AI-for-learning network map
Purpose: this map helps the founder choose where to spend time: which bigger, higher-margin problem in AI-for-learning to pursue, and which investors and builders to meet. Re-read it before taking networking meetings or making a Scribble Works direction call. Refresh it quarterly.
Each figure was checked against its cited page unless marked (unverified).
(a) Investors with recent AI-for-learning deals (one example each)
| Investor | Evidence (example portfolio) | Source |
|---|---|---|
| Owl Ventures | Led Cloudforce's $10M Series A (with Microsoft); over $2.2B in assets under management | Owl |
| Reach Capital | Closed a $265M Fund V for AI in learning, health and work. (Unverified: that a third of the platforms in Anthropic's Claude for Teachers launch are Reach portfolio companies) | citybiz |
| GSV Ventures | In Gizmo's $22M Series A (Apr 2026); co-led SigIQ.ai's $9.5M seed (Apr 2025) | PR Newswire, YourStory |
| Bain Capital Ventures | Joined MagicSchool's $45M Series B | MagicSchool |
| Valor Equity Partners | Led MagicSchool's $45M Series B | MagicSchool |
| Bessemer Venture Partners | Led Brisk Teaching's $15M Series A (Mar 2025) | PR Newswire |
| Insight Partners | Led SchoolAI's $25M Series A (Apr 2025) | GlobeNewswire |
| NextView Ventures | Invested in SchoolAI | NextView |
| Shine Capital | Led Gizmo's $22M Series A (Apr 2026) | TechCrunch |
| New Enterprise Associates (NEA) + AI Fund (Andrew Ng) | Back Kira Learning (kindergarten through grade 12 (K-12), AI-native); Ng is chairman | GlobeNewswire |
| Accel + OpenAI Startup Fund | Speak's $78M Series C at a $1B valuation (Dec 2024, slightly outside the window) | Speak |
| Trilogy Equity Partners | Led Wild Zebra's $6M round (Aug 2026; AI tutor for math and reading) | GeekWire |
| Craft Ventures | Led Starbridge's $42M Series A (AI for selling to government and education) | Starbridge |
| Gates Foundation (philanthropy, not venture capital) | $1B commitment, of which 40% ($400M) goes "toward education, including AI tutoring" | Fortune |
| Learn Capital | Published a 2025 outcomes report on AI and learning. (Unverified: no specific 2025 lead deal found) | PR Newswire |
Context: global edtech venture capital (VC) was $2.6B in 2025, and workforce companies took 38% of deal volume (HolonIQ). The 2021 peak was $16.7B (Tracxn figure, via Rest of World).
(b) Ambitious builders past worksheets
| Company / founder | Why connect | Source |
|---|---|---|
| MagicSchool (Adeel Khan) | Leads teacher-AI distribution; says 6M educators and 10K+ partner schools | MagicSchool |
| Brisk Teaching (Arman Jaffer) | Teacher AI delivered as a Chrome extension; 1M educators | PR Newswire |
| SchoolAI | Student AI spaces for classrooms, bought by districts | GlobeNewswire |
| Alpha School (MacKenzie Price; Joe Liemandt, unverified) | AI school model built on 2 hours of academics a day, growing to about 50 campuses. Houston Independent School District (ISD) runs its Timeback platform at 2 of the 9 schools in a $500K-per-school pilot. Vault context: [[06-reference/2026-02-25-moonshots-ep233-ai-tutors]] | Axios, Houston Public Media |
| Khan Academy / Khanmigo (Sal Khan) | Khanmigo is free for US teachers, paid for by Microsoft | Khan blog |
| Kira Learning (Andrea Pasinetti) | Full-stack, AI-native platform for K-12 districts | GlobeNewswire |
| Wild Zebra (Edan Shahar, Erik Selberg) | Closest analog to [[01-projects/printables-product/index|Scribble Works]]: grades 2-9 at $48/month per child. Also sells to private schools through E3n, the merger of the Educational Records Bureau (ERB) and the Enrollment Management Association | GeekWire |
| Ello (Tom Sayer) | AI reading coach for kindergarten through grade 3, sold to parents | TechCrunch 2023 |
| Amira Learning | AI reading tutor and assessment; says it is in 2,000+ districts | Tracxn (unverified) |
| Third Space Learning | Tutoring sold to schools; scaling Skye, an AI tutor that speaks with students about math; raised about $5.9M in Apr 2026 | New Market Pitch (unverified) |
| Gizmo (Petros Christodoulou) | Gamified AI study app with 13M learners; a consumer growth playbook to learn from | TechCrunch |
