A Player Labs × Delta Gymnastics × Tribe

Three loops. One engine.

A Player Labs
A Player Labs
THE ENGINE Hypothesis Hunter Closes the loop. Surfaces what nobody asked for. PILLAR 01 Lead Nurture PILLAR 02 Member Support PILLAR 03 Manager Intel SIGNALS → ← INSIGHTS

Each pillar closes its own loop. The engine makes them intelligent.

01
Architecture

Tribe inside. Python outside. One bridge.

70/30. Apex orchestrates. Python explores. Headless 360 (TDX, 15 Apr) is the bridge.

Channels
Website chatAgentforce widget — Pillar 1
Parent PortalAuthenticated — Pillar 2
Tribe internal UIManager + coach — Pillar 3
Agentforce Contact CenterVoice — post-conference
Layer 01 Data Cloud
Unified profile Salesforce CRM Marketing Cloud Engagement Xero (2027) Deputy (2027) Meta CAPI via Data 360
↕ governed reads & writes
Layer 02 — the heart Salesforce Tribe org
AI_Recommendation AI_Risk_Score AI_Insight Knowledge Articles Topics + Actions Flows Apex (dispatcher + receiver) Einstein Trust Layer
↕ Headless 360 — MCP tools + Agent API + Platform Events
Layer 03 — the brain External Python service
Hypothesis Hunter Autotelic loops Statistical validation 7-year historical scan Pattern → Insight writeback
Adjacent
Marketing Cloud GrowthIn flight — Girikon delivering
AgentforceTopics, Actions, Atlas Reasoning
Headless 36015 Apr 2026 · TDX
Service Cloud VoiceReplaced by ACC
Closed loop Risk Score → Flow → Case → Outcome → Learning store → Next investigation
02
How the engine learns

The discovery loop.

Self-rewarding. Self-improving. No sycophancy.

AUTOTELIC Discovery Runs nightly. Compounds weekly. 01 Observe 7 years of governed Tribe data 02 Hypothesise Agent proposes 50 03 Validate Python statistical tests 04 Surface AI_Insight in Tribe 05 Learn Outcome → memory

Agents are a dime a dozen. The loop that learns from your data is the moat.

03
What only A Player Labs ships

Salesforce ships scripted agents. We ship discovery agents.

Agent proposes 50 hypotheses. Python validates. Only what survives surfaces.

01 · Operational
Mon/Tue classes drop 47% more than Wed–Fri. Trigger: 2 weeks before school holidays.
n=2,800 p<0.001
02 · Familial · Phase 0 candidate
When one sibling withdraws, 73% of remaining siblings drop within 8 weeks. Independent of NPS or attendance.
n=410 pairs 3.4× baseline
03 · Developmental
Zero schedule changes in first 16 weeks → 1.4× faster milestones, 38% lower Y1 drop.
n=1,800 p<0.01

★ Illustrative shapes. Phase 0 spike validates Pattern 02 on real Delta data.

Python → Headless 360 → Tribe (governed read) → Validate → Platform Event → AI_Insight → Agent surfaces
04
Engagement

What A Player Labs ships.

Three phases · two ripcord gates · 69–85 hours over 9 weeks.

Phase 1 · Validate · Weeks 1–2
Validate
20–25 est. hrs
Spec 4–5h · Spike 12–14h · Sandbox 1–2h · Demo script 3–4h

The architecture survives Rakesh review and the data supports the thesis.

Done when
  • Tech spec signed off by Rakesh
  • At least one pattern at p<0.05 in real Delta data
  • Sandbox Hello-World through Trust Layer with audit log
  • Demo script v0 + risk register
What Delta walks away with
A pattern validated on Delta's own data. A signed-off tech spec. A working sandbox. The Python codebase that found the pattern — Delta's IP from day one.
Phase 2 · Engine · Weeks 3–5
Engine
27–32 est. hrs
Hunter production 24–28h · Ops tooling 3–4h

The autotelic discovery engine running on real data, nightly.

Done when
  • Hunter runs nightly <60 min · 3 patterns surfacing AI_Insight
  • Headless 360 in prod (REST fallback proven)
  • Failover validated — pillars keep working if Hunter offline
  • Ops tooling: trigger-on-demand, structured logs, alerts
What Delta walks away with
An autotelic discovery engine running on 7 years of Delta's data. Validated patterns surfacing nightly inside Tribe. Even before agents ship — the moat is real.
Phase 3 · Demo · Weeks 6–9
Demo
22–28 est. hrs
Topic designs 9–12h · KA toolkit 3–4h · Orchestration + dry runs 10–12h

Three live demos at Niagara Falls on real Delta data.

Done when
  • All 3 Topics pass the Coach-at-the-Gate test
  • Each Topic stands alone if one degrades
  • 3 full end-to-end dry runs cleared
  • Naomi's Knowledge Articles pass agent retrieval
What Delta walks into Niagara Falls with
Three Topics live on Delta's real data. Hypothesis Hunter surfacing patterns no Salesforce competitor in the room can replicate.
🪂 Gate 1
After Phase 1
decide whether to commit Phase 2
🪂 Gate 2
After Phase 2
decide whether to commit Phase 3
How we work
Hourly · weekly ledger · A Player Labs prime
All work lives in your Tribe org.
05