Deep: New AI Agent Product Features Explored
How the world’s top product teams are integrating AI agents into their product feature set. Inspiration from Adobe, Microsoft, Asana, Google, Instacart and more...
🔒DoP Deep goes deeper into the concepts and ideas that are covered in the Weekly Briefing to help you learn lessons from the experiences of top tech companies. If you’d like to upgrade to receive these in-depth pieces of analysis you can upgrade below. New reports added every month.
2025 was dubbed “the year of the AI Agent” and as the year draws to a close, the hype train is still in full motion, albeit with a hefty dose of reality in some cases. The real world impact of agents is still pretty mixed but new research from McKinsey shows that 33% of high performing tech companies are now moving into the scaling phase of their AI Agent efforts.
The release of Claude’s new Opus 4.5 model which beats other models on agentic benchmarks looks set to accelerate the rollout of new agentic features and industry backers now include Jeff Bezos as just last week, it was confirmed that the Amazon’s founder’s venture fund Project Prometheus had acquired an agentic computer firm General Agents.
In this DoP Deep dive, we won’t spend too much time exploring the ambitions of Project Prometheus but we will take a look at what AI agent features some of the world’s leading product teams are actually shipping.
This analysis includes over 20 different examples of new AI agent features together with an in-depth look at what the most up to date research says on AI agent capabilities. If you’re currently thinking about your AI roadmap plans for next year and would like some inspiration based on what other product teams are shipping or would like to get up to date with AI agent capabilities quickly, this Deep dive should help.
Coming up in this DoP Deep:
20+ real world examples of AI agents as product features
How to think about delegation as a core UX pattern: the mental model shift from “users execute tasks” to “users decide what to delegate”
The emerging pattern behind how companies like Ramp and legal tech startup Legora are packaging AI agents as “domain experts” - and why this positioning matters
Microsoft’s vision for an “agentic OS” and what a “headless” future might mean for product teams
The t2-bench test explained: how the industry actually measures whether an AI agent can handle real world scenarios
20+ AI agent features broken down by capability, industry, and how they actually work - with frameworks you can steal for your own product strategy
How this analysis is structured
This analysis includes over 20 different examples of AI Agent product features across various different industries:
Company - the analysis includes examples of new AI Agent features released by product teams from various different companies including Google, Superhuman, Ramp, Monday, Adobe and others.
Feature - typically (but not always), companies decide to name their AI agent features. For example, Asana calls their AI agents Team Mates. Where applicable, each name is included.
Capabilities - each agent comes with different capabilities. AI agent capabilities are broken down into distinct categories (more on that below).
How it works - a detailed breakdown of how the AI agent feature works. The core value proposition of the AI agent is included along with any relevant examples of new technologies. We’ll also discuss the features in more detail in the analysis to get a deeper understanding of how the AI agent feature works - with key takeaways for product teams who might be developing their own AI agent strategy.
Industry - the examples include products from various industries including SaaS, ecommerce, Legal tech and more.
Link to more - link to more information to learn more about the AI Agent feature in more detail.
Agent capabilities explained
Each AI Agent feature is categorized and tagged according to its capabilities. There are over 13 different capabilities including:
Document processing - the ability for AI agents to read, extract, analyze, and manipulate documents across various formats to perform tasks like summarization, data extraction, or content generation.
Voice -the capability for AI agents to conduct spoken interactions, including making phone calls, processing voice commands, and delivering audio responses to users.
Connectors - integrations that enable AI agents to access and interact with external systems, databases, and third-party tools (like Gmail, Slack, CRMs) to retrieve data and execute actions.
Data Analysis - the capacity for AI agents to query, interpret, and derive insights from structured or unstructured data sources, often through conversational interfaces that democratize access to analytics.
Purchasing - the ability for AI agents to autonomously or semi-autonomously complete transactions, including checking inventory, comparing prices, and executing purchases on behalf of users.
Background execution - AI agents with the ability to run tasks asynchronously in the background, continuing work while users focus on other activities and notifying them when complete.
Project management - functionality enabling AI agents to assist with planning, tracking, and coordinating work by generating project structures, updating status, managing dependencies, and facilitating team collaboration.
Coding - the ability for AI agents to write, debug, refactor, and explain code, often with specialized models and interfaces designed for software development workflows.
Video / Image generation - AI agents that create, edit, or enhance visual content including images and videos, typically by analyzing trends, generating avatars, or producing marketing materials.
A closer look at the AI agent product features in more detail - and the implications for product teams
Now let’s take a look at some of the new AI Agent product features in more detail. We’ll explore how the features work as well as what the practical implications of these emerging new AI agent features are for product teams - with some frameworks and mental models to consider for your own product.