NOW SHIPPING: three product lines at unisys ★ ON DECK: IATA ONE Record compliance ★ OPEN TO: tier-1 PM roles · advisory · speaking ★ BENGALURU → THE WORLDNOW SHIPPING: three product lines at unisys ★ ON DECK: IATA ONE Record compliance ★ OPEN TO: tier-1 PM roles · advisory · speaking ★ BENGALURU → THE WORLD
Vol. 09 · Issue 26 · Product, with taste

PRODUCT,
SHIPPED
HOT. 🛬

I'm Nayan Chatterjee - Associate Principal PM at Unisys, opinionated about air cargo, branch banking, and the difference between roadmaps and wish lists. Probably in that order.

// elsewhere
Nayan Chatterjee
Cover · Spring '26№ 001
@ bengaluru, india
The PM Issue
Nayan C.
Years shipping
9+
Previously seen at
Cognizant · HSBC · Morpheus · Unisys
// § 01 - the story

Built in code.
Sharpened in product.

I lead a 5-person product team - two GTM leads, two senior PMs, one associate PM - and work day-to-day with a 40-50 person design and engineering org at Unisys. It's a lot of calendars, but the shipping keeps it honest.

I started in engineering at Cognizant, building analytics for banks and insurers - learned how a thing actually gets shipped, then crossed into product where everything else is just storytelling on top of operations.

HSBC taught me regulated industries at scale. Morpheus taught me consumer edtech in a hurry. Unisys taught me that even cargo manifests and teller screens have a brand voice (it's "trust, but compliant").

Now I own three enterprise product lines - CPS, Failed Milestone dashboard, and UBB - serving Tier-1 airlines and building societies across UK, LATAM, and APAC.

The bit your recruiter wants: 9+ years across SaaS, banking, and cargo. Roadmaps, GTM, regulatory delivery (KYC/AML, PCI DSS, IATA ONE Record). I personally write the frontend code for every prototype I ship - production-ready, API-wired, just needs the backend plugged in.

// § 02 - selected work

RECEIPTS, NOT VIBES.

Numbers attached. Case studies available on request.

UNISYS · CPS · AIR CARGO

Re-engaged a Tier-1 carrier. Shipped four new modules.

Re-opened Air India Cargo, structured an IATA GSS licensing deal, launched CargoIntel · DataGuard · CapacityIQ · CPS CoPilot. +20% users, +15% MAU.

№ 001
UNISYS · FAILED MILESTONE · CARGO IQ

Failed Milestone Dashboard - turned breaches into billables.

Owned Failed Milestone dashboard roadmap for Air Canada & American Airlines. Drove Cargo iQ Spec V7.0 readiness across 23 impact areas. Zero P0s, three quarters running.

№ 002
UNISYS · UBB · BRANCH BANKING

A 4-year roadmap a CFO will actually fund.

Authored the 2026–2029 strategic roadmap for a retail branch platform. $2M+ pension-fund POC. 86+ teller transactions, KYC/AML and PCI DSS embedded in release cycle.

№ 003
MORPHEUS · 2022–23 · EDTECH

+60% subs in two quarters. -15% churn.

Launched a NEET/IIT prep platform, designed a personalization algorithm that cut test anxiety 30%, shipped an Android OMR checker that took 40% off grading time.

Q1
Q3
Subscriptions +60%
Before
After
Churn -15%

Two quarters. Subs up 60%, churn down 15%. Illustrative shape, exact figures on request.

№ 004
UNISYS · CPS · 2026 · AIR CARGO
IN FLIGHT
AIR CARGO

Three surfaces, one cargo drop-off. Whole front end shipped before a line of backend.

Designed and built a three-app cargo booking prototype for a regional US cargo carrier. A self-service drop-off kiosk for walk-in shippers, a carrier acceptance desk for the counter team, and a CPS portal stitching it together. I wrote a 40-point requirement spec, then shipped three working React prototypes on Lovable, API-wired and ready for the backend to plug in. Same pattern as everything else here: spec, working prototype, pressure-test with engineering, then commit the build. De-risk the bet before you scale it.

I wrote the spec, built all three working React prototypes, and pressure-tested the flows with the engineering lead before the team committed the backend build. Front end first, de-risked, then scale.

№ 005 · currently in build
// how every one of these got built

ONE PATTERN. EVERY TIME.

01
Spec
write the PRD and requirements
02
Prototype
build the working React front end myself
03
Pressure-test
walk it through with the engineering lead
04
Commit
only then does the team size and build
05
Ship
measure, learn, repeat

De-risk the bet before you scale it. Same guardrails, every client.

// § 02.5 - what I actually can't show you

CLASSIFIED, BUT REAL.

Enterprise product work lives under NDA. So here's the shape of it, with the sensitive parts pulled. Everything below is real work, redacted on purpose.

