SFR 10-Year OPORD · Deep Dive
Eighteen digital employees, one rule, and a platform that treats every automated system like a governed member of the team. This is how six executives run a $100MM company.
People do hands, trust, and judgment. Reliable software — timers, rules, math — does the daily coordination, every time, auditable, and running even if an AI provider goes down. AI's biggest job is building and improving that software, plus the work only intelligence can do: understanding language, reading photos, drafting plans.
| Tier | What runs there | Examples |
|---|---|---|
| Pure software | Rules, timers, math, dashboards — no AI needed in the daily loop | The Watchfloor's escalation timers · the hiring math · collections sequences · license calendars |
| Hybrid | AI only at the conversation boundary; a deterministic engine does the action | Front desk (AI talks, software books) · recruiter screening · warranty intake |
| AI-core, bounded | Intelligence is the product — run per event, never idling | Scoring a sales transcript · reviewing an install photo set · drafting the weekly FRAGORD (once a week) |
The Human Forefront: humans stay at the forefront of human-to-human things — teammate to teammate when critical, and teammate to customer. Every substantive customer conversation is a human's job (it's why every Field Supervisor is paired with a PM). Coaching, feedback, and hard news are delivered by people. The design test for every automation: does it free a human for a relationship moment, or insert a machine into one? The first is the doctrine; the second is banned.
The banned pattern: an AI that "polls around" checking routine things is a cron job wearing a costume — slower, costlier, and unauditable. If it can be a rule, it must be a rule. AI cost must never scale with revenue.
Each one replaces a role a traditional $100MM contractor hires — and each is governed like a hire: an identity, defined permissions, an audit trail, and a human who owns it.
| Digital employee | What it does | Replaces the need for | Status |
|---|---|---|---|
| AI Front Desk | Answers every inquiry in under a minute, 24/7; books appointments; hands humans the hard conversations | A call-center shift | Building |
| AI Sales Manager | Scores every appointment against the 10-step method; drives weekly coaching | Ride-along hours no manager has | Live |
| AI Dispatcher | Optimizes crew routes and schedules; proposes reshuffles when weather or delays hit | Dispatch coordinators | Partial |
| AI Project Coordinator | Prep instructions, day-before confirmation, day-of ETA, end-of-day photo recap — automatically, every job | 2–3 project coordinators per branch | Building |
| AI QA Inspector | Reviews install photos against standards before the crew leaves the site | Catch-it-later rework | Building |
| AI Proposal QA | Checks every proposal against engineering standards and margin floors before presentation | Estimate review meetings | Building |
| AI Controller | Watches every job's costs daily; flags a margin leak the day it starts | A staff of analysts | Building |
| AI Workforce Planner | Runs the capacity math 90 days ahead — which roles, which branch, when | Reactive scramble-hiring | Building |
| AI Recruiter | Sources, screens, schedules interviews, checks the red flags | Recruiting coordinators | Partial |
| AI Trainer | Role-based training paths, certification gates, refreshers | A training department | Partial |
| AI Watchfloor | The "nothing falls through" engine — a timer on every handoff, automatic escalation on every breach | Managers chasing loose ends | Partial |
| AI Warranty Agent | Intake, triage, scheduling, root-cause tracking that feeds quality back to the field | Warranty admin backlog | Partial |
| AI Collections | Respectful, systematic AR follow-up and payment plans | Awkward, inconsistent chasing | Partial |
| AI Marketing Engine | Generates creative (live today); allocates spend to the channels producing the best customers | Agency retainers | Partial |
| AI Chief of Staff | Drafts the weekly FRAGORD from plan vs. scorecard vs. issues — leadership decides | Meeting-prep hours | Building |
| AI Data Steward | Clean records everywhere — duplicates resolved, quality enforced | Data-entry cleanup | Building |
| AI Safety Officer | Incident logs, certification and license expirations, safety cadence | Compliance falling to memory | Building |
| AI Legal Assistant | Contract templates, lien deadlines by state, licensing calendars | Missed-deadline risk | Building |
SmartOS v3 isn't "a better CRM" — it's the platform the digital employees run on. Six capabilities make that real:
| Capability | Why it matters |
|---|---|
| One event stream | Every business event — lead, booked, sold, scheduled, installed, signed-off, claimed, paid — recorded once, readable by every system |
| Governed digital employees | Every automated system has an identity, defined permissions, and an audit trail — hired, reviewed, and accountable like a person |
| Human-in-the-loop queues | One standard approve/exception inbox for everything the systems propose — people stay sovereign |
| Tested before trusted | Every AI behavior passes an evaluation suite before it touches a customer, and can be rolled back in minutes |
| Everything observable | Every automated action logged and attributable — no black boxes |
| The customer portal | The external face of the whole machine — where the customer sees it working for them |
Owned by: the Director of Business Systems & Technology — every system, every integration, the AI platform, and the data spine. In an AI-first company this is the keystone seat.
