Salesforce Zendesk Integration for Renewal-Risk Alerts
Connect Salesforce renewal dates with Zendesk ticket signals so Customer Success can spot at-risk accounts early, explain the risk, and route a useful next step before renewal.

TL;DR: what you'll build. Renewal risk often slips through the gap between Customer Success and support, because the CSM owns the renewal in Salesforce and the support agent owns the ticket queue in Zendesk. This Salesforce Zendesk integration runs a daily Major app and agent that matches upcoming renewals with open high-priority tickets. CSMs get a reviewable brief before a support fire turns into a renewal problem.
What this renewal-risk alert does
A Salesforce Zendesk integration for renewal-risk alerts combines two kinds of data:
- Salesforce Opportunity renewal dates and Account context.
- Open Zendesk Tickets for the same customer.
The goal is that Customer Success stops finding out about a support fire the same week a contract is up.
The CSM owns the renewal, and the support agent owns the ticket queue. Neither one normally looks at the other's system until a deal is already shaky.
This article is a build spec for a daily Major app and agent that closes that gap. It is written against the Zendesk and Salesforce APIs as documented, not as imagined.
Where this fits in the Major hubs
The pattern sits at the intersection of two Major hubs:
- AI customer service agents for the support side.
- Salesforce renewal automation on the CRM side.
If your team already runs Salesforce automation for pipeline hygiene, this is the same app-plus-agent approach applied to the handoff between CS and support.
Why renewal risk gets lost between Salesforce and Zendesk
A renewal that is 60 to 90 days out lives in a Salesforce Opportunity. The account's support history lives in Zendesk as a stream of Tickets. Nothing forces those two records to meet.
A CSM checks Salesforce for the renewal date and the ARR, glances at a dashboard, and moves to the next account. A support agent closes or reprioritizes tickets with no signal that the account is 45 days from renewal.
The result is a familiar failure mode. A VP escalates a long-open, high-priority ticket in the same week the renewal was supposed to close quietly. The CSM reads about it for the first time in a Slack message from their manager.
Why the native connector does not solve it
Zendesk's own documentation describes its native Salesforce connector as a one-way sync:
- Salesforce accounts become Zendesk organizations.
- Salesforce contacts or leads become Zendesk users.
- Changes made in Zendesk do not flow back to Salesforce.
In Zendesk's words: "Because data syncing is one-way, any changes applied in Zendesk are not reflected in Salesforce."
It is record sync, not risk logic. A CSM still gets nothing that says "this account's renewal is close and its tickets are ugly."
Who owns the problem
The job belongs to the CSM, but it draws in two other owners:
- The support lead, whose queue holds the evidence.
- CS leadership, who needs the subset of alerts that touch the company's biggest accounts.
Build the matching, filtering, and routing as deterministic software, and save the agent for the one judgment call that needs language and context. That split is what makes the alert useful instead of one more dashboard nobody opens.
What data the workflow needs from Salesforce and Zendesk
Neither system carries renewal-risk fields natively. That is why the workflow has to read from both and compute the overlap itself.
Salesforce: authentication and objects
Authentication is OAuth 2.0 through a connected app or, going forward, an external client app. Salesforce is restricting new connected-app creation as of Spring '26 in favor of external client apps. Existing connected apps keep working.
The REST API exposes standard and custom sObjects (Salesforce's name for its record types). This workflow uses two:
- Opportunity: the close date functions as the renewal date in most CS process setups. It also carries Amount for ARR and an AccountId lookup to the related account.
- Account: holds the owner, plus a tier or segment field if your org has configured one. That tier field is a custom field, not a universal API guarantee.
Salesforce: reading records and staying within limits
Retrieving an sObject record is a straightforward GET against /services/data/vXX.X/sobjects/Opportunity/{id}. For a date-ranged pull, use a SOQL query against the Query resource instead.
Org-wide daily API call limits vary by edition and license. Salesforce reports usage on every response in the Sforce-Limit-Info header, documented in Salesforce's API usage headers guide.
Salesforce has no generic outbound webhook. The event-driven options are Platform Events or Change Data Capture over the Pub-Sub API, which is more infrastructure than a daily batch job needs.
Zendesk: authentication and tickets
Authentication is OAuth 2.0 bearer tokens (preferred) or an API token with Basic Auth, per Zendesk's security and authentication docs.
