The industry was switching to agents, so our original design to help customers feel comfortable with AI generated content became obsolete quickly.

Customer-First Designs

Teams used analytic and CSAT tools to evaluated existing customer behavior and sentiment on the features we had recently shipped to GA. They used industry trends and inspiration from peers to determine what AI tools, skills, and agents may appeal to users.

The design lead created an AI prototype of the skills to get some quick usability and sentiment analysis. While her conversations showed that people would prefer more control, and only have topics and outlines generated, she saw the significant shift in AI comfort from her last user interviews, and knew that producing full content would be a better long-term strategy.

Platform Alignment

Many teams at HubSpot were building agents, including a team dedicated to creating an Agent Studio where all agents could be enabled and set up. The Agent Studio team was nervous about building agents with in-context experiences (eg a Blog AI Agent that could be discovered and monitored from the Blog tool), because they did not yet have the time to think about that experience.

I encouraged my team to continue to design an in-context experience and to align among other teams who would benefit from the same. By doing this, we increased the probability of success through user discovery, and also decreased the amount of work put on the Agent Studio team to design new patterns.

Ultimately, we were able to ideate, push our designs further, and align on a core pattern, approved the by agent studio team for consistency.

Impact

  • Released 3 agents within 3 months

  • Design match customers’ mental models and needs that are quickly evolving

  • Helped iterate on generating high quality AI outputs

  • Helping HubSpot maintain its status as Marketing leaders

  • Apply insights across all of Content Hub to ensure quick iteration and integration of research findings

While the concept test of the AI prototype showed that people wanted to still have an “AI Partner” to brainstorm and create, the designer heard a big enough change in sentiment to know that a full content creator was a better long-term bet.

Building AI Agents

Context

HubSpot Content Hub product targets Marketers who creates website pages, landing pages, videos, podcasts and other web content to generate leads.


In 2024, we met our users’ curiousity of AI by releasing step-by-step AI flows that generate blog posts, landing pages, podcasts, and more.


But in 2025, customers expectations and comfort levels with AI were changing rapidly. To stay relevant in the market, we needed to adapt with our customer’s comfort level.

My PM, Eng, and I, as group leads, pivoted our team to creating AI Agents with a few guiding principles:

  1. Learn from 2024 shipped work

  2. Stay close to other teams building agents

  3. Keep scope small to ship in 3 months

Hubspot

My team drove key collaboration cycles across HubSpot with other agent teams to ensure we were building a cohesive, extensible, in-context experience.

Final AI Agent In-Context Design