ATT
AI SHOWCASE
4 JUN 2026
/ S1
ATT · AI SHOWCASE · 4 JUN 2026

what AI actually built. at ATT.

three real projects. what the agents did, and what I no longer have to.

Gabriel Jiang · 4 Jun 2026 01 / 09
CASE 01
NYC
TENDER
/ S2
CASE 01 · NYC TENDER

we prepped the whole tender. from one doc.

the requirement document came in. the agents turned it into the full submission set:

requirement doc→ wrote the backlog→ drafted the project plan→ generated the timeline
MCCY_Backlog.xlsx open ↗
BACKLOGuser stories · click to open
PMP_v1.docx open ↗
PROJECT PLANclick to open
Project_Timeline.xlsx open ↗
PROJECT TIMELINEclick to open
one requirement doc in · a full tender set out
02 / 10
CASE 01
NYC
TENDER
/ S2b
CASE 01 · NYC TENDER

then they asked: can you go faster?

mid-presentation, the client asked if we could shorten the plan. by end of the same day:

01revised the project plan: 3 months to a 2-month MVP
02regenerated the timeline and project plan
03split the backlog into MVP + Release 2
04replied to the client
THE RE-PLANNED SCHEDULE
ORIGINAL 3MO3 MO
re-planned · same day ↓
MVP 2MO2 MO

→ won the tender.

asked to shorten · same day · won
03 / 10
CASE 01
NYC
DELIVERY
/ S3
CASE 01 · NYC DELIVERY

then the agents delivered it.

we translated the implementation plan and user stories into project tasks. then the agents ran the board:

agent configures→ agent tests, with evidence→ [ I VERIFY ]→ Ready for Demo→ feedback→ Ready for UAT
the real board. the agents move the cards. the only column I touch is Internal Review.
odoo · project 2 · MCCY NYC marketing system · tasks live board
The live MCCY NYC project board in Odoo: real tasks moving across Backlog, Ready for Development, In Progress, Internal Review and Ready for Demo columns.
agents build and test · I verify and demo
03 / 09
CASE 02
JTT
DEV
/ S4
CASE 02 · JTT DEV

I write the spec. the rest runs itself.

a real JTT task. the agent wrote the markdown spec, Sara built it, the agent verified the work:

.md spec→ Sara builds it→ agent picks up every 3h→ verifies + evidence
the agent's own verification note to Sara: what works, what is missing, with screenshots.
jtt · project 3 · sprint 2.2.1 · freight info field additions task 51 ↗
Real JTT task 51 in Odoo.
screenshot pending · click to open the live task ↗
markdown in · built, verified, evidenced
05 / 10
CASE 02
JTT
QA
/ S5
CASE 02 · JTT QA

then it writes the test steps.

01the agent writes step-by-step test steps
02assigns them to the user
03the user gets notified
04follows the steps and closes the task
jtt · task 51 · chatter · uat test steps → done task 51 ↗
JTT task 51 chatter: the agent's UAT test steps message to Thomas with seven numbered steps, the verification note to Sara, and stage changes through In UAT to Done.
screenshot pending · click to open the live task ↗
I verify · the agent writes the tests · the user confirms
05 / 09
CASE 03
ATLAS
OPS
/ S6
CASE 03 · ATLAS OPS

nobody is tracking. Atlas is.

an issue surfaces in chat. Atlas logs it to Monday, posts every update back to the team, and tracks it to resolution. nobody touched Monday by hand.

monday.com · IMS 126883the hub
Monday.com ticket IMS 126883 AQS mediawall down, the central log, with the agent's tracked updates.
command post · the issue
AQS Command Post group chat: the mediawall reported down.
ATT PMT · reports backposts
ATT PMT group chat: Atlas posts the IMS 126883 ticket updates back to the team.
project team · resolved
ICA project team chat: the mediawall auto-recovered, resolution tracked.
the issue, the updates, the fix. all logged to Monday, hands-off.
a passing comment in chat · becomes a tracked Monday card
06 / 09
CASE 03
ATLAS
OPS
/ S7
CASE 03 · ATLAS OPS

“batch 27 deployed.”

someone posts that one line in the command-post chat. that is the whole trigger. Atlas does the rest.

chat: ‘batch 27 deployed’→ picks up every task in batch 27→ marks them all deployed→ every timestamp recorded
monday.com · issues & requests · batch 27 all completed
Monday.com Issues and Requests board filtered to batch 27: every ticket marked Completed, all tagged Batch 27.
every task in batch 27, marked deployed with timestamps. 0 humans logged into Monday.
tracking that tracks itself
07 / 09
THE
PATTERN
/ S8
THE PATTERN

I review. I decide. they do the rest.

the same shape under all three. it runs itself, reviews its own work, then pings me for the one call that is mine.

01agent picks up work
02acts
03reviews its own work
04notifies me, pending my review
05[ I verify + decide ]
↻loops back to 01
nex-bot · jtt triage · needs your eyes pending my review
Nex-bot Discord triage notification: a task auto-reviewed and reassigned to Gabriel with new UAT feedback, pending his review, with the task link.
the human stays in the loop · not in the weeds
08 / 09
YOUR
TURN
/ S9
YOUR TURN

what would you hand off?

pick one thing you do every week that you wish you didn't. that is where your agent starts.

let's break into groups and find yours.

Built by Gabriel · with an AI agent team   Nex (Opus) · Sonnet · Haiku 09 / 09
[ ← → ] NAV · [ F ] FULLSCREEN · [ N ] NOTES