Forecast filed Thursday. Wrong by Monday.
Stage updates rely on rep confidence, not the call, the tickets, or the activity history. The forecast is a guess.
Build a Pipeline Review Agent withSalesforce, Gong, and Zendesk Support
One prompt queries your Salesforce pipeline, Gong call sentiment, and Zendesk support health as a single connected dataset. Every deal risk-scored. Every gap surfaced. Your Monday forecast, done before coffee.
Three tools, three tabs.
Zero confidence in your forecast.
Pipeline reviews run on incomplete data because the signals you need are scattered across Salesforce, Gong, and Zendesk.
CRM says one thing. The buyer said another.
Salesforce stage and Gong call sentiment never compare. The last call has the truth — and it's locked in another tool.
Support sees churn first. Sales finds out last.
Ticket spikes and CSAT drops live in Zendesk. The CRM account stays "Green" until the renewal is already gone.
Under The Hood
One prompt. 3 data sources. Already connected.
01
Query open opportunities from Salesforce
stage · amount · close date · last activity · owner
02
Pull call sentiment from Gong on the same accounts
buyer sentiment · competitor mentions · next steps · economic buyer
03
Check Zendesk for open tickets on those accounts
open count · CSAT score · unresolved escalations
output
A single risk-scored pipeline brief
Every deal scored 1–10. Top three risks called out. Suggested next actions per deal.
The Context Store
Your data is pre-joined before the agent runs.
Airbyte connects Salesforce, Gong, and Zendesk Support into the Context Store — a unified data layer where account IDs and structures are already matched.
Your agent queries one surface instead of three APIs. Faster responses, lower cost per query, and results that work because the relationships were built before you asked the question.
Pre-joined accountsUnified schemaSingle API surface
The Prompt
Copy. Paste.
Run your pipeline review.
Two steps. Your data, your results, under 60 seconds.
01install
Connect the Airbyte Agent MCP02copy and run
I want to run a pipeline review that combines CRM, call, and support data.
SETUP
You have access to the Airbyte Agent MCP, which connects to 50+ SaaS tools and lets you query live business data through natural language.
Workflow: List available connectors > connect to Salesforce, Gong, and Zendesk > explore schemas > query and cross-reference data > analyze and deliver.
If a connector isn't linked yet, the error is self-descriptive. Follow the auth flow, then retry.
TASK: Weekly Pipeline Review Brief
Likely connectors: Salesforce, Gong, Zendesk Support.
Steps:
1. Connect to Salesforce, Gong, and Zendesk Support
2. Pull all open Opportunities closing this month from Salesforce with stage, amount, owner, close date, and last activity date
3. For each deal, find the latest Gong calls from the last 30 days. Extract: buyer sentiment, competitor mentions, whether next steps were confirmed, and if the economic buyer was on the call
4. For each deal's account, pull Zendesk tickets from the last 90 days. Calculate: open ticket count, average CSAT, and any unresolved escalations
5. Score each deal's true close probability by combining:
- CRM signals (stage, activity recency, days in stage)
- Call signals (sentiment, buyer engagement, next steps)
- Support signals (ticket volume, CSAT, escalations)
6. Flag deals where CRM stage confidence doesn't match call or support reality
7. For each at-risk deal, recommend a specific next action
8. Output a pipeline review brief: risk-ranked deal list, total pipeline at risk, and a summary I can use to run the forecast meeting
Ask me before any action that sends messages or modifies data in a connected app.The Outcome
A pipeline review that takes your team 3 hours, in 60 seconds
10x
Faster
Your agent queries pre-joined data instead of making live API calls to three tools. What used to take minutes takes seconds.
90%
Cheaper to run
Pre-built data joins mean your agent uses fewer calls and less compute per query. More agents, same budget.
3 → 1
Tools, one query
No API orchestration, no retry logic, no rate limits. The Context Store handles the complexity so your prompt stays simple.
Based on internal benchmarks comparing Context Store queries to sequential API calls across equivalent datasets.
01 · Output
Risk-scored deal list
Every open deal scored on CRM stage, call sentiment, and support health. Not what the rep entered. What the data says.
02 · signal
CRM vs. reality flags
Deals where the stage tells one story but Gong calls and Zendesk tickets tell another. The gaps your forecast never catches.
03 · context
Account health overlay
Support ticket volume, CSAT trends, and open escalations layered onto every deal. The renewal risk your AE can't see from the CRM.
04 · action
Next actions per deal
A specific recommendation for each at-risk deal: schedule a call, loop in an exec, address open tickets, or re-qualify.
05 · summary
Forecast-ready brief
A summary formatted to run your pipeline meeting: total pipeline, pipeline at risk, and the 5 deals that need discussion today.
More agents for your revenue stack
view All use casesHow do I build a pipeline review agent with Salesforce and Gong?
Install the Airbyte Agent MCP, connect to Salesforce and Gong, then paste the prompt above into your preferred AI client. The agent handles the rest.
What Salesforce data does the agent access?
Open opportunities with stage, amount, close date, last activity, and owner. No bulk data export — just the fields your pipeline review needs.
Can I combine Salesforce, Gong, and Zendesk data in one query?
Yes. The Context Store pre-joins your data so the agent queries one unified surface. Account IDs are matched across all three tools before you ask a question.
How is this different from using the Salesforce, Gong, and Zendesk APIs directly?
Direct API calls require orchestration, retry logic, and rate-limit handling. The Context Store pre-joins your data so the agent queries one surface — faster, cheaper, and with pre-built relationships.
How long does setup take?
About 2 minutes. Install the Airbyte Agent MCP, authenticate your tools, and run the prompt.
What AI clients work with the Airbyte Agent MCP?
Claude, ChatGPT, Cursor, Codex, and any client that supports the Model Context Protocol.
Your data is already in these tools. Now give your agents access to it.
Connect Salesforce, Gong, and Zendesk Support and 50+ other tools to the Airbyte Agent MCP. Build agents that actually know your business.
