asking it about health related questions. But the question for OpenAI is whether users trust them with their health data.

Claude for Chrome extension is now widely available; practical ways you can use it

Claude has officially made its Chrome extension widely available in beta for all paying users. The extension works in a browser’s side panel and can perform tasks in the browser much like an AI-powered browser. Despite a bunch of new browser releases last year, Chrome is still firmly the world’s number one browser and if you’re a Chrome user, here are some practical ways you can use the extension for product-related work:

• Design verification – point Claude at a shipped page and your Figma spec and it systematically checks element positioning, colors, and copy, then reports any discrepancies

• Automated QA flows – describe a user journey in plain language (”sign up, submit invalid email, check error message appears”) and Claude executes it in the browser

• Competitive intelligence – schedule recurring tasks to scrape competitor pricing pages, changelogs, or feature updates and output structured summaries to Notion or Docs

• Bug reporting with technical context – when you spot a bug, Claude captures console errors, network failures, and DOM state that can be super useful to pass onto engineering

• Cross-tool data pulls for reviews – ask Claude to pull metrics from your analytics dashboard, cross-reference with shipped Jira tickets, and draft a summary.

Meanwhile…

Google is starting to roll out a new “AI inbox” which gives users a radical alternative to a traditional inbox. The AI inbox sits above the original inbox in the Gmail navigation; a hierarchical decision that suggests that Google thinks this should take precedence over a traditional inbox. The UI consists of suggested to do items from the day and summaries from your most important emails. At the same time, Google has confirmed that it is making other AI features like suggested replies, AI overviews and Help me Write tools all free to users after previously only being available as paid features.

Google’s VP of Product for Gmail, posted a video on LinkedIn outlining the new features and says that Gmail is now in its “Gemini Era”. The AI Inbox certainly feels like a potentially helpful feature - but what happens if the daily overview misses something important?

Copilot is the latest product to adopt the Agentic Commerce Protocol with a new feature called Copilot Checkout - an embedded ecommerce experience that turns conversational searches into purchases within a single conversation. A shopper asks Copilot for recommendations, compares options, and asks follow‑up questions. When intent is clear, Copilot surfaces the right checkout moment and processes the purchase using a merchant’s existing commerce stack. In a post by Nayna Sheth, Microsoft’s Head of Product for Agentic Payments, she says that journeys that include Copilot led to 53% more purchases within 30 minutes compared to those without.

However, rolling out ecommerce features inside conversational products isn’t easy; OpenAI is reported to be struggling with handling structured data, catalogue details and pricing information which has led to some customers being charged the wrong amounts. As with most things product-related, the devil is often in the detail.

Spotify adds more social features, boosting network effects

Spotify unveiled a new social listening feature to its mobile app. Users will see what people are listening to in your sidebar alongside their chats. And, if something sparks your interest, users can quickly tap the song to add it to their library, or react with an emoji.

The good news for people with bad taste in music is that the feature is switched off by default, but it marks another step in the direction towards the company becoming a social media platform (in many ways it already is one). After hiking its prices up recently, the company is likely betting that a deeper social graph will boost its network effects and reduce churn in the long run.

---

Tools you can use

Async - a buzzy new tool that calls itself an AI product manager that lives in Slack. Described by a Stripe PM as something that “feels like your missing, infinite teammate.” Currently in beta but worth a look.

Capacity - introduces what it calls “spec coding” where you define your specifications first, then led AI build production-ready code.

Scouts - AI agents that will monitor specific topics for you. These could be pretty handy for things like competitor monitoring or market research.

---

📈 Product data and trends to stay informed

A new study shows that engineers enjoy working with AI Agents for coding. In this study, “enjoyment” was measured on a 6-point scale comparing coding with agents versus without. The mean score was ~5.11, indicating that developers generally felt “pleased” to “extremely pleased.”

But, there was one important caveat: working with AI agents was enjoyable so long as engineers remained in control and were able to thoroughly review its outputs. Read the full study here.

The percentage of Americans checking LinkedIn more than once daily rose to 4.7% last year from 3.9% in 2020. This week, the company published its “Jobs on the Rise” report and perhaps unsurprisingly, AI engineers took the top spots with AI consultants and strategists in second. Most of the people in that role transitioned from product management.

McKinsey expects to have as many “personalized” AI agents as employees by the end of 2026.

The unintended economic consequences of AI agents are starting to emerge. CSS framework Tailwind has laid off 75% of its engineering team because it became too popular with AI Agents, destroying its primary upsell channel where users would visit their docs to find out more and upgrade to a paid package. Product teams with similar upsell journeys may be vulnerable to similar AI agent disruption.

ChatGPT has lost a further ~4% in AI tool traffic share while Google has increased its share to over 20%. These stats are for web only but they put OpenAI’s pre-Xmas “Code Red” into context.

AI Agents can now outperform humans on PEN testing. Something to consider for your next security audit.

---

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