JuxtMedia Agent Phase 3 — Priority Briefing

Date: 2026-03-05 From: Jonathan (Chief Architect, noboxAI) To: Skip Blankley & Adam Howard / Juxt Media
MVP live • Phase 2 deployed • 7 decisions needed for Phase 3

Phase 3: From Research Tool to Sales Coach

The JuxtMedia Sales Agent MVP is saving Adam ~80% of manual research time, but he has no visibility into deal progression, no follow-up tracking, and no proactive guidance. Skip's vision: evolve the tool from a research assistant into a sales coach that nudges Adam on consistency, surfaces forgotten leads, and provides daily guidance — scalable to multiple commission-only reps.

Skip (Mar 4): “Giving it access to my email account was the biggest fucking unlock.” Gmail integration is the #1 strategic priority. Everything else builds on email visibility.
Time Savings
~80%
Manual research eliminated for Adam
Enhancements
3
Email Tone • Stage Pipeline • Gmail
Decisions Needed
7
Gmail scope, stages, cadence, auto-advance, training data, multi-rep
Accuracy Rate
~80%
20% inaccuracy prevents fire-and-forget

Items Ready to Build Now

These enhancements have no external dependencies. Jonathan can start immediately.

C
Email Tone Refinement
No blockers • Ships first

Adam reported email drafts come across as “too negative or arrogant.” He currently uses 3 mental categories when crafting emails manually. The agent should match these. Root cause: agent references content from related entities (separate YouTube channels, unrelated social accounts) creating misleading context.

3 Tone Presets (Matching Adam's Categories)

PresetDescriptionUse Case
Professional PeerBusiness-to-business, industry-knowledgeable, data-informedCold outreach to established companies
Creative CollaboratorPartnership language, shared creative visionCreative industry leads, past clients
Trusted Local ResourceWarm, community-oriented, helpfulLocal Atlanta businesses, referrals

Key Changes

  • Reframe content gaps as opportunities not deficits: “Your brand has a strong event presence — video coverage could amplify that reach” vs “You don't have event videos”
  • Add tone selector dropdown to DraftEmailModal (default: Professional Peer)
  • Filter context: exclude related-entity data from email generation
  • Tone persists per lead so redrafts maintain selection
Email Draft Flow with Tone Selection
B
Deal Stage Pipeline
No blockers • Foundational for Gmail

Jonathan identified stage management as foundational: “Everything flows from a stage. His visibility should be very affected by what’s active.” Adam specifically requested stage toggles and last-contact sorting. Skip: “Keep stages to 4-5 at most.”

Proposed Stage Model

queued researched contacted engaged qualified closed_won / not_a_fit
StageDescriptionTrigger
queuedImported, not researchedAuto (import)
researchedResearch completeAuto (job completion)
contactedFirst outreach sentAuto (email detection) or manual
engagedReply received / meeting scheduledAuto (email detection) or manual
qualifiedConfirmed fitManual (Adam)
proposalProposal sentManual
closed_wonDeal closedManual
not_a_fitDisqualifiedManual

UI Requirements

  • Stage displayed prominently at top of LeadDetailView
  • StageSelector dropdown component with color-coded badges
  • Stage column in LeadsView data table with sort/filter
  • Visual indicators: red for overdue, amber for approaching due date
  • Overdue detection background job (next_contact_date < NOW())
Stage Pipeline Flow
Implementation Timeline
A
Gmail Integration + Morning Summary
Blocked on scope decisions • Skip's #1 priority

Skip identified this as the biggest unlock. The morning summary transforms the agent from a reactive tool into a proactive sales coach. Jonathan has Super Admin access to Juxt Media Google Workspace (granted live during Mar 4 call).

