Enterprise Architecture · TOGAF 9.2 · IT Governance
M&A IT integration — merging two cargo terminals
Eight visual modules for post-merger IT work. Every idea is framed with aviation-cargo analogies from 13 years at Cathay Pacific Cargo Terminal and 20 years of enterprise architecture delivery.
Mayank ChughDue Diligence → Day 1 → PMITSA · COBIT · ITIL · AI governance
Module 01
Three gates, one analogy: merging two cargo terminals
Due Diligence finds the liabilities. Day 1 keeps the planes flying. Post-Merger Integration decides what survives long-term.
1
Due diligence = pre-merger cargo audit
Before two airlines combine cargo operations, you audit both terminals.
What warehouse systems do they run?
Are any licenses non-compliant?
Is there hidden technical debt nobody documented?
Customs and hazardous-cargo compliance risks?
AI-era: were models trained on data they had no rights to use?
2
Day 1 = one combined ramp
The terminals must operate as one from the moment the deal closes — not the final state, just the minimum to run flights together.
Shared staff badges and access
One unified security perimeter
One set of operational directories
AI-era: push one combined acceptable-use policy, like a unified safety briefing
3
PMI = one warehouse system
Over months, decide which systems survive. That is application rationalization.
Which WMS becomes the standard?
Which gets decommissioned?
Which tracking tool is kept temporarily under a TSA?
Due Diligence finds the liabilities. Day 1 keeps the planes flying. Post-Merger Integration decides what survives long-term.
TSA explained: borrowing the other airline’s ground handling
Key ideaA Transition Service Agreement is a temporary, paid arrangement where the acquired company keeps using the seller’s IT after close — because you cannot build a replacement overnight. You do not own it. You are renting continuity, on a countdown clock.
Concept
Meaning
Cargo terminal example
TSA
Temporary reliance on the seller’s systems
Renting the old airline’s cargo scanners
TSA exit
Deadline to be fully independent
Your own scanners installed and certified
TSA cost
Usually monthly, often above market rate
An airport gate lease at a premium
TSA risk
Missed exit = extra cost + business risk
Flights delayed if your systems are not ready
A TSA is not a solution — it is a clock. The real work is building your own systems before it runs out.
Day-1, Day-2, target state: three different finish lines
Key distinctionDay-1 readiness is not the finished airline. It is the minimum bar to legally and operationally open the gates.
Day 1 · close the deal
Bare minimum
Combined badges work, one security perimeter, nothing catastrophic breaks. Systems underneath can still be two setups duct-taped together — planes still get loaded.
Day 2 · weeks to months
Stabilization
Pain points fixed, TSA processes stabilized, teams start operating as one. Ground crew drop two labeling systems and adopt one interim standard.
Target · 9–18+ months
Destination architecture
TSAs exited, retain/replace/decommission executed, one unified platform. No more renting the old partner’s equipment.
Confusing these three stages is the most common PMI mistake — chasing target-state quality on a Day-1 timeline is how deals fail.
Dependency mapping: why you cannot migrate everything at once
Cargo analogyYou could not cut over warehouse management before the customs interface was ready, and you could not switch billing before both were stable. Hong Kong and Singapore may both depend on the same regional customs gateway — so their cutover dates are linked whether you like it or not.
Wave 1
Foundations
Directories, network, security. Everything else waits on these.
Wave 2
Market cores
Market-specific systems with no cross-market dependencies. Can migrate in parallel, market by market.
Wave 3
Shared platforms
Systems that depend on a regional platform. Wait until that platform is stable across every market that uses it.
Wave 4
Reporting
Analytics layers that need all upstream systems already migrated and stable.
Risk flag. Migrating Wave 3 before the shared platform is ready does not just risk one market — it risks every market that depends on it. Sequencing, not individual system readiness, sets the real timeline.
Coordinating HK and offshore teams on one airline’s systems is the same discipline as sequencing across markets in a merger — just with more countries on the map.
COBIT 2019 in one picture: the control tower vs the ramp
COBIT splits IT decision-making into two layers that must not be confused.
Governance Evaluate · Direct · Monitor — the control tower
Owned by the board and executives. Set direction, choose priorities, watch whether outcomes match intent. Air traffic control does not load cargo — it decides which flights get priority and whether operations are safe.
Management Plan · Build · Run · Monitor — ramp operations
Owned by product, delivery, and engineering. Plan the work, build systems, run the day, monitor performance. The ramp crew does not set airline strategy — they execute within the direction the tower has set.
TOGAF plug-inPhase G (Implementation Governance) and Phase H (Architecture Change Management) are where an EA’s work plugs into COBIT’s governance layer — architecture decisions are evaluated and directed at governance, then executed through management.
Confusing these two layers is a classic failure — like the control tower trying to load cargo, or the ramp crew trying to set airspace policy.
Restore service as fast as possible, even with a workaround. Tracking system down mid-shift: restart the server, switch to a manual backup, get flights loading. You do not need the why yet.
Problem — why it keeps happening
Find and eliminate the root cause. Same tracking system crashed three times this month — memory leak, bad update, or architecture flaw? Fix the cause, not the symptom.
Change — control what gets modified
Assessment, approval, scheduling. The fix is tested and deployed in a controlled window — not pushed live during peak cargo season with nobody signing off.
Process
Question it answers
Timescale
Incident
How do we get it working again, right now?
Minutes to hours
Problem
Why does this keep happening?
Days to weeks
Change
How do we safely deploy the fix?
Scheduled, controlled
SLAs, RPOs, and RTOs — already familiar from enterprise architecture — are the metrics that govern all three processes.
Three tiers of enterprise AI: from public tools to your own vault
Tier 1
Public AI — open tarmac
ChatGPT web, public Claude, no enterprise wrapper. Proprietary code and client data must never be pasted in. A public airport lounge: useful, but you would never leave a client’s cargo manifest on the table.
Tier 2
Private enterprise AI — your own gate
Corporate-wrapped models (Azure OpenAI, Bedrock) in an isolated cloud. Inputs never train the public model. Most day-to-day usage belongs here. A private airline gate: controlled access, still on shared airport infrastructure.
Tier 3
Custom / RAG — your warehouse, your rules
Domain vector stores mirroring the same permission levels as your relational databases. Highest control, used for sensitive proprietary retrieval. Your own warehouse, with security clearances mirrored onto who can access which shelf.
AI FinOps
One enterprise API gateway rate-limits and cost-attributes usage per department — like a single fuel account that tracks which flights burned which fuel.
Shadow AI mitigation
An open AI service catalog plus a 48-hour review gives teams a fast sanctioned path — so nobody sneaks in an unauthorized tool, the way an unsanctioned baggage cart bypassing security would.
Tier 2 and 3 (isolated environments, permissioned vector retrieval) are the same principles as self-directed RAG work, scaled to enterprise governance.