Chapter 10 Resources

Scaling People With Policy, Process, and AI

← Back to Book Resources

Key Concept

The goal isn’t to replace people with systems. It’s to make people more effective with better systems.

The New Hiring Question: “Do we need a human, a bot, or some combination—and what specs should either one follow?”

Figures (Full Resolution)

Figure 10.1: Rails vs Ramps Onboarding

Rails vs Ramps Traditional ramp-based onboarding takes 12 weeks. Rail-based onboarding with guided workflows achieves productivity in 2 weeks.

Downloadable Resources

Templates & Guides

Quality Is Conformance to Spec

Philip Crosby’s insight becomes critical when blending human and AI workers:

Worker Type Spec Unclear Spec Clear
Human Exercises judgment (inconsistent) Follows spec (consistent)
Bot/AI Cannot function Executes precisely
Best Outcome Variable quality Predictable quality

The businesses that scale successfully define specs clearly enough that either a trained human OR a well-configured bot can execute them—and produce the same quality output.

Rails vs Ramps: The Training Revolution

Traditional Ramp (12 Weeks to Productivity)

Week 1-2:   Read the manual. Shadow Jennifer.
Week 3-4:   Handle simple invoices with supervision.
Week 5-8:   Gradually take on more. Make mistakes. Get corrected.
Week 9-12:  Finally productive, but still asks questions.

Result: 12 weeks, high senior staff burden, 23% error rate, some new hires quit.

Rail-Based Onboarding (2 Weeks to Productivity)

Day 1 AM:   30-minute system overview video
Day 1 PM:   Begin processing. System guides each step:
            - Suggests vendor match (confirm or correct)
            - Suggests category (confirm or correct)
            - Shows required documentation (upload prompt)
            - Routes to appropriate approver (automatic)
Day 1 End:  45 invoices processed. 2 policy exceptions handled correctly.
Day 5:      200+ transactions, 94% first-pass accuracy
Day 10:     Fully productive. Handles routine work independently.

Result: 2 weeks, 4 hours training time, 6% error rate, zero turnover.

AI-Infused SOPs

Traditional SOP: “Verify vendor exists, confirm amount matches PO, code to appropriate category, route for approval if over $500.”

AI-Infused SOP: The staff member opens an invoice. The system:

Step Traditional AI-Infused
1. Vendor Look up manually System suggests match
2. PO Match Compare documents Auto-comparison (if connected)
3. Category Remember rules AI recommends based on history
4. Approval Remember thresholds Auto-routes if needed
5. Complete Hope it’s right System validates

The human confirms, adjusts if needed, moves on. The “SOP” is embedded in the workflow, not a separate document nobody reads.

Job Descriptions That Include Rules

Traditional Job Description

“Responsible for accurate expense coding.”

What does “accurate” mean? How do you measure it? What happens when it’s not?

Rules-Connected Job Description

“Responsible for expense coding that meets Tax Ready standards: – 95%+ consistency rate – Zero suspense account usage – Documentation within 10 days

System monitors accuracy and surfaces exceptions for review.”

The employee knows: – What “good” looks like – That the system is measuring – That exceptions surface—not hide

This isn’t surveillance. It’s clarity.

The Override Conversation

When the system blocks something, teach proper escalation:

1. READ the block message
   └── What specifically is the issue?

2. FIX if you can
   └── Missing documentation? Upload it.
   └── Wrong category? Correct it.

3. ESCALATE if you can't
   └── What you tried
   └── What the system said
   └── Why you think it's legitimate

4. DOCUMENT if override approved
   └── Reason recorded in audit trail

Overrides aren’t bad—but they should be rare and documented.

Case Study: 12 Weeks to 2 Weeks

Client: Regional accounting firm, 22 staff

Before (Ramp-Based)

Metric Before
Time to productivity 12 weeks
Senior staff training time 15-20 hours per hire
Error rate (first 90 days) 23%
Turnover during training 2 of 5 hires quit

After (Rail-Based)

Metric After Improvement
Time to productivity 2 weeks 83% reduction
Training time 4 hours per hire 80% reduction
Error rate (first 90 days) 6% 74% improvement
Turnover during training 0 100% improvement

Managing Partner: “We stopped trying to transfer 20 years of tribal knowledge in training. We embedded the knowledge in the system.”

Becoming the Architect

From The E-Myth: Work on your business, not just in your business.

Role Focus Time Spent
Brick Layer Doing the work In the business
Architect Designing systems On the business

The architect doesn’t lay every brick. The architect designs buildings so anyone following the plans can lay bricks correctly.

Key Question: “How would someone else do this correctly if I weren’t here?”

If you can’t answer that question clearly, you’re not ready to scale.

Pro Tips

Written Policies Enable Delegation: > Written policies don’t just improve compliance—they let you delegate confidently. When the “right way” is documented, you can hand work to junior staff, contractors, or offshore resources without sacrificing quality. The policy becomes the training.

Key Employee Leaves? > When a key employee leaves, don’t panic about what’s “in their head.” If your policies are documented and your systems enforce them, the transition is about access handoff—not knowledge archaeology.

Key Takeaways

  1. The hiring question has changed: “Human, bot, or both—and what’s the spec?”
  2. Quality is conformance to spec—same standard for humans and bots
  3. Rails beat ramps: Guided workflows beat manuals nobody reads
  4. AI-infused SOPs embed training in daily work
  5. Job descriptions connected to rules create clarity, not surveillance
  6. If you can’t articulate the spec, you’re not ready to scale

Your Next Step

Identify one process that depends on “tribal knowledge”—information that exists only in someone’s head. Document it well enough that a new person could follow it.

Want help building AI-guided rails? Apply for a complimentary Tax Ready Assessment – we’ll show you what rail-based onboarding looks like for your processes.

← Chapter 9 | Back to Book Resources | Next: Chapter 11 →

Scroll to Top