Every CRM, every CPQ, every ERP built for the last 30 years runs forecast-first. Sales sets an ambition → ops scrambles to deliver → margin is discovered at month-end. Allometry inverts it. The deployed asset, the realized margin, the Pulse itself — those drive what gets quoted, targeted, grown, and priced. The physical teaches the commercial.
Thirty years of enterprise software has one direction of flow: commercial → physical. Ambition cascades down. Reality absorbs the shock. We reversed the arrow.
Every CRM, CPQ, ERP, FSM built since 1990 assumes this direction. It's why month-end exists. It's why the scheduler becomes the bottleneck. It's why 30% of quotes are guessed.
Every closed job trains the scorer. The scorer governs the next quote. The next quote creates the next job. The loop runs in reverse — that's why margin compounds instead of leaks.
Salesforce, HubSpot, and every FSM incumbent is now announcing "AI agents" bolted onto their legacy architecture. That doesn't work. When your data model is account-rooted and your UI is pipeline-rooted, adding agents on top just automates the wrong flow faster.
Salesforce is introducing a headless architecture that decouples the data layer from the user interface, enabling AI agents to access and act on customer data across systems.
"The Pulse" is the primitive the customer sees — one number, one signal per account. Underneath, it's a swarm of specialized agents, each scoring one dimension, cross-referencing each other's outputs, voting into a composite score in real time. Multi-agent by design, not by feature.
realized-margin RAG · cost-floor enforcement · per-address rollup
LTV × retention prob × location expansion curves
portfolio concentration · supply concentration · anomaly detection
unbuilt-address inference · multi-site readiness · renewal timing
Each loop's four modules are themselves agents in the larger swarm. Margin Protect (02) is an agent. Risk Scoring (15) is an agent. Outbound ABM (10) is an agent. 16 modules × 4 agents per loop = 64+ agents running concurrently per customer. The "product" is the swarm's coordinated output. The UI is the transcript.
Four structural reasons incumbents can't just "add AI" and close the gap. None of them are about model quality. All of them are about architecture debt.
Legacy CRMs root on Account → Contact. Changing the root to Location means re-keying every table, every query, every downstream system. It's not an AI refactor — it's a company-ending rewrite.
Pipeline-first UI, forecast-first reporting, quota-first incentives. Incumbents can add agents, but the agents operate on top of the wrong loop. Automating commercial → physical faster just leaks margin faster.
Per-address margin accuracy requires realized margin data keyed to locations. Salesforce has billions of contact records and zero address-level margin records. You can't fine-tune your way out of a data schema gap.
Their installed base is the anchor keeping them commercial-first. Every one of their 150K+ customers is running the old flow. Breaking it breaks them. We have zero switching cost from the old model because we never adopted it.
The same idea, compressed into different surface areas. Pick the one that lands for your audience:
If your margin varies by address, your accounts are multi-location, and you're discovering margin at month-end — the inversion is worth 30 minutes of your time. We'll show you what your Pulse would look like.
Book a Demo → See the Pulse