Skip to main content

4. Operational Execution

What this flow tells you​

Whether the business actually does what it sold. The gap between "what we sold" and "what it really took" is the most expensive data in the company — close it and pricing, planning, and capacity all sharpen.

In SaaS: uptime, request latency, support resolution, deployment cadence, service-level adherence. In services: timesheet accuracy, milestone delivery, scope adherence. In physical: WIP, capacity utilisation, on-time-in-full, rework rate.

Data sources​

  • Work orders / jobs — planned vs actual start–finish, routing steps, resource usage
  • Capacity and utilization — machines, people, queues, bottlenecks, downtime causes
  • Delivery performance — on-time-in-full, cycle times, error rates, rework, returns
  • Service-level events — uptime, incident response, SLA breach, customer-impact severity

Decision gates​

  • Job acceptance — does capacity exist for this work? Owner: capacity planner OR scheduling agent.
  • Escalate stalled job — at N minutes/hours/days past plan, route to human. Owner: incident process.
  • Re-route around bottleneck — when a resource is saturated, divert to alternative. Owner: ops + AI agent.
  • Quality hold — defect rate above threshold blocks downstream release. Owner: quality function.

AI capabilities required​

  • Memory — state across long jobs (build pipelines, multi-day incidents, long projects); resume after interruption without re-asking.
  • Perception — read system telemetry, log streams, sensor data; classify normal vs anomalous.
  • Action — restart service, scale capacity, dispatch resource, open incident, request approval.
  • Escalation — name the bottleneck, summarise context for the human picking it up; never leave a human cold.

Crypto-rail instruments​

  • Nautilus TEE attestation — work-completed proof issued by a trusted execution environment; immutable signed evidence that operational milestones happened.
  • On-chain SLA witness — a contract that observes service-level events and pays out automatically on breach (no claim cycle, no negotiation).
  • Tokenised capacity — sell idle capacity as on-chain bookings; demand-side gets verifiable allocation, supply-side gets a tradable inventory.
  • Action chaining via Blinks — Blinks bundle instruction + payment; one click triggers the operational action with the funding attached.

Gauge​

On-time-in-full (OTIF) at the job class, weekly. SaaS variant: monthly service availability vs SLA target. One number, one trend line, one alert.

Kill signal​

OTIF < 95% for two consecutive weeks OR cycle time > 2x planned for any job class OR incident response > SLA in any P1 case. The flow is broken — capacity, process, or both.

Skill coverage​

Shipped: workflow-mapper — map any workflow into A&ID blueprint. constraint-mapping — classify Real/Artifact/Hybrid constraints with ranked impact.

Green-field: capacity-allocation agent with on-chain SLA settlement; incident-response composite that opens, summarises, and routes without losing context; tokenised-booking matcher.

Full matrix: skills-matrix.

Upstream / downstream​

5P slice​

5P inspection: Principles (close the loop between sold and delivered), Performance (OTIF + cycle time), Platform (work-management + observability + AI agent + TEE), Process (accept → plan → execute → attest), Players (operators + ops engineers + AI agent on memory and escalation).