Case Study · Synthetic Triage Agent

SYNaTRA: Synthetic Triage Agent

The client wanted in-call coaching. They didn't have the data foundation to support it yet, so I built synthetic triage first, on data they already had, and let the team prove they could scale it before I built the harder thing on top.

Synthetic triage Phased foundation 2 to 19 paths in a month Team-scaled

Internal workflow tool; identifiers redacted.

The Problem

The ask was ahead of the data that would have to support it.

An in-call coaching agent with no foundation underneath it

  • The client wanted an in-call coaching agent, but they didn't have the data foundation to support one yet.
  • Tickets weren't being categorized consistently, so the resolution data that did exist couldn't actually be leveraged.
  • Without consistent categorization, there was no reliable signal to train a coaching agent on, in-call coaching wasn't a data problem waiting to be solved, it was a data problem that hadn't been created yet.

The Approach

Not "that's not possible." What's possible now, and what it builds toward.

Reframe, don't refuse

  • Instead of telling them in-call coaching wasn't possible yet, I reframed the problem: start with synthetic triage using the resolution data they already had.
  • Built out the common resolution paths first, the patterns that already existed in the data, just uncategorized.
  • Trained the team to add resolutions to tickets consistently, so the data feeding the system would keep improving after I left.

A phase they could stand on

  • Delivered real value on day one with triage, not a promise of value once the "real" project started.
  • Positioned them for the next phase, in-call coaching, on a foundation that would actually exist by the time I got there.
  • Scaling the agent was tied directly to how much the data improved, which put the team in control of the timeline instead of me.
  • Synthetic triage
  • Resolution path modeling
  • Ticket categorization training
  • Phased delivery

The Results

Nineteen resolution paths in a month is not a small number.

From manual categorization to a system the team scaled themselves

  • Moved from manual, inconsistent categorization to automated triage that freed up real time and money that used to go into re-categorizing tickets after the fact.
  • In the first month alone, the team scaled from two common resolution paths to nineteen. Nobody taught them how to add the other seventeen.
  • I came back to build the more complex in-call coaching agent on top of that foundation, the phase the client originally asked for, now actually buildable.
Starting resolution paths
2
Common resolution paths identified at project start.
Resolution paths after 1 month
19
Scaled by the team themselves, unprompted.
Growth timeframe
1 mo.
Nearly tenfold increase in resolution paths in the first month.

What the Team Owns Now

The next phase was built on a foundation they understood, not one they were handed.

The triage system, and the habit that feeds it

  • The team owns the triage agent and the discipline of consistent ticket categorization that keeps it accurate.
  • They proved this by scaling resolution paths nearly tenfold in a single month, without waiting on me to do it for them.

A real foundation for what comes next

  • The in-call coaching agent I'm building now sits on data and categorization habits the team built themselves, not on assumptions I made for them.
  • A vendor hands over a feature and moves on. A team that understands its own system grows into the next phase without waiting for someone else to build it for them.
“Delivering your vision that your team can actually own.”
↑ Back to top