Carmee Pathway Coordination - AI Automation Assessment
Date: Feb 16, 2026
Context: ZTAG Pathways Ownership & Authority Governance document
Question: How much of Carmee's work can/should AI automate?
Executive Summary
Verdict: 60-70% of Carmee's pathway coordination work can be AI-automated (routine classification, quote generation, form filling). The remaining 30-40% MUST stay human (relationship building, judgment calls, strategic handoffs).
Key insight: Carmee's role is pathway stewardship, not sales execution. AI can handle stewardship logic (rules, eligibility, routing). Humans handle relationship cultivation and strategic escalation.
Recommended split:
- AI handles: Classification, eligibility checks, quote generation, form completion, template customization
- Carmee handles: First contact, relationship building, judgment calls, Kristin handoffs, exception handling
Task-by-Task Analysis
✅ HIGH AI SUITABILITY (Automate These)
1. Pathway Classification (Determine which pathway applies)
Current process: Carmee reviews inquiry, determines if it's Pathway 1 (Grant-Funded School), Pathway 2 (3rd-Party Nonprofit), or Pathway 3 (City/Municipal)
AI capability:
- Parse inquiry email/form
- Extract: Organization type, funding source, program structure
- Match to pathway rules
- Output: "This is Pathway 1 (Grant-Funded School) - EDU pricing applies"
Confidence: 95% - This is pure logic, no judgment needed
Implementation:
- Email → Gemini Flash classification ($0.0001/call)
- Decision tree: Public school district? → Pathway 1. 501(c)(3)? → Check supervision → Pathway 2 or STOP
- Output sent to Carmee for confirmation
Value: Saves Carmee 15-20 min per inquiry (currently ~10-15 inquiries/week = 2.5-5 hrs/week saved)
2. Eligibility Verification (501(c)(3), supervision, website checks)
Current process: Carmee manually verifies nonprofit status, checks if program is supervised, reviews websites for educational discount eligibility
AI capability:
- 501(c)(3) verification: Query IRS Tax Exempt Organization Search API (free)
- Website verification: Web scraping to check if org is for-profit vs nonprofit
- Supervision check: Parse program description for keywords (supervised, drop-off, rental, Playmaker)
- Output: "✅ Verified 501(c)(3). ✅ Supervised program. ❌ For-profit business detected → Professional pricing only"
Confidence: 85% - Automated checks work for 85% of cases, edge cases need human review
Implementation:
- API: IRS Tax Exempt Org Search
- Web scraping: Brave Search + web_fetch for homepage analysis
- Keyword detection: "drop-off", "rental", "supervised", "Playmaker"
- Flag edge cases for Carmee review
Value: Saves Carmee 10-15 min per 3rd-party inquiry (~5-7 inquiries/week = 1-2 hrs/week saved)
3. Quote Generation (EDU vs Professional pricing, bulk discounts)
Current process: Carmee calculates pricing, applies EDU discount or Professional pricing, applies bulk discounts, creates quote
AI capability:
- Input: Pathway classification, quantity, buyer type
- Logic:
- Pathway 1 → EDU pricing
- Pathway 2 → EDU if 501(c)(3) + supervised + in-school; else Professional + optional 10%
- Pathway 3 → EDU (pending confirmation)
- Bulk discounts: [need discount schedule]
- Output: Quote PDF or draft for Carmee approval
Confidence: 95% - Pure math and rules, no judgment
Implementation:
- Template: Quote generation script (pricing table → PDF)
- Bulk discount tiers (need to extract from existing quotes)
- Draft sent to Carmee for final approval before sending
Value: Saves Carmee 20-30 min per quote (~8-12 quotes/week = 3-6 hrs/week saved)
4. Template Customization (Grant language, funding proposals)
Current process: Carmee customizes funding proposals using approved templates
AI capability:
- Input: Grant type (ESSER, ELOP, Title funds), buyer info, quantities
- Template library: Approved funding packets
- AI fills in: District name, quantities, pricing, timeline, key requirements
- Output: Draft funding proposal for Carmee review
