Latest AI Week at Zuora

Once we heard about Amplitude's successful AI Week, we knew we had to host our own for our Product & Engineering group. We even managed to get Wade from Amplitude to come talk to us.

Our AI Week sessions included:

  • LLM Internals - what's inside the black box and why does that matter?
  • Prompt Best Practices - good prompts aren't easy.
  • Our internal AI architecture - so we can ensure the uniformity and centralized management needed for compliance, security and to keep our costs under control.
  • UI Toolkit for AI - Our UX team created our 'UI Toolkit for AI', one design language and toolkit to bring our AI based solutions to life. Whether a team is building AI for Assistant, Insight or Agent experiences, we now have one design language.
  • Developer Reference Store - a prototype repository to ensure the highest consistency from prompt based features.
  • Aman Singla, from Immersa , discussed effective context management, we need our LLM to reason effectively through tasks, not just retrieve answers.
  • Wade Chambers, from Amplitude , shared their experiences in shipping AI-based products.
  • Best practices based on hard won experiences for fine tuning and testing all our AI-solutions.
  • Existing AI-based project review - we already have many in flight AI related projects, so here we stepped back to review what we learned and what we will do better. And we set a goal of 50 such projects by fiscal year end (Zuora's fiscal calendar ends in January).
  • Our CISO reminded everyone of our AI Policy so we know where the guard rails are.

And we saw Windsurf, Claude Cursor, Gemini, Vercel, and Trupeer used in our AI Week 'hands on' practical sessions.

To learn more about these sessions, use the feedback link in the corner and tell us your interest.

We're not new to the AI game. Zuora started leveraging machine learning long before ChatGPT's big debut. Back in 2021, we leveraged our database of tens of millions of anonymized payments across different card networks, countries, and payment gateways, to predict when the next payment retry should be attempted as that was the time the retry is likeliest to succeed. This became our Smart Retry feature and our customers measure Smart Retry's success using the additional dollars our 20% higher retry collection rates provide.

We've already shipped other AI-based features. From an AI assistant for querying metrics that doesn't necessitate sharing your tenant's data with the LLM to our Order Entry feature that accepts a PDF of a contract or signed quote and builds a Zuora Billing Order for you to review before invoicing.

Reminder: Our public AI policy document guides all this work. We'll always be clear when we are using AI and we'll safeguard your data just as we always have. AI should augment and empower humans, not replace them.