DeadheadDesk is pre-launch. So this page has no logo wall, no glowing quotes, no suspiciously round numbers. What it has: the measurement system every future claim will come from, the blanks we intend to fill in public, and a log of exactly how the agents fail. Read the labels.
IF A NUMBER EVER SHOWS UP ON THIS PAGE UNLABELED AND UNMEASURED, EMAIL THE FOUNDER. HE'LL FIX IT AND THANK YOU.
Median draft time
0 SEC
Inbound timestamp → priced draft sitting in the rep approval queue.
SAMPLE DATAItems processed this week
0
Quote emails read + carrier check-ins handled. Same unit as your invoice.
SAMPLE DATAEval-suite accuracy (nightly)
0%
Field-by-field grading against a corpus of real freight emails, every night.
SAMPLE DATAMedian draft time. The clock starts when the email hits your quote inbox and stops when a priced draft lands in the approval queue. Rep approval time is reported separately — we don't hide the human gate inside the headline number.
Items processed. Every quote email read and every carrier check-in handled counts as one item. It's the exact unit on the invoice: 500/month included, $1.50 each after. No creative accounting between the dashboard and the bill.
Eval-suite accuracy. Every night, the agents re-run a corpus of real, permissioned, redacted freight emails with known correct answers — origin, destination, equipment, dates, price band — and get graded field by field. A scenario passes only if every load-critical field is right. If the pass rate sags, a drift alarm pages us. You find out from the weekly report, not from a wrong quote.
THE TILES ABOVE SHOW SAMPLE VALUES FROM OUR TEST HARNESS — PRE-LAUNCH, THERE IS NO CUSTOMER TRAFFIC TO SHOW. PILOT CLIENTS SEE THEIR OWN TILES, LIVE, FROM WEEK ONE.
We mystery-shop brokerages with realistic spot-quote emails and time the replies. Before the full 50-brokerage study publishes, here's the shape of the early data — and a question for you.
Step one: guess where you land
On a typical weekday quote request, how fast does your desk send a priced reply?
METHOD: IDENTICAL DRY-VAN QUOTE REQUESTS SENT TO BROKERAGE QUOTE INBOXES DURING AND AFTER BUSINESS HOURS. BUCKET = TIME TO FIRST PRICED REPLY. "NEVER" = NO REPLY WITHIN 5 BUSINESS DAYS. EARLY-SAMPLE FIGURES SHOWN; THE FULL STUDY (N=50, METHODOLOGY AND ANONYMIZED RAW BUCKETS) PUBLISHES ON THIS PAGE.
We could mock up glowing case studies tonight. Instead, here are the actual one-pagers — blank, waiting on three design partners. The metric rows are already locked. The partner fills them in at pilot day 30, wins and misses both, sign-off required before anything publishes.
| Median response latency | → |
| Loads covered / mo | |
| After-hours capture rate | → |
| FTE-hours redeployed / wk |
PUBLISHES AT PILOT DAY 30 · SAME LANES, SAME STOPWATCH · MISSES INCLUDED
Claim this slot →| Median response latency | → |
| Loads covered / mo | |
| After-hours capture rate | → |
| FTE-hours redeployed / wk |
PUBLISHES AT PILOT DAY 30 · SAME LANES, SAME STOPWATCH · MISSES INCLUDED
Claim this slot →| Median response latency | → |
| Loads covered / mo | |
| After-hours capture rate | → |
| FTE-hours redeployed / wk |
PUBLISHES AT PILOT DAY 30 · SAME LANES, SAME STOPWATCH · MISSES INCLUDED
Claim this slot →DESIGN PARTNERS GET PRIORITY INTEGRATION AND A SAY IN THE ROADMAP. IN EXCHANGE, THEIR NUMBERS GO ON THIS PAGE — GOOD OR BAD. THAT'S THE DEAL.
Every agent fails somewhere. The product isn't "it never fails" — the product is failure that gets caught, measured, and fixed before it costs you a load. These are real failure classes from our test corpus, with the monitoring response for each.
| Failure class | What it looks like | How the harness catches it |
|---|---|---|
| Sender format drift | A regular shipper updates their TMS; their quote emails change layout and extraction confidence sags. | Per-sender confidence tracking + nightly regression. Drift alarm inside 24 hours; corpus updated; suite re-run green before the next shift. |
| Ambiguous equipment | "Need a van" — dry or reefer? 53' or sprinter? | Never guessed. The draft asks the confirming question in plain text, and the ambiguity is flagged to the approving rep. |
| Stale or outlier market rate | The rate feed hiccups and returns a number 40% off your lane history. | Sanity band against your own TMS lane history. Outside the band, the draft is held and flagged — it does not price. |
| Garbage carrier replies | A driver answers a check-in with a photo of a BOL, a voicemail transcript, or just "yes." | Parse-confidence threshold. Below it, the raw message routes to a human with the full load context attached. No status gets written on a guess. |
| TMS / API breakage | An auth token expires at 2 AM and writes silently stop. | Heartbeats on every connector. Silent-failure alerting pages us, not you — and any gap shows up in the weekly report either way. |
| Model behavior shift | An upstream model upgrade quietly changes how drafts read or price. | Model versions are pinned. Candidates run in shadow against the full eval suite before any swap; the diff is reviewed by a human first. |
| Double-brokering red flag | A carrier's MC, callback number, and dispatch email don't line up. | Hard escalation, always. The agent never green-lights a suspect carrier — it packages the evidence and hands it to your rep. |
THIS LOG GROWS. EVERY NEW FAILURE CLASS FOUND IN PRODUCTION GETS A REGRESSION TEST IN THE NIGHTLY SUITE AND A ROW ON THIS PAGE.
Pilot clients get this every Monday before coffee. Here's the condensed template, filled with sample data so you can see exactly what gets counted — including the week something broke. The full walkthrough lives on the How It Works page.
| Line item | Count | Δ vs prior wk |
|---|---|---|
| Quote requests read | ||
| Drafts held for rep approval | ||
| Approved unchanged | ||
| Approved with edits | ||
| Rejected by reps | ||
| After-hours drafts (18:00–08:00) | ||
| Carrier check-ins completed | ||
| Exceptions escalated to humans | ||
| Nightly eval suite | ||
| Margin on approved quotes that booked |
ALL FIGURES SAMPLE DATA FROM THE TEST HARNESS. "MARGIN" = QUOTE DRAFTED BY THE DESK → APPROVED BY A REP → SHIPPER BOOKED. EVERY LINE IS A COUNT OR A DOLLAR — NO VIBES.
01 · Baseline · day 0
Before anything switches on: your current reply latency, pulled from your own sent-mail timestamps plus an outside mystery shop. That's your "before," on paper. You keep it even if you walk.
02 · Shadow grading · wks 2–3
The agent drafts silently on your real inbox while reps work as usual. Every draft is graded against what the rep actually sent — extraction, price, tone. Misses become regression tests.
03 · Before/after · day 30
Same inbox, same lanes, same stopwatch. The pilot's go/no-go reads off this delta — and design-partner deltas publish on this page with sign-off, wins and misses both.
NOTHING ON THIS SITE BECOMES A MARKETING CLAIM UNTIL IT'S A MEASURED DELTA WITH A DATE ON IT.
The response-time audit mystery-shops your own quote inbox and plots your desk on the distribution above. You keep the number either way — it's the "before" column of your case file, whether or not we ever fill in the "after."
PILOT FEE CREDITS AGAINST THE $7,500 SETUP. EVERY SEND GATED BY YOUR REPS FROM DAY ONE.