Signal Based Account Prioritization for Procore Sales Reps
Replaced gut feel prioritization with a live signal engine that tells reps who to call first and why, then automates the alert straight to Slack.
situation
Reps were prioritizing on instinct. Nothing surfaced which accounts were showing real buying intent versus which were simply next on the list. The goal: help reps prioritize early using real world signals tied to construction activity, so effort goes toward accounts more likely to convert.
result
Built a signal based tiering system that pulls construction companies across Canada and the US, enriches them, and layers on two live signals: project manager hiring as a proxy for active or upcoming projects, and news coverage of new construction projects, read and interpreted by AI rather than keyword matched. Two or more signals flags Tier 1 and hot. One signal is Tier 2. Tier 1 triggers an AI drafted Slack message to the rep, so the account surfaces as a ready to action item instead of sitting in a report nobody opens. The logic branches for new prospects versus existing CRM accounts, so the same engine works either way.
impact / results
Cuts the manual research a rep would normally burn an hour on down to a Slack ping with the reasoning already attached. Reps stop guessing which of hundreds of accounts to call first and start every day working the ones already showing real signals of a coming purchase, the exact shift that shortens sales cycles and lifts win rates on the accounts that matter most.
tech stack
Clay (filtering, enrichment), Claygent (AI read news summaries), Find Jobs (hiring signal), Slack (rep alerts), meta prompting (tiering logic).
ConstructConnect via Salesforce sync for the richer, non delayed signal, web scraping to backfill job and project signals where clean API access does not exist, Salesforce integration for production triggers, a scoring model with more than two signals for finer tiering.
process map
reflections
The pseudocode and the build matched on signal logic and the Slack output. The gap was data access. ConstructConnect, the source with the least lag, does not integrate with Clay, so I substituted hiring and news signals. They work, but they arrive later than direct access would.
The real unlock was realizing ConstructConnect does integrate with Salesforce. So the right architecture is not "Clay talks to ConstructConnect," it is "ConstructConnect populates Salesforce, and Clay reads Salesforce." A path around an integration that does not exist. That is the lesson AlphaForge kept pushing: anticipate where the build will break before you start, and get creative about the route around it rather than stalling on the one that is not there.