The Orchestration Triangle in Practice: The Quarter the Numbers Came Back Too Clean
- Jeff Abbott
- Jul 10
- 3 min read
Priya runs customer support at a software company big enough to have dashboards and small enough that she still knows which customers shout. Eighteen months ago her team deployed an AI triage system, and by every measure it was working: the assistant now resolved 78 percent of tickets without a human, response times had collapsed from hours to seconds, and satisfaction scores held steady. The quarterly review wrote itself, and the proposal on her desk followed the data to its logical end: retire the phone line, reduce Tier 2 by half, let the system do what the system demonstrably did.
She should have felt vindicated. Instead she felt the thing the dashboard had no row for: the numbers were too clean. Her most experienced agents had gone quiet in retrospectives. The tone of the renewal calls she sat in on had cooled in a way she could hear but not cite. She’d been in support for fifteen years; something below her conscious pattern-matching was raising its hand.
Chapter 8 of the book has a name for what she did next. The Orchestration Triangle asks you to draw three vertices — Data, Intuition, Context — and mark, honestly, where a decision is currently landing. Then, like a conductor, bring in the sections you’ve been ignoring; the aim isn’t to pick a winner but to bring three kinds of knowing into harmony.
Data — what the numbers reveal
78% autonomous resolution. Response time down 96%. CSAT flat at 4.3. Cost per ticket down two-thirds. Every measurable arrow pointing the same direction.
Intuition — what experience is muttering
Resolved isn’t the same as solved — the numbers feel too clean. My best people have stopped raising things, which in support is never calm; it’s surrender. And I can hear something in the renewal calls that I can’t yet footnote.
Context — what the situation actually holds
The customers who phone are disproportionately the oldest accounts — the ones whose renewals fund the company. Two of them were promised, at signing, by name, that they’d always reach a person. Support is where this company’s reputation was built; it isn’t just a cost center with good manners.
The decision, she saw when she marked it, was parked hard on the Data vertex — not because the data was wrong, but because it was the only voice being allowed to speak in complete sentences.
So before signing off on the cuts, she followed her intuition’s one citable lead and pulled the tickets the AI had marked “resolved” for her twenty largest accounts. A meaningful fraction had been closed not because the customer was satisfied but because the customer had stopped replying — abandonment, wearing resolution’s clothes. The clean number had a hole in it exactly where her unease had pointed, and the context named which customers were falling through it.
The rebalanced decision kept most of what the data recommended — the triage system stayed, Tier 1 stayed lean — and restored what it had steamrolled: the phone line survived for named accounts, two senior agents became “orchestrators” who audit the AI’s closed tickets weekly, and “resolved-but-abandoned” became a metric with its own row on the dashboard.
The Score Sheet — the habit worth keeping
The triangle diagnoses one decision; the Score Sheet is what turns it into a practice. After each significant decision, Priya now writes a single line with four entries:
The decision — one phrase, no justification
Which voice led — Data, Intuition, or Context
Which voice was suppressed — and what it was trying to say
What happened — filled in later, honestly
Her first line reads:
Tier-2 reduction — led: Data — suppressed: Intuition (“resolved isn’t solved”) — outcome: found abandonment hiding inside resolution; kept the phone line for named accounts.
Over a quarter, the sheet becomes something no dashboard offers: a record of which kind of knowing you habitually overweight, and what it costs you. The discipline, she says, is less about distrusting the data than about noticing, early, when only one section of the orchestra is playing.
The numbers, incidentally, stayed good. They were simply true now, which is a different thing.
Priya is a composite — drawn from conversations with operations and support leaders integrating AI systems, with details changed. The too-clean quarter comes around for everyone.
Try it yourself
Take one live decision where AI or analytics is a loud voice. Draw the triangle, mark honestly where the decision is landing, and give the ignored vertices one hour of standing to speak — one hour of pulled tickets, one conversation with whoever holds the context. Then start your own Score Sheet: one line per decision, four entries. Harmony, not veto.
Comments