Interactive specifications mapping technical Zod mutations, parallel search packs, and event pipelines.
Validates all core representative profiles and target URLs prior to launching intensive background actions. Ensures runs do not fail midway due to missing schemas or invalid structures.
In v1, there was no upfront schema check. If a Company URL or Persona Target was malformed, the pipeline would crash minutes into the execution, wasting tokens and user time.
export const runPrep = mutation({
args: {
companyUrl: v.string(),
contactUrl: v.string(),
voiceProfile: v.id("voice_profiles")
},
returns: v.object({
status: v.string(),
runToken: v.string()
}),
handler: async (ctx, args) => {
if (!args.companyUrl.includes(".")) {
throw new Error("Invalid Company URL");
}
// Proceed to Step 2...
}
});
Adaptive Scenario Complexity Routing: By mapping GTM scenario classes directly to custom Zod-validated configurations, we remove unnecessary database lookups and prevent token burn on cold outreach.
| Scenario Class | Scenario Family | Research Depth | Default Output Pack | HITL Review Gate |
|---|---|---|---|---|
| Cold outreach / intro call | Outreach | Light (2-3 queries) | Call Prep Dashboard + Agenda + Question slider | Brief summary check |
| Follow-up / reschedule with value | Follow-up / Reschedule | Light (2-3 queries) | Follow-up email + Optional one-pager | Brief summary check |
| Post-call follow-up (any scenario) | Follow-up / Reschedule | Light (2-3 queries) | Summary + Email; add slides/one-pager for Full | Brief summary check |
| Partner intro / discovery prep | Discovery / Intro | Standard (4-5 queries) | Partner research brief + Joint messaging draft + Agenda | Full inline editor review |
| Discovery call (customer) | Discovery / Intro | Standard (4-5 queries) | Findings brief + Discovery Qs + Trap/Next-step Qs + One-pager | Full inline editor review |
| Company research only | Research / Campaign build | Standard (4-5 queries) | Org chart + Tech stack + Findings brief | Brief summary check |
| Campaign build (persona/industry) | Research / Campaign build | Deep (6-8 queries) | Topic research + Strategy overview + Outreach slides/infographic | Full inline editor review |
| Phase | Scenario Family | Required Research Packs | Default Outputs | Opt-in Extras |
|---|---|---|---|---|
| Prep | Discovery / Intro | Account + Persona Packs | Call Prep Dashboard; Meeting Agenda; Discovery questions; Strategy overview | Org chart; Tech stack; Qualification questions |
| Meeting | Discovery / Intro | Reuse prep packs (Cached) | Meeting slides/one-pager; Question slider | Solution builder; Infographic |
| Follow-up | All families | Transcript intake | Transcript → synthesis feed into Follow-up pack | Skip if no recording available |
| Prep | Outreach | Account + Persona + Opportunity | Outreach strategy overview; outreach slides/infographic; trap + next-step questions | Partner joint messaging; campaign variants |
| Any | Research / Campaign build | Account (deep) + Persona | Industry or persona campaign brief; Research memo | Inventory collector; Contract collector |
| Follow-up | Follow-up / Reschedule | Opportunity + Transcript (if available) | Synthesize follow-up; Follow-up email; Custom one-pager | Custom slides/infographic; Reschedule value brief |
GTM Workflow v2 Optimization Objectives: Re-scaffolding GTM processes to collapse human steps, dynamically restrict parallel search depth, and establish closed-loop transcript pipelines.
Executed an 8x4 query grid for every single company context regardless of scenario, causing massive run latencies (3+ mins) and unnecessary token burn.
Forced users to progress through three distinct, separate checkpoints (Review Trends, Review Findings, Review Formats) with redundant UI prompts.
Presented 27 standalone ad-hoc output checkboxes on execution, creating choice paralysis for representatives planning basic calls.
The "Follow-Up" workflow block sat completely empty on the board as a blank placeholder frame, neglecting recordings and manual transcript indexing.
Locks scenario complexities dynamically (Light, Standard, Deep). Simple outreach scales down to 2 quick queries, saving token budgets and reducing runtimes.
Consolidates all briefing reviews, draft objections, and meeting formats onto a single dashboard screen, allowing representatives to make quick edits or prompts.
Standardizes output deliverables into semantic packages (Prep, Meeting, Follow-Up packs) mapped directly to scenario families.
Fires automated listeners to ingest raw Zoom/Google transcripts, extract commitments, log decisions, draft emails, and push Linear alerts in one loop.
Standardized Drive Schema Mapping: In Step 6 of Call Prep and Step 4 of Follow-Up, the system constructs structured client folders early in the pipeline. This mitigates naming collisions and connects assets directly back to CRM task dashboards.