Project 01 · Cloud Data · Marketplaces
Lakehouse Marketplace: governed data exchange across clouds.
Premise: every Fortune 1000 enterprise runs 50–500 ongoing data exchanges; median time-to-production is 14 weeks at $80K–$300K per exchange per year. Platforms like Snowflake Marketplace, AWS Data Exchange, and Databricks Marketplace solved discovery; they have not solved the four hard problems that follow: cross-cloud federation, regulated-data clean rooms, producer-side monetization, and (newly urgent) agent-aware policy enforcement at query time.
Turns data sharing into a six-minute listing: usage-based metering at the credit-consumption level, clean-room privacy by default, cross-cloud federation as the architectural default, and a query-time policy plane that lets producers grant AI agents the same access as humans without losing row/column-level controls or lineage. Producer console treats every dataset as a SKU; consumer experience is "subscribe in 30s, query in 5min." Every agent query carries a signed audit record back to the data owner: the policy layer is the moat that makes regulated-data sharing safe enough to scale. North-star: compute revenue attributable to shared listings ($M / quarter). Liquidity counters: % listings with > 5 active subscribers, time-to-first-query, clean-room workloads / quarter, and policy-violation incidents per million queries (audited monthly). Wedge: convert customers' largest in-flight bilateral data exchanges into platform-listed SKUs.
Concept for
Snowflake
Databricks
AWS
Microsoft
Sharing · Clean rooms · Governance
$2.9B ARR ceiling
Project 02 · D2C Insurance · Post-sale Lifecycle
Lifeline: the customer account as a renewable product surface.
Premise: D2C insurers spend $180–$420 fully-loaded CAC per bind, hold the customer ~5.4 years on average, and convert < 6% into a second product over the lifetime. Every life event that should change coverage (marriage, child, mortgage, salary jump, divorce, claim) is observable in the customer's life and invisible to the carrier; the customer account UI is a PDF download and a billing date. Cross-product attach is a marketing-budget line item, not a product surface; claims, the moment of highest emotional bandwidth, are run by ops as a cost center rather than by product as a growth event.
Lifeline turns the post-sale surface into a product owned by Product, not Marketing. Three surfaces, one event spine: the customer account shows current coverage against detected life context with one-tap adjust within risk class; the CS console shows the same context with a drafted next-best-action, audit-trailed; the claim flow is instrumented as the highest-NPS moment in the product. Open-banking, payroll, and customer-declared signals enter as signed envelopes; an immutable audit ledger doubles as the experimentation substrate, regulator-grade. North-star: net revenue retention on the existing book, audited monthly against a hold-out cohort: NRR > 118% from book actions alone. Counters: coverage-gap precision > 92%, cancel-save rate > 22% (catches nudge fatigue), claim-NPS > 65.
Concept for
PolicyMe
Wealthsimple
Lemonade
Ladder
Lifecycle · Retention · Cross-sell
~$118M / yr per-carrier uplift
Project 03 · Search & Ads · Agentic Commerce
Atlas: when search becomes action.
Premise: search ads rest on three breaking assumptions: that a human reads the SERP, that a click is a meaningful unit of intent, and that publishers get a share of the upside. By 2027–28, projected ≥30% of high-intent commercial sessions will be at least partially agent-mediated (Gartner / eMarketer agent-commerce forecasts); ad and publisher revenue from those sessions is currently zero.
Atlas is a parallel ad platform: human SERP (existing inventory, defended) and agent inventory (a new auction over agent-readable units). Agents return cryptographically-signed receipts for downstream conversions; verified attribution becomes a contractual artifact, not a probabilistic model. Publishers whose content informed an agent's answer collect a defined revenue share (a 35% pool, daily-granular, exposed in a publisher console). North-star: verified agent-mediated conversions / week. Counter: brand-safety incidents per million impressions, capped at 1.5. Sequencing: publish the open Agent Ad Protocol before the bidder ships; own the standards layer first, the auction second.
Concept for
Google
Microsoft (Bing)
OpenAI
Anthropic
Adobe
Ads · Agentic commerce
~$12B / yr stranded value
Project 04 · Content Commerce · API Platform
Pier: the Open API layer for content commerce.
Premise: TikTok Shop is the fastest-growing surface in content commerce, ~$20B GMV in 2024 across the US and SEA, but the seller integration layer is years behind Shopify's. A Shopify Plus merchant who wants to test TikTok Shop today rebuilds their catalogue manually (4+ hours, < 60% coverage on first attempt), wires fulfillment routing by hand, and discovers that 3P seller tools (Yotpo for reviews, Rithum for multichannel inventory, Zendesk for support, Klaviyo for marketing) each have piecemeal API access with no shared partner contract. Shopify's moat is its 8,000+ app marketplace; TikTok Shop has no equivalent.
