Notion Faces, Stripe’s 12 product questions and Microsoft’s Copilot controversy
Plus: 2025’s fastest growing jobs report, Lessons from Spotify on how to measure internal AI tools, brain connected productivity tools.
Hi product people 👋, I hope your 2025 is off to a good start so far.
Coming up this week, we take a look at a controversial new pricing strategy for Microsoft Copilot as well as a less controversial growth strategy from Notion.
Plus, a series of new announcements from Nvidia as it looks to start the year with a bang, a study that shatters many of the assumptions product teams have about how to build an AI bot and a new accessory from CES to turn heads on the return to the office - inspired by a mischievous Japanese spirit.
Enjoy the rest of your week!
More here: Watch on YouTube | Follow on Substack Notes | Upgrade to unlock the full newsletter
Key reads and resources for product teams
As we gear up for a new year and all the unknown unknowns that come with it, here’s some of the most popular posts from the Department of Product in 2024 that you may have missed to get up to speed:
The AI playground 2024: 10 ways to get practical, hands-on experience
How to use Claude Artifacts to build simple product prototypes
(Department of Product)
Google research paper - What are AI agents?
A new whitepaper published by Google scientists explains how agents work, their key components (models, tools, and orchestration), and how they differ from standalone AI models. The paper also provides an overview of different tool types (Extensions, Functions, and Data Stores) and targeted learning techniques that enhance agent performance. Additionally, it offers practical guidance with code examples and real-world applications using Google's Vertex AI platform. Examples include: a travel agent for booking holidays, customer service agents and smart home controls. (Google Research)
Case study - How Spotify uses confidence scores to measure its internal AI tools
A new case study from Spotify's Financial Engineering team shares how they use confidence scores to measure the performance of their internal GenAI applications. (Spotify Engineering Blog)
Process - How to make everyone in your company great at data
Most leaders understand the importance of good data to their operations, too many fail to recognize the critical role that people play in creating it - and even make it harder for people to do the right things when it comes to data. (Harvard Business Review)
UX - A rubric to check your error message quality
NN Group’s scoring rubric offers a systematic approach to assess error message quality across three key dimensions: visibility, communication, and efficiency. Each dimension contains four guidelines that can be scored from excellent (4 points) to poor (1 point), ultimately generating an overall letter grade that reflects the user experience. An 'A' grade indicates a top-tier experience where users can easily recover from errors, while lower grades point to specific areas needing improvement. (NN Group)
Resource - Stripe’s 12 Product Review Questions
When Stripe launches a new product, team members ask themselves these 12 questions. (
‘s Substack)New product features and innovation this week
Notion has released a simple - but creative - new product growth hack that seems to be gaining traction. Faces allows you to create your own Face in the style of Notion drawings and share it with others. The faces aren’t Notion branded but since the company is synonymous with this style, it’s an effective way to indirectly grow awareness of Notion. You can see it in action here.
Nvidia has launched “Agentic AI Blueprints” to automate work in enterprises. Unveiled as part of a major keynote from the company’s CEO this week, the blueprints are a set of ready-to-use templates and tools designed to help developers and businesses create advanced AI agents capable of performing complex tasks autonomously. Blueprints are integrated with NVIDIA’s AI Enterprise software platform and developers can use these blueprints as starting points to build custom AI agents tailored to specific business needs.
Some examples of blueprints include: PDF to Podcast, a blueprint that allows developers to create podcasts from PDF documents, Multimodal PDF Data Extraction designed for enterprise retrieval-augmented generation (RAG) and Document Research Assistant, a blueprint that harnesses NVIDIA NIM microservices and NeMo Retriever to help content creators produce high-quality blogs through automated research, outlining, and content generation.
Meanwhile…
Auth0 has announced a new integration with price management startup Stigg (who we mentioned a few weeks back). It connects Auth0’s identity management with Stigg’s entitlement enforcement capabilities which means you don’t have to create independent entitlement / permissions logic. Super helpful for SaaS companies who are looking to introduce / refactor permissions.