The Tickets endpoint under the Ticketing API returns the fields this workflow needs:
- Status.
- Priority.
- Created at and updated at timestamps, which give you ticket age.
- Requester and
organization_id, which you match back to the Salesforce account.
CSAT, where your plan collects it, comes through the Satisfaction Ratings endpoint, not the ticket object itself.
Zendesk: rate limits by plan
Rate limits are documented per plan:
- Team: 200 requests per minute.
- Growth and Support Professional: 400 requests per minute.
- Enterprise: 700 requests per minute.
- Enterprise Plus: 2,500 requests per minute.
- High Volume API add-on: available for accounts that need more.
Every response carries the X-Rate-Limit and X-Rate-Limit-Remaining headers. A 429 response comes back with a Retry-After header that tells you how long to back off.
Why a nightly pull beats a webhook here
Zendesk supports native webhooks and Zendesk Integration Services (ZIS) inbound webhooks for event-driven triggers. A webhook is like a doorbell: Zendesk rings your system whenever something changes, so you need someone listening at the door all day.
For this use case, a nightly pull against the Tickets endpoint, filtered by organization and priority, is simpler to reason about. It is also cheaper to run than standing up a webhook listener.
Compliance: scope access like the source systems
Your org may be subject to GDPR Article 28 processor obligations or operate a SOC 2 program under the AICPA Trust Services Criteria. If so, the one requirement that matters here is access scoping.
The snapshot you build combines customer PII and account financial data. It needs the same confidentiality controls as the source systems, not looser ones because it is "just an internal dashboard."
Split the work between the app, agent, and workflow
Each part has one clear job, which keeps the output repeatable and auditable.
The workflow: the deterministic backbone
The workflow is a daily schedule trigger, a fixed sequence of API calls, and an explicit approval gate. On each run it:
- Pulls Opportunities renewing in the next 90 days.
- Resolves the related Account.
- Fetches open high-priority Zendesk Tickets for that account's organization.
- Writes a timestamped snapshot.
- Hands off to the agent.
None of this needs judgment, so none of it runs through a model.
The app: where state lives
The app holds a managed Postgres table for snapshots, with one row per account per day. Every source ID is preserved, so a CSM or an auditor can trace a risk flag back to the exact Opportunity and Ticket records it came from.
The app also serves the CSM-facing queue: a filtered, sortable list of today's at-risk accounts, with drill-down into the evidence.
The agent: one narrow job
Given a snapshot's recorded evidence, the agent writes a short, grounded explanation and a draft brief for the CSM. It has firm limits:
- It does not re-fetch data.
- It does not decide whether an account is "actually" at risk beyond what the deterministic filter already flagged.
- It never writes back to Salesforce.
- It never contacts the customer.
That boundary is the whole point. The matching and filtering should never vary from run to run. Explaining evidence and drafting prose is the one part that benefits from language judgment.

Renewal Watch architecture: trigger, connectors, app, agent, and routing
Build your Salesforce Zendesk integration for renewal-risk alerts on Major
Follow these steps in order. The example prompts follow the list.
- Connect Salesforce and Zendesk through their official Major connectors (Salesforce via OAuth connected app, Zendesk via OAuth or API token), scoping Salesforce to read access on Opportunity and Account and Zendesk to read access on Tickets and Organizations, so Major's credential proxy holds the credentials instead of any app or agent code. You should see both connectors listed as connected with read-only scopes on exactly those objects.
- Build the app by first modeling the data: create a
renewal_snapshotstable (accountid, opportunityid, renewaldate, arr, accounttier, snapshotdate, openticketcount, highestpriority, oldestopenticketagedays, and sourceticketids as an array) and arenewal_alertstable (snapshotid, status of pending, reviewed, escalated, or dismissed, csmowner, reviewedby, reviewedat), keeping the originating Salesforce and Zendesk record IDs on every row, then prompt the Renewal Watch app build using the first example below. You should see an internal app called Renewal Watch with a snapshot table, a CSM queue filterable by tier and renewal window, and a detail panel linking back to source Salesforce and Zendesk records, with nothing in the UI unsourced. - Add the agent with the second example prompt below, so it reads one renewal snapshot row at a time, uses only that row's renewal date and ticket evidence, reports missing data as unknown rather than zero risk, and drafts a two-paragraph brief for the account's CSM without altering Salesforce or contacting the customer. You should see a draft brief tied to a single snapshot, citing only the evidence stored in that row.