Confidence: 80% - Templates work well, but customization requires some judgment
Implementation:
- Library: Store approved funding packet templates (Word/PDF)
- AI fills blanks: District name, contact, pricing, timeline
- Carmee reviews before sending (human approval required)
Value: Saves Carmee 30-45 min per funding proposal (~3-5 proposals/week = 2-3 hrs/week saved)
5. Form Completion (Vendor onboarding, procurement paperwork)
Current process: Carmee completes vendor onboarding forms for city/municipal purchases
AI capability:
- Input: Vendor form (PDF or online)
- Extract: Required fields (company name, address, tax ID, contact, insurance, references)
- Auto-fill from ZTAG company profile
- Output: Completed form for Carmee review
Confidence: 90% - Most forms are standardized
Implementation:
- OCR: Extract form fields from PDF
- Auto-fill: ZTAG company database (name, address, EIN, insurance info, references)
- Carmee reviews and signs
Value: Saves Carmee 30-60 min per form (~2-4 forms/month = 1-2 hrs/month saved)
⚠️ MEDIUM AI SUITABILITY (Augment, Don't Replace)
6. Grant Language Coordination (Help decision-makers write grant applications)
Current process: Carmee provides grant language and coordinates with school decision-makers
AI capability:
- Generate sample grant language (ESSER, ELOP, Title funds)
- Suggest key talking points (SEL, STEM, inclusive play, equity)
- Draft email to decision-maker with grant language attachment
Confidence: 70% - AI can draft, but human relationship context matters
Why human-in-loop:
- Tone matters (formal vs casual, district culture)
- Relationship history (have we worked with this district before?)
- Strategic timing (when to push, when to wait)
Recommendation:
- AI drafts grant language + email
- Carmee reviews, personalizes, sends
- Saves 15-20 min per grant coordination (~3-5/week = 1-2 hrs/week saved)
7. Bulk Discount Negotiation (Structure district-wide implementations)
Current process: Carmee structures bulk pricing for district-wide rollouts
AI capability:
- Calculate bulk discount tiers
- Suggest multi-school package (e.g., "5 schools = 15% off, 10 schools = 20% off")
- Draft implementation timeline
Confidence: 75% - Math is easy, but judgment matters (how aggressive to discount?)
Why human-in-loop:
- Strategic pricing (don't leave money on table, but also don't lose deal)
- Competitive intelligence (what are competitors offering?)
- Long-term value (is this a foothold district for future expansion?)
Recommendation:
- AI calculates standard bulk pricing
- Carmee approves final discount level
- Saves 10-15 min per bulk deal (~2-3/week = 30-45 min/week saved)
❌ LOW AI SUITABILITY (Keep Human)
8. First Contact / Relationship Building
Current process: Carmee engages initial inquiry, builds rapport, understands buyer needs
AI capability: ❌ Very limited
Why this MUST stay human:
- Trust building - Buyers want to talk to a human, not a bot
- Nuance detection - Is this a tire-kicker or serious buyer? What's their urgency?
- Rapport - Warm introductions, personal connection, understanding school culture
- Unstructured needs - "We're looking for something to help with..." (requires listening + translation)
Recommendation: Keep 100% human. AI can draft responses, but Carmee sends them.
Time: 20-30 min per first contact (~10-15/week = 3-5 hrs/week)
9. Judgment Calls (When to escalate, when to stop, exception handling)
Current process: Carmee decides when to close pathway without escalation, when to escalate to Quan, when to involve Kristin
AI capability: ❌ Limited (can flag, but shouldn't decide)
Why this MUST stay human:
- Strategic exceptions - "This doesn't fit our rules, but it's a strategic partner" (requires business judgment)
- Relationship risk - "This buyer is difficult, should we walk away?" (requires emotional intelligence)
- Escalation timing - "When do I loop in Kristin?" (requires understanding of team dynamics)
Recommendation: Keep 100% human. AI can flag edge cases ("This doesn't fit Pathway 1 or 2, needs review"), but Carmee decides.
