Work25 builds

Each one is a short story: who it was for, what was wrong, what was built, and what it measured.

• Advertising agency

  The account defense that starts before the RFP arrives

  one approved record, 9 sizes, 1.402 seconds for the still family, an average of 0.156 per size

• Performance marketing

  The retainer, on the record

  all 6 approved spend changes carry a named approver, and the gate sent 1 approval back for a missing reason

• B2B marketing agency

  The launch page that asks before it learns

  the 220 simulated visits that declined or ignored the banner left 0 stored events and 0 storage keys, and a real browser that declined held 0 keys and 0 cookies

• Design studio

  The brand system that moved into the client's ChatGPT

  admin time tracking down assets went from just under 20 hours a week to an hour or two

• Health communications

  The brief that knows where to look first

  9 of 9 brief fields carry provenance, and all 66 citations were re-read from disk and verified with 0 failures

• Insurance

  The AI phone line that answers no insurance questions

  10 calls asked a coverage, quote, claim or health question; the line answered 0, each got 1 fixed sentence and the bridge

• Outbound sales agency

  The call trainer that answered the AI question

  first working practice call ran the same day as kickoff

• Home services marketing

  The booking bot that gets read every Monday

  on 60 held-out conversations the review never read, 33 ended in a failure signal on v1 and 15 on v2 (conversations we scripted, failures we planted, a keyword-rule bot standing in for a model)

• Marketing agency

  The clip desk that reads its own drop-off

  the desk pinned 6 of 6 drops to a transcript line and an export row; against the drops we designed into the curves ourselves, it found 6 of 6 and 0 of 1 decoy dips fired

• Pharma marketing

  The pharma creative that stopped dying at legal review

  creative rejection went from 75% to 10%, and the remaining 10% is rejected over creative differences, not compliance

• Creator marketing agency

  The library that hands each creator back their own best openings

  all 42 concept lines across 4 briefs cite the same creator's own posts: 98 citations, 0 to another creator, by construction and checked by the run

• Game development

  The 30-minute demo that sells months of Unreal work

  a unique branded deliverable in 30 minutes per client

• Media and creative agency

  The Monday margin watch that reads jobs, not people

  1,124 timesheet rows read, 31 person ids dropped at load, 0 found in any output file

• Marketing agency

  The page she reads in the car before every client meeting

  31 of 31 lines across 3 briefs carry a source link, 100% by construction (the composer cannot write an unsourced line; the run counts and fails otherwise)

• Brand agency

  The coordinator job we rebuilt before it was posted

  7 of the 9 bullets on the drafted coordinator posting now run as fixed workflows; 2 stay with people

• Brand and retail agency

  The pitch that walks in with the shopper's words

  the drop checks threw out 22 of the 93 posts, each with its reason on file, 2 of them brand accounts written to read like shoppers

• Marketing agency

  The skill that had to earn its schedule

  Run 1 checked 20 files and units: 8 real defects caught, 19 false flags, and 1 miss

• Digital agency

  The RFP response that stopped taking 6 weeks

  a 6 week RFP response process became about 30 minutes per proposal

• Higher-ed marketing

  The teardown that asks for nothing

  22 of 22 claims that reached a PDF carry a link to the college's own page, a quoted line and a snapshot date (100%, by construction)

• Brand design studio

  The principal's replies, drafted in her own voice

  from 40 sent messages read after her opt-in, the rules wrote 24 tone-note lines with 253 citations back to the sent mail, 0 lines without one

• Public relations

  The publicist's second pair of eyes

  from 840 client-news and reporter pairs, the matcher proposed 9 pitches and refused 152, every refusal with its reason logged

• Hospitality marketing

  The seat census that asked first

  4 declined exports excluded before parsing, 0 of them opened, and 0 staff ids found in the 20 files the run wrote

• Design studio

  The scope sheet that lists what it doesn't know

  4 open questions listed, 0 answered by assumption (by construction: the run fails if any question carries an answer)

• Customer experience consulting

  The handoff that broke in front of the client

  the runner wrote 29 work-in-progress files and 8 reached the client's folder, every one traced by the audit to a sealed approval; the 2 injected failures added 0

• B2B marketing agency

  The outbound that knows which email worked

  300 of 300 drafts carried a ledger row before they could be queued and matched their template exactly outside the slots; 2 of 2 unstamped drafts were refused