Pier is a tiered Open API + a governed Partner App Marketplace. Tiered access (Lite for trial sellers, Standard for catalogue sync, Premium with paid SLA for high-volume) over a single REST + webhook contract. Bidirectional catalogue sync with deterministic conflict resolution at field level (price, inventory, content-moderation tags, listing status). Partner App Marketplace where 3P tools publish apps with declared permission scopes; sellers install in clicks. A webhook-driven audit ledger doubles as the compliance substrate (TikTok-specific moderation rules) and the experimentation substrate. North-star: seller onboarding time (sign-up to first SKU live), under 1 hour. Counters: catalogue assortment coverage rate > 90%, fulfillment success rate > 97%, brand-safety incidents per million SKUs capped at 1.5. Wedge: 50 design-partner Shopify Plus sellers + 5 anchor 3P tools at launch (Yotpo, Rithum, Zendesk, Klaviyo, ShipStation).
Concept for
TikTok Shop
Shopify
BigCommerce
Amazon Multichannel
API Platform · 3P Integration · Content Commerce
~$4–8B GMV uplift potential
Project 05 · Financial AI · Local LLM Systems
Earnings Call Analyzer: private financial intelligence from public transcripts.
Premise: public companies file 4 earnings calls a year, over 2,000 S&P 500 calls annually, each 15,000–25,000 words of dense forward guidance, risk disclosures, management tone, and KPI prints. Today the options are bad: manual analyst review (2–4 hours per call, subjective, can't scale across a portfolio), expensive analyst-product subscriptions (Bloomberg, FactSet, AlphaSense), or cloud-API LLMs that send market-sensitive transcripts to a third party. Each option fails on at least one of speed, cost, privacy, or scale.
A working local-first AI dashboard: ingest a transcript (.txt or .pdf) and return a structured executive view in ~60 seconds on a CPU-only 8 GB-RAM laptop. Output: executive summary with green / amber / red sentiment badge, KPI cards (sentiment, confidence, theme count, risk count), theme-importance bar chart, risk-severity heat chart, bull / bear case decomposition, and evidence quotes with tag pills. Sidebar filters (tag, severity, confidence) re-render the dashboard live, without re-running the model. Stack: Streamlit + Ollama running Qwen2.5 7B-Instruct + Python PDF parser + JSON-schema constrained decoding + Plotly. The product point is the local-first tradeoff: cloud APIs win on speed (~5s vs ~2–4 min) and reasoning depth, but local inference wins on privacy (transcript never leaves the machine), zero marginal cost, offline reliability, and full prompt / schema control. North-star: time-to-insight per call. Counters: summary accuracy, hallucination rate, schema-drift incidents, analyst-override rate.
Concept for
Bloomberg
FactSet
S&P Global
AlphaSense
Morgan Stanley
Snowflake
Local LLM · Financial AI · Privacy-first AI
~60s time-to-insight · $0 per call
Project 06 · Fintech SaaS · Agentic AI
Copilot Ledger: agentic bookkeeping for SMBs.
Premise: the ~35M US small businesses (SBA 2024) spend ~$154B / yr on bookkeeping and adjacent compliance. LLMs can plausibly automate 80% of the mechanical layer, but every entrant has lost trying to sell automation to the SMB owner: routing around the accountant (the actual buyer with professional liability exposure).
Copilot Ledger sells the AI to the accountant, not around them. Every transaction is auto-classified with a published confidence score; low-confidence entries route to an approval queue triaged by materiality; every entry, auto-approved or human-approved, is signed for a regulator-grade audit trail. North-star: hours saved per accountant per week. Counter-metric: error rate on auto-approved entries, audited monthly against a held-out human-labeled set, capped at 0.4%. Channel-led GTM: managing partners of mid-size firms (20–200 SMB clients) are the buyer; the SMB is the workload. The accountant channel becomes the moat and the regulatory shield.
Concept for
Intuit
Microsoft
Xero
Sage
Agentic AI · Compliance
12+ hrs / accountant / wk
Project 07 · Cloud FinOps · Enterprise SaaS
Aperture: AWS cost intelligence for multi-account orgs.
Premise: the average Fortune 500 enterprise spends $90M–$400M annually on AWS distributed across 200–2,000 accounts; 25–40% of that spend is recoverable. AWS's own tools work in single-account contexts and become unusable above 100 accounts. Existing FinOps vendors built dashboards for a finance team that no longer exists. The actual users are a triad: FinOps engineer, Cloud CoE architect, and CFO finance partner.
Aperture is workflow-shaped, not dashboard-shaped. Each role gets the workflow they actually run (anomaly triage, RI/SP commitment-cycle planning, monthly chargeback close), with cross-account scope as a primitive, not an enterprise-tier upsell. Customer-zero is Aperture's own AWS bill, instrumented and self-attributed. The product disappears into the rituals already running on the customer's calendar. North-star: annualized $ saved per customer, audited against a customer-attested baseline. Counter: false-positive rate on anomaly alerts < 12%; allocation coverage > 95% by month six.
