Agentic Demand Index · v3 · 8 pulses · multi-basket value-capture-layer architecture
Agentic Demand Index
Where are agentic AI workloads emerging? Eight operational pulses — each measuring a different value-capture-layer (App Builders · Tollbooths · Infrastructure · Endpoints · Security · Industrial) with its own multi-basket equity-gaps. The hardware buildout (CCST · TAISP · MHP · STP) tells you what supply looks like. These pulses tell you whether demand is showing up to consume it — and which value-capture-layer reacts first.
Agentic Demand Index — equal-weighted average of the 10 live pulses below (10 of 10 reporting). Each pulse counts once; a pulse with no current reading drops out rather than being filled in.
Share of 70,000+ US businesses with a paid AI transaction in June 2026, per the Ramp AI Index — actual card + bill-pay spend, not market pricing. The pulses above read how markets price agentic demand; this reads whether businesses are actually paying for it. Divergence between the two is the signal. Feeds EAP as its M4 · Ramp-Adoption module. Vendor split: Anthropic 42.4% · OpenAI 39.5% · Google 6.4%.
Open forward-deployed engineering roles across 6 AI value-chain companies (OpenAI, Anthropic, Databricks, Palantir, Scale, Sierra), counted daily from their public job boards — 328 including deployment/applied-AI roles, 12.28% of all their open positions. The demand pulses read how markets price agentic uptake; FDE hiring reads whether vendors are staffing to deploy it. Full read on the Forward Deployment Pulse.
Share of the 8.8 trillion tokens routed through the open model market (2026-08-08) that ran on Chinese frontier models, from OpenRouter's public usage data — what builders actually run, not what markets price. Month 64.3% → week 61.2% → day 65.4% is the acceleration read; European models carry 0.3%. Full four-bucket split (US · China · Europe · RoW), top models and method on the Token Split Pulse.
69 points · since 2026-06-02
Enterprise Infrastructure Pulse
AI Factory: GPU · Server · Network · Power
Enterprise Workflow Pulse
SaaS Agents: Workflow · Tools · Observability
Consumer Cloud Pulse
Cloud Agents: Tokens · Apps · API
Enterprise Application Pulse
App-Developers: 5 sub-baskets, vertical splits
Consumer Endpoint Agent Pulse
Devices: Smartphones · AI-PCs · Wearables · XR
Industrial / Physical Agent Pulse
Physical: Factory · Robotics · FPGA · Auto
Tollbooth Pulse
Monetization: Cloud + Neo-Cloud + API + Data Tollbooth
Agentic Security Pulse
Security: Identity · Endpoint · Network · Data Sec
Forward Deployment Pulse
Supply side: FDE hiring across the value chain
Token Split Pulse
Model demand: CN vs US token usage (OpenRouter)
How the family is structured
The headline. The Agentic Demand Index is the equal-weighted average of the live pulse composites — each pulse counts once, and any pulse not currently reporting is dropped and the rest renormalised (never a neutral placeholder). It reads the breadth of agentic demand across the eight value-capture layers and moves as the pulses below move.
Each pulse runs as its own Cloudflare worker on its own cadence (CCP hourly, EWP weekdays, EIP daily, EAP weekly, CEP & IAP monthly). Each writes a composite score 0-100 plus per-module breakdown to KV. The composite reads the momentum regime of demand: ≥70 = Surging, 40–69 = Scaling, <40 = Emerging — how hard demand is moving, not how far it has spread (a high reading can precede broad adoption by years). Once enough history accumulates, a falling trend shows as a cooling arrow on the series.
Each pulse also computes an Equity Gap — the difference between its composite 3-month momentum and the 3-month momentum of its equity basket. Positive gap = operational signal is improving faster than stocks are pricing in (bullish divergence). Negative gap = stocks have priced in more than the operational signal supports (bearish divergence).
Two family-wide sub-modules apply across pulses: Maturity Score tags each stock T1/T2/T3 by disclosed volume (demo / early-production / mass-production), and Legacy-vs-Agent Decomposition separates net agentic ARR growth from legacy ARR decay for stocks with both businesses running parallel (CRM · NOW · SAP pattern).
Family reading is a diary entry on our process, not investment advice.
FAQ · from the current data · as of 2026-08-08
Quick answers
What is the Closelook Agentic Demand Index?
The equal-weighted average of 10 operational pulses — EIP, EWP, CCP, EAP, CEP, IAP, TBP, ASP, FDP, TSP — each tracking a different value-capture layer of agentic AI demand, from enterprise infrastructure to consumer devices to model-token usage. A pulse with no current reading drops out of the average rather than being filled in with a neutral placeholder. It is a research diary on where agentic demand is showing up, not investment advice.
What is the current Agentic Demand Index reading?
At the last build of this page, the Agentic Demand Index read 64.7 out of 100 — a Scaling momentum regime — based on 10 of 10 pulses currently reporting, as of 2026-08-08.
What do the Agentic Demand Index momentum bands mean?
Each pulse and the family aggregate read on the same 0-100 scale: 70 and above is Surging, 40 to 69 is Scaling, and below 40 is Emerging. The band names a momentum regime — how hard demand is moving — not how far adoption has spread; a high reading can precede broad adoption by years.
What is the Ramp AI Index real-spend cross-check?
The share of 70,000+ US businesses with a paid AI transaction, from Ramp's card and bill-pay data: 55.0% in June 2026 (+0.8pp month over month, +12.2pp year over year). It measures actual business spend rather than market pricing, and feeds the Enterprise Application Pulse as its Ramp-Adoption module.
What does the Forward Deployment Pulse add to the family?
FDP is the supply-side complement to the demand pulses: 228 forward-deployed engineering roles open across 6 AI value-chain companies as of 2026-08-09, counted daily from their public job boards. Where the other pulses read what markets price, FDP reads whether vendors are staffing to deploy agentic AI.