derr.ai
contact
derr.ai
> /projects/deal-risk-surfacing
# title

Deal-risk surfacing across Salesforce, Gong, and email

# context

Outbound and pipeline at a B2B SaaS commercial team. AEs were missing deal-risk signals that emerged across multiple data sources. Calls, emails, field changes. Risk surfaced too late and deals slipped at quarter close.

Forecasting suffered because at-risk deals were invisible until the AE had a feel for it.

# build

A daily-cron agent that pulls from Salesforce via a custom MCP. OpportunityFieldHistory, MEDDPICC custom fields, related Tasks. Combined with Gong transcripts from the last 3-5 calls and email engagement signal.

A risk categorization engine surfaces five named risk types. Champion fade. Decision-maker disengagement. Close-date drift. Competitor mention spikes. Activity gaps.

The agent writes a Slack DM to the AE with the specific evidence the agent saw plus a suggested next move. It writes a Task back to the Opportunity in Salesforce. It never overwrites field data. Pushback in the Slack thread feeds back into the prompt-template tuning loop.

# outcome
  • Deal-slip rate on flagged opps dropped meaningfully versus the unflagged baseline
  • Forecasting accuracy improved because at-risk deals stopped surprising the team at quarter-end
  • 90%+ AE acknowledgment rate within 30 days
  • False-positive rate dropped below 10% after 90 days as the categorization improved from corrections
# what this proves

Salesforce stays the system of record. The agent writes Tasks only. Multi-source data fusion makes the risk signal sharper than any single source alone.