Discovery · Jobs and opportunities

Which branch

Pick the tree branch that gets this week's prototype from interview counts and ticket rates, and name the result that would move you.

casemediumExercise · 3 min · 9 of 59

On this page

Three opportunities, forty interviews, one FDE week. The branch with the most mentions is not the branch that clears the queue, and the exercise is to show why with the numbers.

The situation

The tree's outcome is tickets carried overnight at a 14-agent tenant: 130 today, under 20 by quarter end. The tenant handles 560 tickets a day, a reply takes 3 minutes, and the 4 pm queue of 130 is about 390 agent-minutes of work at that speed. Forty interviews this quarter across 9 tenants produced three opportunities:

A. Replies to repeat questions take as long as new ones: 26 of 40, from 7 tenants. A drafted reply is claimed to save about 1 minute on every reply; nobody has timed it.
B. The wrong macro gets picked out of 200: 18 of 40, from 5 tenants. It happens on 1 ticket in 6 and costs about 90 seconds each time.
C. An escalation waits on a second person: 31 of 40, from all 9 tenants. 1 ticket in 12 escalates; the agent spends 4 minutes writing the summary, and a drafted summary would save 2 of them. The customer waits 20 minutes.

Sales wants C, because every tenant raised it. The FDE can build one prototype this week.

The question

Which branch gets this week's prototype, and what result in the next two weeks would move you to a different branch?

A strong answer · Rubric · Where answers go wrong

Write your answer first, then open

A strong answer

Convert each opportunity into agent-minutes a day, then into replies against the 130. A: 560 replies × 1 minute = 560, or 280 replies at the new 2 minutes. B: 560 ÷ 6 ≈ 93 tickets × 1.5 minutes = 140. C: 560 ÷ 12 ≈ 47 escalations × 2 minutes = 94. B and C leave most replies near 3 minutes, so their 234 minutes buy roughly 80 replies, short of 130. A clears the queue alone while the saving s stays above about 0.57 minutes, where 560s ÷ (3 − s) reaches 130; half a minute buys only 112.

C has the most mentions because every tenant escalates, but a mention is a conversation, not a minute, and the customer's 20-minute wait is a different metric from the one at the root — first-response time, which belongs on the tree as its own outcome, not as evidence for this one.

So A gets the prototype, as opportunity solution trees sets it up: three agents, one afternoon, 60 tickets, cut if fewer than 30 drafts are inserted. Two results would move me. A measured saving under 0.57 minutes means A alone no longer clears the queue, and B's week of macro-pick logs runs next to make up the gap. Or a ticket log showing escalations are 1 in 4 here rather than 1 in 12: C triples to about 280 minutes, A's return at half a minute, and if T1 measures A that low, C's lower model risk wins, since the agent who handled the ticket corrects the summary. C stays on the tree as next quarter's first branch, and "31 of 40" goes to the roadmap conversation with sales as evidence of spread, not as a decision.

Rubric

Converts each opportunity's rate into agent-minutes a day, then into replies against the 130 in the backlog, with the arithmetic shown.
Chooses A because its minutes clear the outcome, not because of its count or because it uses a model.
Explains why C's 31 of 40 does not win: mentions are conversations, and the customer's wait is a different metric.
States the prototype test for A with a cut rule written before it runs.
Names at least one measured result, with its number, that would move the prototype to B or C.
Keeps C on the tree with its evidence rather than deleting it.

Where answers go wrong

"C, obviously — 31 of 40 and every single tenant." A count is not the outcome's unit; 94 agent-minutes a day is about 31 replies at 3 minutes against a 130-ticket backlog, and the answer never did the division.
"A, because drafting is the AI strategy." The right branch for a reason that would have picked it with any numbers; no minutes, no cut rule, and a reviewer cannot tell the choice from luck.
"Prototype all three this week and compare." One FDE, one week, three half-tests and no pass line for any of them; and B's test is a week of macro logs, not a prototype at all.