Intent Solutions · Sample Output
This is what the operator actually gets: researched companies, verified contacts, a compliance block, a grounded draft, and the saved run record. The record shapes are the real schema from the repository.
Labeled sample
Every company, contact, and message on this page is fictional; the .example domains are reserved for documentation and cannot belong to a real business. The field names and structure come from pipeline_core/models.ts; the values were written for this demo. No customer data exists here.
Input
Phase 1 · Research, deterministic
One researcher agent per domain, fanned out. Connectors run in the order written in code and stamp their source on every fact.
Harborline Logistics
harborline.example
Freight forwarding, 180 employees. Opened two new port offices this year; hiring for a dispatch operations team.
source: apollo, exa
Corely Freight Systems
corely.example
Trucking software and brokerage, 95 employees. Recent press on late-delivery penalties in refrigerated lanes.
source: apollo, exa
Mapletide Supply Co
mapletide.example
Regional distributor, 60 employees. Job posts show a first dedicated sales-operations hire last quarter.
source: apollo
Phase 2 · Enrich, deterministic
Email verification comes from the enrichment connectors. Before anything is drafted, the compliance gate screens every contact; a blocked contact is recorded on the run and never reaches the drafting step.
Jordan Reyes
VP Operations, Harborline Logistics
j.reyes@harborline.example
Enrichment adds context the draft can use: Series B raised last year, dispatch team doubling per the hiring data above. Verified by Hunter, enriched by People Data Labs.
Priya Natarajan
Director of Operations, Corely Freight Systems
priya.n@corely.example
Verified email, LinkedIn profile matched. Fit signal: owns the refrigerated-lane performance problem named in the research.
Sam Okafor
COO, Mapletide Supply Co
s.okafor@mapletide.example
On the operator's do-not-call list, so the gate stops this contact before drafting. The block is recorded on the run record with its reason, which keeps the audit trail intact.
Phase 3 · Outreach, the only model calls
Two model calls per contact: a fit score, then a draft. The draft below is grounded; each personalized line traces to a research fact shown above, and the trace list ships with the draft.
To: j.reyes@harborline.example · Subject: Two new port offices and one dispatch queue
Jordan,
Opening two port offices in a year usually means dispatch is now coordinating across time zones with the same headcount. Congrats on the growth, and on doubling the dispatch team behind it.
We help operations teams at freight companies your size cut coordination time between dispatch and the dock without new software for the drivers. If that is on your plate this quarter, I can send a two-paragraph summary of how the last rollout went.
Worth a look?
Save · validated only
The validator mints the only record type the store accepts, then the run lands in a local file the operator owns: ~/.intent-outreach/runs.jsonl. Field names below are the real schema; values are this sample.
The cost figure is the meter's total for the four model calls in this sample (two contacts drafted, two calls each) at Sonnet-class rates, assuming roughly 2,000 input and 500 output tokens per call. The five skipped connectors had no keys set, so they were skipped and recorded as such.
← Back to the executive brief · How it works · What it touches