Director, Digital Support

View my resume

My vision for the role is to build a lightweight support system that relies on customer journeys across different Cursor touchpoints (help center, in-app, forums, etc) to self-update& maintain with minimal human oversight as Cursor continues to evolve and grow.

For that, I'm prioritizing the following challenges with a proposed strategy to tackle each one of them:

Challenge 1: Run the help center

Information architecture, content quality, findability, and the workflows that keep help accurate as the product ships fast.

The current site relies on 8 main topics and a (very) useful chatbot that I'm sure provides a lot of insight as to what is being asked and clicked. I believe the following strategy can help push the current site to the next level of autonomy & helpfulness.

1) Measure

Current articles & right hand chat do not ask for explicit issue resolution; adding a simple thumbs up & down would give us enough signal to prioritize unhelpful experiences and drive content quality improvement. What gets measured gets managed.

2) Autonomy

The way I would prioritize content design & good information architecture would be based on two main pillars:

  • Help center chat queries: Distinguishing issues by using frequency, pain points, and papercuts as parameters to prioritize content update.
  • Cursor agent chat queries:This is the biggest lever. Analyzing direct user queries about product usage itself (distinct from actual coding questions) will highlight key customer journeys and give deep insight into where users struggle when building inside Cursor. Potential engagement & retention analysis could be built on top of these since we can track how many of them gave up or continued with their coding journey.

Essentially, we need to analyze conversation logs of unsuccessful, platform-dependent user queries to the Cursor agent. Building an automation that runs this analysis on a weekly basis will automatically rank the help center topics that need immediate attention. By understanding frequency patterns paired with new product updates, there is an untapped potential for predictive support to completely avoid a bad experience.

To keep it agile, I would also monitor and audit change logs from the latest releases to cross-match and clean up any irrelevant or outdated information in the help center. This creates a flywheel effect: change log → auto-updates → identifying new issues from agent chats → updates.

3) Human & Agent Discoverability

As the majority of this discoverability increasingly comes from AI agents inside (and outside) Cursor, one of our main priorities must be serving agent support requests together with human requests. Ideally, we could understand the exact share of human/agent traffic to prioritize and adjust our effort on each side of the flywheel. With recent updates giving Cursor more portability (e.g. Cursor SDK), agents are now the ones actively crawling through platform and support documentation.

I also believe there should be a tight loop between the help center & cursor.com/docs as they fundamentally need to serve the same truths about the platform and feed off each other. Finding a way to connect these two would be our second-order priority, coming just after automating the help center self-update loop mentioned above.

Challenge 2: Own the forum

Scaled support and community surface — moderation strategy, health, tooling, and how forum signal flows back into product and support.

As a support lead at Stripe and community leader for aibuilders.mx, I can understand why this is the most difficult challenge in terms of information density, variety, & complexity. Mainly because of the diverse spectrum of users cohabiting the same space (from beginners to hardcore SWEs).

Communicating customer pain points to product teams is always tough, especially when they are scattered as anecdotes across a forum. To solve this, I would take a dual approach: create a segmentation framework for moderators to rely on for tactical, complex issues, while simultaneously piloting a RAG assisted bot trained on the internal Cursor codebase & docs to help resolve highly technical platform bugs or simple questions.

For the product feedback loop, forum signals would flow back into Product based on frequency and severity. Most of our initial effort would go into building an algorithm that weighs and balances these two variables, utilizing that same team of moderators to continuously audit, intervene, and calibrate it.

Challenge 3: Define the metrics and operating rhythm

Adoption, resolution, quality, customer effort, and community health.

I've always believed that KPI tracking should be simple. As a product grows, it's only natural to want to track every single metric, but I've found that having just a handful of core metrics across quality, efficiency & delivery is more than enough to set a clear direction and have a clear threshold to understand when something needs immediate attention. (I personally use the Andon Light system for this)

All in all, even though it isn't explicitly mentioned as a channel, I would heavily index on in-app, proactive support. Having the model predictively understand when a user is hitting a platform-level issue and stepping in to help is the single biggest lever Cursor holds today for enabling an excellent user experience.