JuxtMedia Agent Phase 3

1

Gmail Integration Scope for Phase 3

Skip identified Gmail access as the #1 unlock. The question is how deep: read-only scanning, read + activity tracking (match sends to leads), or full send capability from the app. Jonathan has Super Admin access to Juxt Media Google Workspace.

A — Read-Only Inbox/outbox scan only. See what emails exist but no automatic lead matching or activity tracking.

B — Read + Track Recommended Detect sends/receives, match to leads, populate email_activity table, power the morning summary report. No sending from the app.

C — Read + Track + Send Full email capability from the app. Adam can send directly from the agent. Adds complexity and risk given the 20% inaccuracy rate.

2

Morning Summary Format

The morning summary is the “sales coach” daily touchpoint — 3 categories (tracked deals, untracked deals, slip-through-cracks). Question: should it live in the app dashboard, arrive as an email digest, or both?

A — Dashboard View (In-App) Recommended Adam sees the morning summary when he logs in. Keeps him in the tool. 3-column card layout with actionable items.

B — Email Digest Sent to Adam's inbox each morning. No need to open the app. Risk: becomes another email to ignore.

C — Both Dashboard view + email digest. Maximum visibility. More to build but ensures Adam sees it regardless of whether he opens the app.

3

Stage Model — How Many Stages?

Skip said “4-5 max” but the full model has 8 stages (queued, researched, contacted, engaged, qualified, proposal, closed_won, not_a_fit). Need to decide how granular the pipeline should be for Adam’s daily workflow.

A — 8 Stages (Full Model) queued, researched, contacted, engaged, qualified, proposal, closed_won, not_a_fit. Maximum granularity, most to manage.

B — 5 Stages Recommended queued, researched, contacted, qualified, closed. Balances granularity with Skip's “keep it simple” directive.

C — 4 Stages new, active, proposal, closed. Simplest possible. May lack visibility into early pipeline stages.

4

Follow-Up Cadence

When a lead hasn’t been contacted within a threshold, it gets flagged as overdue. The question is whether to use a fixed default (simple) or let Adam/stages customize the timing.

A — Fixed 3-Day Default Recommended 3 days for all stages. Simple, consistent. Iterate based on Adam's actual usage patterns before adding complexity.

B — Configurable Per Stage Different cadence per stage (e.g., “contacted” = 3 days, “qualified” = 5 days). More accurate but more settings to manage.

C — Manual Per Lead Adam manually sets follow-up date for each lead. Maximum flexibility but highest manual effort.

5

Auto-Stage Advancement

When Gmail detects that Adam sent an email to a lead, should the tool automatically advance the stage (e.g., queued → contacted), or should Adam confirm? This is key to the “sales coach” pattern — the tool can detect behavior and suggest actions.

A — Manual Only Adam clicks to advance every stage. No automation. Simple but adds friction to daily workflow.

B — Fully Automatic Email detected = stage advances immediately. No confirmation. Risk: false positives from email thread noise.

C — Hybrid (Auto-Suggest, Adam Confirms) Recommended Auto-detect shows a notification: “Email sent to [lead] — advance to contacted?” Adam clicks to confirm. The “sales coach” pattern.

6

Sales Training Data Timing

Skip has Loom videos and Zoom transcripts from past Juxt sales calls. This content could train the agent on Juxt-specific sales patterns. But Phase 3 already includes Gmail + stages + email tone. How much should we take on?

A — Ingest Now (Phase 3) Ingest Skip's recordings as part of Phase 3. Adds scope but gives the agent Juxt-specific sales knowledge immediately.

B — Playbook Doc First Skip writes a “sales playbook” document. Lighter than video transcripts, structured, and faster to ingest.

C — Defer to Phase 4 Recommended Gmail + stages are enough for Phase 3. Training data is a Phase 4 unlock once the pipeline is stable and Adam is using stages daily.

7

Multi-Rep Foundation in Phase 3?

Skip’s vision: “3-4 freelance contractors doing commission-only sales outreach.” Full multi-rep requires user login, role-based access, and escalation flows. A lighter option: just add a “Send to Skip” button that validates the escalation pattern without full engineering.

A — Full Multi-Rep (Phase 3) Add user login + “Send to Skip” escalation flow. Validates the multi-rep model but significantly expands Phase 3 scope.

B — Defer + “Send to Skip” Button Only Recommended No login system. Just a “Send to Skip” button on qualified leads. Tests the escalation pattern without multi-rep engineering.

+

Google Workspace OAuth Confirmation

Skip: please confirm that jonathan@noboxai.com has the OAuth consent screen configured for Gmail API access on the Juxt Media Google Workspace. Jonathan was granted Super Admin access during the Mar 4 call — this confirmation ensures the API connection can proceed.

Confirmed — OAuth is configured, proceed The Google Workspace OAuth consent screen is set up. Jonathan can proceed with Gmail API integration.

Not Sure — Need to check Unsure if OAuth consent screen is configured. Jonathan will investigate and set it up if needed.

JuxtMedia Agent Phase 3 — Decisions

Decisions Summary