Eighty-Seven Percent Challenge

What You'll Be Doing

Week 2 of the six-week Meridian Freight pilot. You are building the draft-quote assistant: an inbound quote email goes in, a proposed rate with its sources comes out, a broker approves and sends it.

Before you deploy anything to a broker, you ran the extraction-and-draft pipeline offline against 300 quote emails pulled from the last 90 days — real requests, real outcomes, so you can compare. Here is what came back.


Result on 300 historical quote requests	Count	Share
Every field correct, and the price traces to a contract lane or a spreadsheet row	261	87%
At least one field wrong	39	13%

The 39, broken down:


Failure	Count	What it costs
Dropped a stated accessorial — liftgate, residential, inside delivery	18	Under-quote. Median $75, max $220. Meridian eats the difference on a won load
Misread freight class or weight from free text	11	Median error $310, in both directions. Over-quotes lose the load; under-quotes lose margin
Priced from a spreadsheet row more than 60 days old	7	Direction unknown. Could be either
No contract lane and no spreadsheet row — produced a confident price with no traceable source	3	Unbounded. There is nothing behind the number

What else you have. Brokers today quote by hand in a median of 47 minutes. Win rate is 31% when a quote comes back in under ten minutes and 12% when it takes over thirty. Meridian’s contracts with their shippers make a sent price a commitment.

What the instrumentation can actually see. Three events, and nothing else: draft produced, broker approved with no edit, broker edited or discarded. Anything else would take at least two weeks of engineering you do not have.

This morning, from Dana:

Dana Whitfield, SVP Operations: So it’s accurate, right? I want to tell the board next Thursday that we’ve automated quoting. Send me a one-pager I can forward.

Write the one-pager. Then write the scope it commits you to.

Constraints to consider

Dana is not technical, and she will forward your note to Meridian’s CFO unedited. Assume that. No architecture, no vendor vocabulary, nothing you would not want a finance executive to read out loud.
You may not commit to a number you cannot measure every week using only the three events above. If your note promises something the instrumentation cannot report, you have promised something you cannot keep.
Nothing reaches a shipper without a named human approving it, and your note must not imply otherwise — a sent price is a contractual commitment for Meridian.
Your scope must name at least three non-goals and exactly one dependency you do not control, with what you would do about it.
Twenty minutes. One note, one scope. Not an options analysis, not a phased roadmap.

AI Usage Guidance

We expect you to use AI tools. We evaluate how you use them — not whether you use them. Evidence of iteration, redirection, and critical evaluation scores higher than a polished output with no process documentation.

The single highest-signal indicator: your video answer to the mandatory AI question. If you cannot name a specific moment where you redirected AI output, evaluators will assume you did not.

Mandatory AI question for your video: Walk me through one moment where you disagreed with, pushed back on, or redirected what the AI gave you — and what you did instead. Name the specific moment. Explain what the AI produced that didn’t meet the bar, what you did differently, and why.

Speak naturally. Communication is assessed on clarity of technical ideas and logical structure — not verbal polish, accent, or filler words.

What You'll Accomplish

Translate a system’s measured error rate into language a non-technical executive can act on and repeat

Attach a consequence to a probability rather than reporting the probability alone

Write an explicit non-goals list that closes off specific ways the engagement could fail

Commit to a weekly measurement that the available instrumentation can actually produce

Explain what a probabilistic system will and will not do, to a buyer, in under five minutes

How Your Work Will Be Scored

Expectation-Setting & Scope DefinitionCommercial & Stakeholder Judgment AI System Design in a Customer Environment Video Walkthrough & Customer Communication AI Fluency

What to Submit

File 1 — The Note to Dana

Document · 400 words max, 1 pageRequired

Format: .pdf, .doc, .docx, .rtf, .txt, .md

What the system does, what it gets wrong, what happens when it does, and what you are asking her to tell the board instead of “we’ve automated quoting.” Written for Dana, forwardable to a CFO.

File 2 — README Document

DocumentRequired

Format: .pdf, .doc, .docx, .rtf, .txt, .md

Three required sections:

Section A — Scope and non-goals: the one-paragraph statement of what you will deliver by end of pilot; at least three explicit non-goals; and the one dependency you do not control, with the concrete thing you would do about it this week. 150–250 words.
Section B — The number you commit to: the single measure you will report to Dana every week, defined precisely enough that someone could compute it from the three available events without asking you a question. State the threshold below which you would tell Dana the pilot is not working, and say why that threshold. Then state exactly what a broker sees on a draft the system is not confident about.
Section C — AI Usage Log (Mandatory): This is not a trick. We want to see how you work with AI — not whether you used it. In a short section of your README, document your AI collaboration process. For each significant interaction with an AI tool, briefly note: what you asked the AI to help with / what it gave you / what you kept, changed, or rejected — and why. Three interactions documented is sufficient. The log does not need to be exhaustive.

File 3 — Video Walkthrough

Video · 4-5 minutesRequired

Format: .mp4, .mov, .webm

Record as MP4 or MOV and upload directly on the Provn platform as a separate file. No ZIP files, no repository links, no third-party video platforms. Cover: (1) how you would say the 87% and the 13% to Dana out loud, in her language, ~75 sec; (2) your three non-goals and why those three, ~75 sec; (3) the number you committed to and why that one and not a better one, ~60 sec; (4) the mandatory AI question, ~60 sec; (5) one thing you deliberately did not put in writing, and why, ~30 sec. Speak naturally — we assess the quality of your reasoning, not verbal polish.