- Set the workflow trigger as a daily scheduled run early in the morning, before the CS team's day starts, that pulls Opportunities with a close date inside the 90-day window via SOQL, resolves each Account, queries Zendesk Tickets by organization_id filtered to status:open and priority:high or urgent, writes the snapshot rows, then calls the agent once per new or materially changed snapshot. You should see fresh snapshot rows dated today and new briefs only for accounts whose evidence is new or has materially changed.
- Test the failure paths: force an API error and confirm the account is marked stale instead of skipped, fire the same Opportunity update twice and confirm the app upserts on accountid plus snapshotdate, check that an ambiguous account match lands in manual review, and confirm an account with no open high-priority tickets stores zero while a failed Zendesk query shows "data unavailable." You should see one queue entry per account per day, stale or unmatched accounts clearly labeled, and no false "no risk" results.
- Roll out with permissions and the human review gate so every drafted brief lands in the CSM's queue as "pending review," never as a sent email or a Salesforce field update, with CSMs seeing only their own book and top-tier accounts (by whatever segment field your org already uses) routing a copy to a CS leadership view on the same approval gate. You should see CSMs approving, editing, or dismissing briefs from their own queue, and leadership seeing top-tier copies without any autonomous escalation.
Build an internal app called Renewal Watch with a table for daily account snapshots (Salesforce Opportunity and Account fields, plus Zendesk ticket counts and priority) and a CSM-facing queue view, filterable by tier and renewal window, with a detail panel that shows the source Salesforce and Zendesk record links for every flagged account.
Build an agent that reads one renewal snapshot row at a time. Use only this snapshot's renewal date and listed ticket evidence. Explain the signals and call out any missing data as unknown, not as zero risk. Draft a two-paragraph brief for the account's CSM. Do not alter Salesforce records and do not contact the customer.

Daily snapshot, agent rationale, and the human review gate
Governance and IT control
Access and credentials
- SSO and role-based access: CSMs see only their own book of accounts in the queue. CS leadership sees the escalation view. Neither role ever sees the underlying Salesforce and Zendesk credentials, which are exposed only to the workflow's scoped connector.
- Scoped secrets via the credential proxy: the Salesforce and Zendesk OAuth tokens live in Major's credential proxy, not in agent prompts or app code. They are scoped to read-only access on the specific objects this workflow needs.
A credential proxy works like a hotel front desk that holds your room key. Staff can open the one door they are allowed into, but nobody walks around carrying a master key.
Records, history, and change control
- Audit logs: every snapshot row and every agent-drafted brief is logged with its source record IDs. Every CSM review action (approve, edit, dismiss, escalate) is logged with a timestamp and the reviewing user, so a renewal retro can reconstruct exactly what evidence existed and who saw it when.
- Managed database and storage: snapshots and review history live in Major's managed Postgres, not in a spreadsheet or a model's context window. The history survives past any single conversation or run.
- Versioned deploys: changes to the matching logic, such as the 90-day window or the ticket priority filter, go through a versioned app deploy, not an ad hoc prompt edit. A change in what counts as "at risk" stays reviewable and reversible.
Approval required
Any action that reaches the customer, or writes back into Salesforce, requires a person. The agent's output is always a draft in a queue, never a sent message or a committed CRM update.
Failure handling and edge cases
Each case below has a defined response so the queue never misleads a CSM.
API errors and rate limits
- Salesforce or Zendesk API errors: a failed call during the daily pull should not silently skip an account. Log the failure against that account and retry with backoff. If it still fails, mark the account's snapshot as stale rather than omitting it, so a CSM sees "data unavailable" instead of a false "no risk."
- Rate limits: respect Zendesk's Retry-After header on a 429, and watch Salesforce's Sforce-Limit-Info header proactively. If the account list is large, batch the Zendesk ticket pulls by organization rather than firing one request per account.
Duplicates and matching
- Duplicate or near-duplicate events: the same Opportunity might trigger two snapshot writes in a day, for example when a field update fires twice. The app should upsert on accountid plus snapshotdate, not create duplicate queue entries.
- Ambiguous account matching: Zendesk's own sync logic logs an unsuccessful match and does nothing further when more than one record matches. This workflow should do the same and flag an unmatched or ambiguous account pairing for manual review rather than guessing.
