Proof of Practice · The Human–AI Loop
Test • Build • Codify • Share
A Loop in Motion
A single real cycle — from ambiguous input to clear decision — showing exactly how the Triad moves through the work.
The loop doesn’t end when something is produced. It ends when a human decides what to do with it.
Context
Before we start
What follows is a real Loop cycle — the kind that runs inside the Triad every day. The scenario is simplified for clarity, but the structure, the handoffs, and the decision points are exactly how it works in practice.
This is not a tutorial. It’s a demonstration. The goal is to show what the Loop actually feels like — not just what it looks like on a diagram.
The input
An ambiguous customer request
The Triad
Maura · CP · Soph
The output
A clear direction — not just a faster answer
Step 1 of 8
The Input
A customer request arrives. On the surface, it looks like a clear ask.
Customer request
“We need a way to make our customer onboarding faster. Can you help us build something?”
This looks like a solution request. It isn’t. It’s a symptom. The gap between what customers ask for and what they actually need is where most AI-accelerated teams lose the thread.
Before anything is built — before a single prompt is written — the human layer steps in.
Step 2 of 8
Human — Reframing the Problem
This is the hardest, most senior work in the whole cycle. It’s also the step most teams skip.
Reframed
“Teams lack visibility into onboarding progress, leading to delays, duplicate work, and inconsistent customer experiences.”
Speed without clarity just produces faster misalignment.
Step 3 of 8
CP — Divergence
Using the reframed problem, CP expands the solution space. The goal is not one answer — it’s a wide range of possibilities worth evaluating.
Option A
Lightweight status tracker
Low friction, fast adoption. Solves visibility without restructuring workflow.
Option B
Automated milestone alerts
Removes coordination overhead. Requires integration with existing tools.
Option C
Shared onboarding dashboard
Single view for both teams. Higher build cost, higher alignment payoff.
Option D
Structured check-in cadence
Human-first. No tooling required. Tests the hypothesis before building anything.
CP’s job is breadth — not to choose, but to surface what’s possible. A wide range is the input to the next step.
Step 4 of 8
Soph — Convergence
Soph stress-tests the options against constraints, tradeoffs, and what will realistically hold up in practice.
Constraints applied
Team of 8 — mixed technical/non-technical
CRM already in use — no new platforms
Timeline: improvement needed in 30 days
Viable directions
Option A — fastest to test, lowest friction. Start here.
Option D — run in parallel to validate before building anything.
Step 5 of 8
Human — Decision
Two viable options. The human evaluates against judgment, context, and what actually matters to this team right now.
Speed vs adoption
What matters more right now: moving fast or bringing the whole team along?
What not to solve yet
Deliberately choosing not to solve the integration question. That’s a future loop.
Decision
Start with Option A — lightweight tracker inside the CRM they already use. Run Option D (structured check-ins) in parallel for 2 weeks as the human validation layer.
Steps 6–7 of 8
Build → Feedback Loop
Human + AI in motion. The Triad builds a working first version — then puts it in front of real people before deciding what comes next.
Build output
A working first version — not a spec, something usable. In their CRM within the week.
Reality check
Two fields tracking the wrong thing. One check-in step redundant with an existing meeting. AI synthesizes patterns. Human decides what to act on.
Step 8 of 8
The Decision Point
The loop doesn’t end with output. It ends with a decision. Three options — only the human can make this call.
↩
Iterate
Re-loop with feedback. Sharpen before expanding.
↗
Expand
The core works. Extend to the full team or use case.
✓
Ship + monitor
Good enough to run. Watch the data. Next loop is scheduled.
In this case: Iterate, then ship.
Fix the two fields. Remove the redundant check-in step. Run it live. This is the layer I’m most focused on — ensuring that speed amplifies the right work, not just more work.
What this shows
This is not prompt engineering.
It’s a system for maintaining problem integrity while increasing speed. The Triad doesn’t remove the human — it puts human energy exactly where it matters most: reframing the problem, making the final call, deciding what to do with the output.
The human led:
Problem reframing (Step 2) and the final decision (Step 5). The highest-leverage steps in the whole cycle.
AI expanded:
Solution space (CP) and tradeoff analysis (Soph) — days of work compressed into hours.
The result:
A better outcome than either human or AI could have reached alone — and a clearer problem definition for the next cycle.
© 2026 Maura K. Randall · thehumanailoop.com · AIGal.io
thehumanailoop.com 