Morning Summary — 3 Categories

Category 1
Tracked Deals
Leads in the agent that need action today. Sorted by urgency — overdue follow-ups first.
Category 2
Untracked Deals
Email activity not reflected in the agent. Potential new leads from inbox. “Add to Agent?” action.
Category 3
Slip-Through-Cracks
Leads emailed once then forgotten. No follow-up within threshold. The behavioral nudge.
Gmail Analysis → Morning Summary Pipeline
Adam (Mar 2): “There are some times where I'll get a lead and I may email them, and then a week goes by and I completely forgot about it.” — This is exactly what Category 3 catches.

7 Decisions for Skip & Adam

These decisions are blocking Phase 3 engineering. Each has a recommended option based on the synthesis of both calls.

1. Gmail Scope for Phase 3
➤ Recommended: B — Read + Track

A) Read-only (inbox/outbox scan)

B) Read + track (detect sends, match to leads)

C) Read + track + send (full email from app)

Sending from the app adds significant complexity and Adam isn't ready for fire-and-forget given the 20% inaccuracy rate. Read + track gives full morning summary capability without send risk.

2. Morning Summary Format
➤ Recommended: A — Dashboard View (in-app)

A) Dashboard view (in-app BriefingView)

B) Email digest (sent to Adam's inbox)

C) Both (dashboard + email)

Dashboard view keeps Adam in the tool. Email digest as Phase 4 addition once daily usage is established.

3. Stage Model — How Many Stages?
➤ Recommended: B — 5 Stages

A) 8 stages (full model: queued, researched, contacted, engaged, qualified, proposal, closed_won, not_a_fit)

B) 5 stages (queued, researched, contacted, qualified, closed)

C) 4 stages (new, active, proposal, closed)

5 stages balances granularity with Skip's “keep it simple” directive. Enough to distinguish pipeline position without overwhelming Adam.

4. Follow-Up Cadence
➤ Recommended: A — Fixed 3-Day Default

A) Fixed 3-day default for all stages

B) Configurable per stage

C) Adam manually sets per lead

Fixed 3-day default for MVP. Iterate based on Adam's actual usage patterns before adding complexity.

5. Auto-Stage Advancement
➤ Recommended: C — Hybrid (auto-suggest, Adam confirms)

A) Manual only (Adam clicks)

B) Auto-detect from email activity (fully automatic)

C) Hybrid (auto-detect shows notification, Adam clicks to confirm)

Hybrid gives Adam visibility into email-detected events while keeping him in control. The tool suggests, Adam decides. This is the “sales coach” pattern.

6. Sales Training Data Timing
➤ Recommended: C — Defer to Phase 4

A) Ingest Skip's Loom/Zoom recordings now

B) Skip writes a “sales playbook” doc for ingestion

C) Defer to Phase 4

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?
➤ Recommended: B — Defer, add “Send to Skip” button only

A) Yes — add user login + “Send to Skip” escalation flow

B) No — defer full multi-rep. Plant the seed with a “Send to Skip” button only.

Phase 3 is already substantial. The “Send to Skip” button validates the escalation pattern without full multi-rep engineering.

Complete the input form to capture these decisions in a structured format. Responses will be integrated directly into the Phase 3 engineering plan.

Phase 4 Items (Activate After Phase 3)

ItemActivates WhenCurrent Bridge
Video Asset Detection FixAfter email tone ships (same codebase)Adam works around 20% inaccuracy with manual review
Sales Training Data IngestionAfter Gmail + stages are stableSkip has Loom/Zoom content ready; pipeline doesn't exist yet
Multi-Rep / Commission RepsAfter Phase 3 proves “sales coach” modelSkip's vision: 3-4 freelance contractors doing commission-only outreach

Phase 3 → Phase 4 Transition Gates

Cross-References

Mar 2 Feedback SessionAdam + Skip tool evaluation, 41 minutes
Mar 4 Strategy SessionSkip + Jonathan — Gmail vision, sales coach direction
Phase 3 Synthesis & HandoffCross-call synthesis with ADW spec
ADW Handoff SpecFull engineering spec for all 3 enhancements