An upsert is like a sign-in sheet with one line per person per day. If you sign in twice, the second signature replaces the first instead of adding a new line.
Empty and stale data
- Empty ticket results: no open high-priority tickets is a real, useful signal, not an error. Store it as zero, distinct from "Zendesk query failed."
- Stale snapshots: if the last successful Zendesk pull for an account is more than a day old, show that age in the CSM queue. Nobody should treat three-day-old ticket data as current.
What the agent must never do
The agent must never contact the customer, write to Salesforce, or invent evidence not present in the snapshot. Missing data is reported as unknown, not folded into a risk score.
What to measure
- Snapshot coverage: the share of renewing-in-90-days accounts that got a complete, non-stale snapshot each day.
- Time from flag to CSM review: how long a pending alert sits before a CSM opens it.
- Escalation accuracy: the share of leadership escalations a CSM confirms versus dismisses as noise. This tells you whether the ticket-priority filter is tuned right.
- Manual touches avoided: how many accounts a CSM would otherwise have had to check by hand across both systems each week.
- Review-to-resolution time: for escalated accounts, how long between the alert and the renewal actually closing or churning, tracked in Salesforce as it already is today.
Before you go live
- [ ] Salesforce is connected with read-only scope on Opportunity and Account, and Zendesk with read-only scope on Tickets and Organizations, both through Major's credential proxy.
- [ ] The 90-day renewal window and the open, high or urgent priority ticket filter are set in a versioned app deploy, not in a prompt.
- [ ] Every snapshot row stores its source Salesforce and Zendesk record IDs, and the app upserts on accountid plus snapshotdate.
- [ ] Failed API calls mark the account as stale with "data unavailable," and empty ticket results store zero.
- [ ] Ambiguous or unmatched account pairings route to manual review instead of being guessed.
- [ ] The agent's prompt forbids writing to Salesforce, contacting customers, and inventing evidence, and it reports missing data as unknown.
- [ ] CSMs see only their own book, top-tier copies route to the CS leadership view, and every brief starts as "pending review."
- [ ] Audit logs capture every snapshot, drafted brief, and CSM review action with a timestamp and user.
Start building on Major
Build a renewal-risk app that keeps Salesforce renewal context and Zendesk ticket evidence together, then gives CSMs a reviewable daily brief.
Related articles
- NetSuite Integration: Automate the Manufacturing Handoff
- AI Agents for Real Estate: 6 Workflows That Stay Governed
- AI Agents for Business: 6 Workflows That Run Reliably
About the author
Ritvij Gautam is Head of Growth at Major.
Last reviewed November 6, 2026 by Ritvij Gautam.
Frequently asked questions
- Can Salesforce and Zendesk be integrated for renewal-risk alerts?
- Yes, but not through native sync alone. Zendesk's Salesforce connector copies accounts and contacts one way into Zendesk as organizations and users; it does not compute risk. A renewal-risk workflow reads Salesforce Opportunity and Zendesk Ticket data separately, matches them by account, and applies its own 90-day and ticket-priority logic on top.
- What data should a renewal-risk workflow use?
- Salesforce Opportunity (renewal/close date, ARR, account reference), Salesforce Account (owner, and tier if your org tracks it as a custom field), and Zendesk Tickets (status, priority, age, and organization). Tier and CSAT fields depend on how your org configured each system, not on a universal API guarantee.
- Can Zendesk tickets sync back to Salesforce?
- Zendesk's documented Salesforce integration syncs Salesforce accounts, contacts, and leads into Zendesk, one way. Zendesk's own help docs state that changes made in Zendesk are not reflected back in Salesforce through that sync. Ticket-to-Salesforce sync, where it exists, is a separate, narrower feature scope; check current Zendesk documentation before relying on it.
- How should an AI agent handle a high-risk renewal?
- It should explain the recorded evidence in a snapshot, draft a brief for the CSM, and stop there. It should not contact the customer, change Salesforce records, or invent risk signals beyond what the deterministic workflow captured. A person reviews and approves before anything reaches the customer or the CRM.
- Why not just build a Salesforce-Zendesk dashboard instead of an agent?
- A dashboard still requires someone to check it daily and interpret raw numbers. The agent turns a snapshot's open tickets and renewal date into a short, evidence-grounded brief a CSM can act on immediately, while the dashboard-equivalent data and matching logic stay in deterministic app code that runs the same way every day.