Time: 10-15 min per edge case (~3-5/week = 30-60 min/week)
10. Kristin Handoff (Strategic district-level rollout planning)
Current process: Carmee determines when to involve Kristin for district-level strategy (only if school agrees to monthly Playmaker commitment + needs district introduction)
AI capability: ❌ Very limited
Why this MUST stay human:
- Relationship readiness - Is the school ready for Kristin's strategic involvement? (requires feel)
- Timing sensitivity - Kristin is capacity-constrained, don't overload her (requires judgment)
- Strategic value - Is this district worth Kristin's time? (requires understanding of 10-year mission)
Recommendation: Keep 100% human. AI can flag potential Kristin handoffs ("School confirmed monthly Playmaker commitment, may be ready for district strategy"), but Carmee decides when to loop her in.
Time: 15-20 min per potential handoff (~1-2/week = 15-40 min/week)
Summary Table
| Task |
AI Suitability |
Time Saved |
Who Decides |
AI Role |
| Pathway Classification |
✅ High (95%) |
2.5-5 hrs/week |
AI + Carmee confirm |
Automate |
| Eligibility Verification |
✅ High (85%) |
1-2 hrs/week |
AI + Carmee edge cases |
Automate |
| Quote Generation |
✅ High (95%) |
3-6 hrs/week |
AI draft → Carmee approve |
Automate |
| Template Customization |
✅ High (80%) |
2-3 hrs/week |
AI draft → Carmee approve |
Automate |
| Form Completion |
✅ High (90%) |
1-2 hrs/month |
AI fill → Carmee sign |
Automate |
| Grant Language |
⚠️ Medium (70%) |
1-2 hrs/week |
Carmee (AI drafts) |
Augment |
| Bulk Discount |
⚠️ Medium (75%) |
30-45 min/week |
Carmee (AI calculates) |
Augment |
| First Contact |
❌ Low |
3-5 hrs/week |
Carmee |
Human only |
| Judgment Calls |
❌ Low |
30-60 min/week |
Carmee |
Human only |
| Kristin Handoff |
❌ Low |
15-40 min/week |
Carmee |
Human only |
Total time saved: 10-18 hrs/week (out of ~20-25 hrs/week total pathway work)
Carmee's remaining work: 7-10 hrs/week (relationship building, judgment, strategic escalation)
Net effect: Carmee can handle 2-3x more inquiries with same time investment, OR focus on higher-value strategic work
What SHOULD Be AI-Automated (Priority Order)
Phase 1: Quick Wins (Implement This Month)
- Pathway Classification - Email → AI → "This is Pathway 1" (2-5 hrs/week saved)
- Quote Generation - AI drafts quote → Carmee approves → Send (3-6 hrs/week saved)
Expected impact: 5-11 hrs/week saved, Carmee focuses on first contact + judgment calls
Implementation time: 4-6 hours (Matt Wolfe learning Week 2 could include this)
Phase 2: Deeper Automation (Next Month)
- Eligibility Verification - IRS API + web scraping (1-2 hrs/week saved)
- Template Customization - Auto-fill funding proposals (2-3 hrs/week saved)
Expected impact: Additional 3-5 hrs/week saved
Implementation time: 6-8 hours
Phase 3: Augmentation (Month 3)
- Grant Language Drafting - AI drafts, Carmee personalizes (1-2 hrs/week saved)
- Bulk Discount Calculation - AI suggests, Carmee approves (30-45 min/week saved)
Expected impact: Additional 1.5-3 hrs/week saved
Implementation time: 4-6 hours
What Should NOT Be AI-Automated
Keep 100% Human:
- First contact / relationship building - Trust, rapport, human connection
- Judgment calls - Strategic exceptions, when to escalate, when to stop
- Kristin handoff timing - Capacity-sensitive, strategic value assessment
Why: These require emotional intelligence, relationship context, strategic judgment that AI can't replicate.