CONFIDENTIAL
PRD · v2.4
Cargo Booking Spec
Objective
Requirements
Acceptance criteria
40 reqsclient redacted

Product spec · air cargo · 40 requirements · client redacted

CONFIDENTIAL
Milestone Intelligence

Delivery dashboard · Tier-1 carrier · figures under NDA

CONFIDENTIAL
Platform Roadmap
Q1Q2Q3Q4
Foundation
Scale
Intelligence
Agents
4 yearsnames withheld

4-year platform roadmap · branch banking · names withheld

Want the unredacted version? Bring a real brief and an NDA. Link at the bottom.

// § 03 - opinions, unsolicited

FIVE THINGS I'LL DIE ON THE HILL FOR.

№ 01
A bad product cannot be out-roadmapped.
№ 02
If your PRD needs a glossary, it isn't one.
№ 03
OKRs > KPIs. The verb matters more than the number.
№ 04
Compliance is a feature. Ship it like one.
№ 05
The best PMs write release notes their grandma could grade.
// § 04 - the personal stack

IN ROTATION.

in rotation · tools
Shipping.
  • · Jira · Confluence · the unholy trinity
  • · Figma · for the artifacts I owe design
  • · Datadog · Pendo · Mixpanel · the truth tellers
on the desk · domains
Reading.
  • · IATA ONE Record · Cargo iQ · e-AWB · FSU
  • · Teller platforms · KYC/AML · PCI DSS
  • · Goal: 0 critical regressions for Q4
in the brain · stack
Coding.
  • · React · JS
  • · Python · SQL · for the days I miss it
  • · AI/ML (PGP, Great Lakes) · for the next layer
// also tagged as
#ShipsBeforeSlides#PRDPoet#NotAProjectManager 🙃#ZeroBugsZeroChill
// § 04.5 - live prototypes

DROP THE NEEDLE.

Working prototypes. Click through any of them.

// clickable lovable prototypes
// § 05 - now building · in flight
Currently in build

AGENTIC AI
DELIVERY OS.

● IN BUILDOne agent from this thesis is already in build - the Branch Teller Assistant.

An end-to-end agentic AI roadmap I'm authoring right now - the blueprint for how we'll deliver client implementations going forward. Twelve+ task-specific agents on a shared, governed platform: status reports, change impact, requirements, code, tests, triage. Same five-step pattern, reused. Same guardrails, every time.

North star
Cut implementation effort. Lift quality. Make the next client engagement ship faster than the last.
The platform - shared across every agent
Agent runtime
Container apps + functions, per-agent isolation
Orchestration
Semantic Kernel + MCP-mediated tools
Model layer
Reasoning / workhorse / classify - tiered by task
Per-tenant RAG
Hybrid vector + BM25, one index per client
Guardrails
PII, prompt-injection, audit, prompt registry
Eval harness
Accuracy sampling + cost-per-agent telemetry
Every agent follows one pattern
01Trigger
02Gather
03Reason
04Act
05Learn
12 agents on the build list
trigger · model tier
A01
Workhorse
Status Report
Cron
A02
Workhorse
Meeting Minutes
Meeting end
A03
Reasoning
Change Impact
Ticket webhook
A04
Reasoning
Requirements Extraction
Doc upload
A05
Workhorse
Backlog Seeding
Post-extraction
A06
Workhorse
Code Co-pilot
IDE
A07
Workhorse
Test Case Generator
Story ready
A08
Workhorse
Automated Code Review
PR opened
A09
Classify
Defect Triage
New defect
A10
Workhorse
Branch Config Generator
Profile upload
A11
Workhorse
Incident RCA
New incident
A12
Classify+Work
Self-Serve Assist
User chat
90-day MVP - build the platform, ship one agent well, then replicate
1
Days 1 - 30
Platform + first agent
  • Provision model layer + scaffolding
  • Guardrails: PII, content, audit
  • Ship Status Report agent end-to-end
  • Security signoff on deployment pattern
2
Days 31 - 60
Replicate + baseline
  • Ship Meeting Minutes + Change Impact
  • Baseline hours, TAT, defect rate
  • Start eval dataset + accuracy sampling
  • Pick pilot for parallel run
3
Days 61 - 90
Prove + decide
  • Compare pilot vs. baseline
  • Hours / story-point, PM effort, CR TAT
  • User acceptance survey
  • Go / no-go for Wave 2
// Reusable across every client engagement. Same scaffolding, same guardrails, agents swapped to fit the domain. The goal isn't 26 agents - it's a delivery operating system.
// § 06 - contact

SEND A
REAL BRIEF.

No "quick chat to pick your brain." No "exposure." Bring a problem and I'll bring an opinion - and probably a 90-day plan.