| Speak | AI voice tutor valued at $1B | Speak |
| Multiverse (Euan Blair) | AI workforce skilling; $2.1B valuation; first cash-positive quarter in Q1 2026 | Tech.eu |
| Starbridge (Justin Wenig) | Monitors public agencies and schools for bids, budgets and contacts for sales teams; Clever and Instructure are customers. The school-data plumbing is the same kind of data engineering phData sells | Starbridge |
| Primer / Prenda (microschools) | Microschool networks; Keith Rabois led Primer's Series A (figures may be dated) | Education Next |
(c) Margin lens: where the money flows
- Districts and states (the largest pool). US public schools spent $927B in 2020-21, in constant 2022-23 dollars, about $18.6K per pupil (National Center for Education Statistics). Federal Elementary and Secondary School Emergency Relief (ESSER) money has ended. Districts "are likely to cut vendors first, such as tutoring companies or afterschool programs" (April 2024 forecast, K-12 Dive). Louisiana's model tutoring contract makes 50% of its value depend on student outcomes (Louisiana Department of Education). FEV Tutor collapsed when the relief funds ended (The 74).
- Education Savings Accounts (ESAs) and vouchers. Private-choice programs allocated $8.2B (EdChoice 2025; data through Dec 2024) (EdChoice). Arizona's ESA has 92K students and about $1B in yearly spending (Common Sense Institute). A federal tax-credit scholarship starts 2027-01-01, and states must opt in (IRS).
- Employers. US training spend was $102.8B in 2025. Spending on outside products and services rose 29% to $16B (Training magazine).
- Parents (consumer). Khanmigo charges parents $4/month (Khanmigo). Wild Zebra charges $48/month (GeekWire, above).
Ray's read (analysis, unverified): The buyer ranking is employers first, then premium services that ESAs can pay for, then district software with outcome data, then consumers. No margin figures are cited; the ranking orders buyers by pool size and ease of sale. The consumer tier looks squeezed at the bottom. Scribble Works' proposed test price is $9 vs $12 per month ([[01-projects/printables-product/2026-09-11-business-opportunity-pricing-and-acquisition]]), close to Khanmigo's $4. Wild Zebra's list price is 4x our $12 arm (5.3x the $9 arm); no data yet on what parents pay.
(d) Three candidate bigger problems (Ray's analysis, unverified)
- Outcomes evidence for school AI. SchoolAI says more than 400 school districts use it (GlobeNewswire), and Louisiana's model contract pays on results. Gates says the AI tools it funds will be evaluated by independent researchers before they scale (Fortune). Instructure's LearnPlatform already offers "rapid-cycle evaluation" (Instructure), but a learning-management-system vendor that owns the evaluator is not a neutral judge of rival tools. Hypothesis: the open gap is a vendor-neutral layer that joins usage and assessment data across tools. This fits phData-style skills: data integration, warehouse modeling and governed student data are the core of the work.
- Infrastructure for ESA-funded AI microschools. Alpha is growing to about 50 campuses, and Houston ISD runs Timeback at 2 of the 9 schools in its $500K-per-school pilot (Axios; Houston Public Media, above). ESA dollars and the 2027 federal credit add fuel. The problem at founder scale is the operating system for small AI-first schools: compliance, ESA billing and outcome reporting. Scribble Works' offline, kids-on-paper stance could become one module of it.
- AI upskilling for employers. Employers spent $16B on outside training products and services in 2025, up 29% (Training magazine). Workforce companies took 38% of edtech deal volume (HolonIQ, above), and Multiverse is valued at $2.1B (Tech.eu, above). This is the closest fit to the founder's phData day job and the furthest from kids.
Open questions / next pass
- Check existing vendor-neutral data and evaluation layers (for example the Ed-Fi Alliance standard) before treating problem 1's gap as open.
- Get primary sources for each row marked (unverified). Use Form D related persons as a free proxy for cap tables.
- Map warm paths: which of these investors are in the RDCO or phData network (check 03-contacts).