Concept for
Amazon (AWS)
Microsoft (Azure)
Google Cloud
FinOps · Enterprise SaaS
> $4M / customer / yr saved
Project 08 · Capital Markets · Trader Tools
Tempo: a sub-100µs order-routing console for multi-venue equities.
Premise: the trader workstation hasn't changed shape since the Bloomberg Terminal of the early 1980s. For HFT and stat-arb desks, decision latency is the largest unsolved cost in execution edge, and almost all of it lives in the UI, not the wire.
Tempo redesigns the front-end as a co-designed surface of the OMS: every interaction has a published microsecond budget, every render is deterministic, and post-trade reconstruction is signed by gesture. The latency budget is treated as a contractual SLA, not a brag, instrumented at the kernel via eBPF and exposed in customer dashboards. North-star: P50 keystroke-to-fill latency. Counter-metric: post-trade reconstruction MTTR (sub-20s, T+1, audit-grade). Wedge: small-team prop desks (10–50 seats); compliance becomes a buyer, not a blocker.
Concept for
Jane Street
Citadel Securities
Two Sigma
IMC
HFT · Market Microstructure
~$14–22M / yr slippage avoidance
Project 09 · Edge AI · Developer Platforms
EdgeForge: GPU inference at the world's doorstep.
Premise: AI inference is centralized in three regions; users in São Paulo, Mumbai, and Lagos pay 220–340ms of round-trip tax per token. Hyperscalers won't fix it (edge GPU economics are worse than dense regions). The CDNs can fix it but historically lacked GPU access. Whoever bundles silicon, model registry, and routing first wins the developer.
EdgeForge is the inference platform where the developer never picks a region. Upload a model, set a latency budget; the platform places each request across edge / regional / centralized GPU pools. Pricing is per request with a public price card and SLA per latency tier. Developer-experience moves: one-command deploy, LoRA adapters as first-class, token-level latency tracing in the same view as cost-per-1M-tokens. North-star: paid inference requests / week. Counter: P99 cold-start < 600ms; routing-tier SLA met > 95%. The bundle is the product: silicon partnership, CDN topology, and registry surface in one developer flow.
Concept for
Cloudflare
NVIDIA
Fastly
Vercel
Edge · GPU economics
P99 first-token < 80ms
Project 10 · Creator Economy · Marketplaces
Loop: creator monetization that compounds.
Premise: mid-tail creators (50K–2M followers) are the economic engine of every short-form video platform, and the largest unforced error. Median earnings sit at $400–$2,800 / month with a coefficient of variation north of 0.7. They churn not from low reach but from financial-planning collapse; quarterly churn runs 18–25%, and platforms spend 2–4× LTV re-acquiring replacement supply.
Loop treats creator earnings as a financial product, not an ad-revenue residual. A stable monthly subscription floor is funded by viewer subscriptions; performance ads are layered on top; per-creator algorithmic transparency is exposed in a creator console where the predicted next-month floor is committed to within ±10%. Brand-safety AI matches advertisers to creator inventory at the content-embedding level: defensible to a CMO, tolerable to the creator. Recommender-system PM craft: creator-side, viewer-side, and advertiser-side metrics balanced as a single objective. North-star: median mid-tail subscription floor. Counter: brand-safety incidents / 1M impressions; floor-prediction accuracy > 92%.
Concept for
Meta
TikTok
YouTube Shorts
Snap
Marketplaces · Recommender systems
~$3.4B / yr churn protection
Project 11 · Industrial AI · Energy Operations
Wellhead: industrial AI for upstream operations.
Premise: upstream oil & gas operates > 1.2M producing wells globally; unplanned downtime costs the industry $50–80B / yr. Pure-ML approaches have largely failed because reservoir engineers reject any tool that asks them to trust an output without a physical mechanism. Methane-intensity regulation (EPA NSPS OOOOb/c, EU Methane Regulation, OGMP 2.0) is a parallel forcing function with a calendar deadline.
Wellhead is physics-informed by construction and edge-deployed by default. Predictions arrive with auditable mechanism explanations (mass balance, multiphase flow regime, equipment kinematics) that an engineer can defend in a morning meeting. ESG measurements are signed and lineage-tracked, exportable in EPA OOOOb/c / EU Methane / OGMP 2.0 formats; they are the artifact filed with regulators, not an input to a separate report. The runtime operates through 30+ days of disconnected operation; cross-operator federation happens via aggregated gradients, never raw telemetry. North-star: unplanned downtime hours per asset / quarter, audited against a pre-deployment baseline. Wedge: ESP failure prediction at one major operator's brownfield asset.
Concept for
Saudi Aramco
ExxonMobil
Shell
Occidental
Industrial IoT · ESG
25%+ downtime reduction