New Supervised Mode takes Zuora AI from assistant to trusted team member

By Shealyn Nosal, Senior Product Marketing Manager

September 22, 2026

Engineers had their unlock a while ago: point an AI coding assistant at a codebase, and one developer ships work that used to take a team. Finance has been watching that shift from the sidelines, since “AI, go touch the ledger” is very different from “AI, go write some code.” If you’re lucky, the wrong suggestion in your general ledger gets caught by an auditor, and by the SEC if you’re not.

Today, Zuora AI is making that shift possible for finance. Now with Supervised Mode, AI can complete real quote-to-cash work, from write-offs to revenue releases and batch subscription changes, while a person still signs off on every one. We’re continuing to supercharge finance productivity with an operator that can actually complete tasks for you, governed the way finance already operates. 

And customers are already seeing results. 

“Because I’m spending less time chasing data, I can spend more time thinking strategically about scalability, process improvements, and how we want the business to operate.”

— Stacey Jones, Business Process Manager, Order to Cash Systems, ActiveCampaign

“I’m looking for efficiency and convenience. I found it in Zuora AI.”

— Chris Skalet, Senior Manager, Finance, Comcast Technology Solutions

What’s new: from answering questions to doing the work

Since our last announcement, Zuora AI has shipped more than 20 new supervised write actions. These are real actions an agent can now propose and, with user sign-off, execute. The single biggest expansion belongs to the Revenue Analyst Agent, which now can execute on 16 new actions such as:

• Deferring or releasing revenue on a contract or performance obligation, with a preview before anything posts
• Updating revenue forecasts, with current versus proposed shown side by side
• Managing revenue allocation and switching allocation treatment (carve-in versus carve-out)
• Applying or releasing revenue and transfer holds, voiding billing events, and processing revenue impairment

Other Zuora AI agents are now more capable too:

And completing any of these tasks just got easier. Zuora AI now features a curated library of more than 150 prompts across 16 categories, organized by product area and by the page you’re already on, so one click takes you from a blank chat box to useful work. You can also browse our full agent library here. 

25,000 conversations, real work done

Almost 75% of our customers are already using Zuora AI, generating more than 25,000 real working conversations. And over 1,200 of those asks were to take real action, not just answer a question.

ActiveCampaign’s team made Zuora AI part of its daily billing workflow for natural-language data queries, invoice and payment lookups, account and batch updates, and test-environment setup:

“Zuora AI turned me from the person answering every billing question into the person fixing why those questions kept coming up. That’s the shift from reactive to strategic.”

— Stacey Jones, Business Process Manager, Order to Cash Systems, ActiveCampaign

Hootsuite’s Revenue Accounting Operations team reconciled an account in under two minutes. And GitLab’s Revenue Systems Manager said Zuora MCP returns “clean, auditable answers. Not in days. In seconds.”

Here are a few more examples of how real teams are putting Zuora AI to work:

• A billing team applied a single payment across more than 40 invoices, by invoice number and exact amount, in one pass
• A revenue team updated the current and renewal term across 130 subscriptions in a single request instead of 130 separate edits
• A revenue team ran 1,194 individual payment-application actions across 727 distinct payment records on a single account — clearing more than $36,000 in accumulated unapplied credits, some as small as a few dollars each
• A finance team asked Zuora AI to split one $28,239 payment across two invoices. Confirmed complete in under a minute
• A billing team ran a supervised, two-phase batch job (a 3-subscription pilot to surface edge cases, then the full 1,911-subscription file) to resume and cancel subscriptions safely at scale.

AI that finance can trust

AI only earns a place in finance if it can answer three questions every time: what it did, why it did it, and who signed off. That’s the standard Zuora AI is built to meet. Every action runs through Supervised Mode, so a person reviews and approves before anything posts. Larger changes route automatically to a second reviewer through Approval Controls, and admins can set granular permissions by action and by role. The AI Audit Trail ties every result back to the conversation that produced it, and Watch Tower gives admins a single view of that activity across the org. Zuora MCP carries the same permissions and guardrails into the tools your team already uses, like Claude, Cursor, and Codex, so a read-only tenant stays read-only no matter where the request comes from. For finance teams, that level of governance is the reason AI can move from answering questions to taking action.

The real unlock

Finance is finally getting its version of the engineering unlock, safely. And it’s closing a real gap, since 97% of SaaS finance leaders say manual order-to-cash work still holds their team back.

A chatbot that only answers questions was never going to do this. Once AI can safely take the action, not just describe it, the constraint on finance capacity stops being headcount and becomes how much work a team hands off. It’s a real shift in how much a team can take on, without adding more headcount.

Want to stay up to date with the latest? Join our AI mailing list or talk to an expert to learn how Zuora AI can supercharge your team’s productivity.