Microsoft is reportedly forcing users to use Copilot in some territories and some users aren’t happy about it:
“It was very keen to be used, and this was irritating to me as a user,”
This is how one user described being forced to use Copilot across Microsoft 365 products according to the WSJ. In Australia and other Southeast Asian countries, Microsoft is adding Copilot to all products - and raising prices as a result - whether users want it or not. This is likely a test that Microsoft is running in one territory to collect users feedback but it still feels like a major strategic blunder from Microsoft which could backfire. Surely if users don’t want to use an AI Assistant, they shouldn’t be forced to do so? In this case the user reportedly decided to ditch Microsoft in the end and switch to Google Docs instead.
Meta is removing third party fact checkers in favor of the Community Notes style moderation model adopted by Twitter / X. The VP of product at X, Keith Coleman, was the original creator of the concept of Community Notes and it was originally called Birdwatch. More on that here.
2025: the year of the AI agent?
If 2025 is indeed the “year of the AI agent”, what exactly are companies building AI agents for? Here’s a summary from a recent WSJ piece:
Tools you can use
Add to Sheets - a Chrome extension that lets you save text, links, image URLs, page links, and more from anywhere on the web to a Google Sheets spreadsheet without switching tabs.
TestSprite - an end-to-end QA product which lets you automate the testing of your entire product.
Project Ambience - get deep work done with no distractions.
📆 Upcoming live programs
If you're looking to level up your technical skills this year, the new 2025 live Web Technologies program kicks off February 1st.
Set over 4 weeks, you'll get the chance to work with an instructor and other participants from all over the world to get a deeper understanding of the technical aspects of product development.
The Web Technologies program is designed to help you become more technically confident - not transition into engineering. Early bird pricing gets you $150 off.
📈 Product data and trends to stay informed
LinkedIn has revealed the fastest growing jobs in 2025. Perhaps unsurprisingly, AI engineers take the top spot.
Nvidia has made over $1bn of investments in startups, as it tries to extend its ecosystem moat around the GPU business.
Bluesky’s growth is slowing down. Following a major bump after the US election, growth has plummeted from 284% to just 10% month on month. It now has just over 26 million users but activity measured by likes has also dropped.
Paid message revenue is one of Meta’s fastest growing segments - and WhatsApp is driving that growth. 90% of Indian consumers now message with a business at least once a week, and 66% report being frustrated with businesses that don’t offer messaging. Report on the rise of WhatsApp.
A new study suggests most users don’t want AI to seem human. Instead, users react more positively to AI which emphasises the humans that were involved in its development. Something to bear in mind if you’re working on your own AI features or bots right now.
Despite its hefty price tag, OpenAI is losing money on its $200 pricing tier. Sam Altman says he personally chose the price point but paid users are simply using it too much. OpenAI ended the year with $3.7 billion in revenue - but it spent $8.7 billion to achieve that, roughly equating to a ~$5 billion loss. A corporate restructuring and price rises are reported to be on their way.
The world has run out of data to train models. According to Elon Musk, this happened last year and the shift will now focus on using synthetic data instead. Gartner estimates that 60% of the data used for AI and analytics projects in 2024 was synthetically generated.
Other product news in brief
👋Uber’s CEO has resigned from self-driving truck startup Aurora
💰Anthropic is looking to raise more money, valuing it at $60 billion - up from $18 billion last year.
🎧Google is launching a “Daily Listen” podcast which is AI generated and personalised based on a user’s preference.
🧠A new startup wants to boost your productivity by connecting to your brain
Paid subscribers get the full DoP Substack including: DoP Deep dive reports to learn from the world’s top tech companies and The Knowledge Series for sharpening their tech skills, AI tutorials for putting AI into practice at work.
I was 100% influenced to make a Notion face! 🥰