Carmee's time on these: 4-6 hrs/week (irreducible human work)
Loss Function Alignment
Does this protect Carmee's role?
- ✅ Yes - Frees her from tedious classification/quote work
- ✅ Yes - Lets her focus on relationship building (her strength)
- ✅ Yes - Increases her capacity (can handle 2-3x inquiries)
Does this respect human empowerment?
- ✅ Yes - AI handles tedious logic, Carmee handles judgment + relationships
- ✅ Yes - Carmee's authority isn't reduced, it's amplified (more time for strategic work)
- ✅ Yes - This is AI as tool (like Matt Wolfe), not AI as replacement
Does this accelerate escape velocity?
- ✅ Yes - Pathways run more autonomously (less Quan involvement)
- ✅ Yes - Carmee can scale without hiring another person
- ✅ Yes - Routine decisions automated, strategic decisions human-led
Integration with Matt Wolfe Learnings
This fits perfectly into Matt Wolfe implementation plan:
Week 2 (Feb 25-28): Build Carmee pathway automation alongside Fathom todo extraction
- Tuesday: Pathway classification logic (2-4 hours)
- Wednesday: Quote generation automation (2-4 hours)
- Total: 4-8 hours (fits within existing Week 2 calendar blocks)
Week 3 (Mar 7): Eligibility verification + template customization
- Additional 4-6 hours (can add a separate block or extend Council briefing day)
Week 4 (Mar 11-14): Integration testing
- Test with real inquiries, validate accuracy, measure time saved
Recommended Next Steps
Immediate (This Week):
- Extract 10-15 sample inquiries from Carmee's recent work
- Document current pricing rules (EDU vs Professional, bulk discount tiers)
- Collect approved funding proposal templates
Week 2 (Feb 25-28):
- Build pathway classification tool (email → pathway determination)
- Build quote generation automation (pathway + quantity → quote PDF)
- Test with sample inquiries, validate accuracy
Week 3 (Mar 7):
- Add eligibility verification (IRS API, web scraping)
- Add template customization (auto-fill funding proposals)
Week 4 (Mar 11-14):
- Integration testing with Carmee (real inquiries)
- Measure time saved, adjust rules/logic
March 12 (Carmee's 6-Month Review):
- Report: "Carmee now handles 2x inquiries with AI support, time saved = 10-15 hrs/week"
- Graduation: Carmee's authority confirmed, AI tools validated
Cost Analysis
AI costs (estimated):
- Pathway classification: Gemini Flash ~$0.0001/inquiry × 50/month = $0.005/month
- Quote generation: Gemini Flash ~$0.0001/quote × 40/month = $0.004/month
- Eligibility verification: IRS API free, web scraping ~$0.01/check × 20/month = $0.20/month
- Template customization: Gemini Flash ~$0.001/template × 15/month = $0.015/month
Total AI cost: ~$0.25/month (negligible)
Time saved: 10-18 hrs/week × $25-30/hr (Carmee's effective rate) = $250-540/week = $1000-2160/month
ROI: ~10,000x
Final Verdict
How much CAN you automate? 60-70% of Carmee's pathway work
How much SHOULD you automate? 60-70% (automate logic, keep human judgment)
What stays human?
- First contact (trust, rapport)
- Judgment calls (strategic exceptions, escalation timing)
- Kristin handoff (relationship readiness, capacity sensitivity)
Net effect:
- Carmee handles 2-3x more inquiries with same time
- OR: Carmee focuses on higher-value strategic work (district relationships, complex deals)
- Escape velocity: Pathways run autonomously, less Quan involvement
This is exactly the AI-leveraged scaling model you're building.
Start with Phase 1 (classification + quote generation) in Week 2 of Matt Wolfe plan.
Owner: Minnie
Next: Extract sample inquiries, document pricing rules, schedule build time