Joseph Deville — GTM systems shipped at Cortex

The problem I was hired into

A GTM org running on manual dispatch, stale account prioritization, and integrations that failed in front of customers instead of before them. The ask was automation. The actual constraint was that nothing produced evidence of its own value, so every improvement was an opinion.

I built each system with two outputs: the thing it did, and the record of what it touched. The second output is why the work survived a budget cycle.

What shipped, and what it wrote to

SE request routing, in Slack

Sales engineers were matched to deals by hand, which meant a dispatch bottleneck sitting between a rep asking for help and getting it. I built a Slack app that read deal context and routed the request automatically, in the channel where the ask already happened.

Wrote to: CRM opportunity records, Slack threads
Measured by: Time from request to assigned SE, requests handled per week

Signal-driven BDR lead feed

Prioritization lived in spreadsheets that went stale the day they were built. I integrated enrichment sources with the CRM and internal databases to put a ranked, refreshed account list inside the daily workflow reps already used, rather than in a tool they had to remember to open.

Wrote to: CRM lead and account records, rep daily queue
Measured by: Accounts worked from the feed, meetings sourced, opportunities created

Competitive intelligence, inside the deal

Competitive intel usually lives in a dashboard nobody opens. I built monitoring with AI analysis that wrote findings directly onto the deal record, so a rep prepping a call saw the intel in the place they were already looking.

Wrote to: CRM deal records
Measured by: Competitive deals with intel attached at close, win rate on those deals

Demo preflight validation

Integration failures were being discovered live, in customer-facing sessions. I engineered a preflight check across CRM (SOAP auth), call recording, the application database, and issue tracking, run before every demo. Failures surfaced in advance instead of on the call.

Wrote to: Pre-session status checks, issue tracker
Measured by: Failures caught pre-session, demos lost to integration errors

POV deck automation and product feedback logging

Two manual handoffs between sales and product, both of which quietly dropped information. I automated proof-of-value deck generation and built an auto-logger for product feedback requests, standardizing the data moving between the two systems.

Wrote to: Product feedback backlog, opportunity records
Measured by: Decks generated, feedback items logged and routed

How the attribution actually worked

INSTRUMENT

Measure at build, not after

Every system logged what it touched from day one. Reporting bolted on afterward never reconciles, and you end up arguing about attribution instead of scaling what works.

TRACE

Influence, stated as influence

Each system had a defined touch on a record. The claim was influence on those opportunities, not sole credit, and the logic was published alongside the number so it could be challenged.

GATE

Humans in the critical path

Anything writing to records reps depend on, or reaching a customer, kept a review checkpoint. Automation earns trust by being interruptible.

Five systems, and the attribution built into each one before it shipped.