Rene Charbonneau · AI & Digital Transformation Leader

Agentic MRO. AI agents on the shop floor, so the ramp funds itself.

  • Claude Certified Architect, I build multi-agent systems at work and at home
  • Built FAA-signed maintenance & in-flight systems at Air Canada

For FTAI (NASDAQ: FTAI) that means one thing: digitize the Maintenance, Repair & Exchange (MRE) shop floor so turnaround time drops, parts cost drops, and the module ramp funds itself. Below is how I read the maintenance machine, and what I would do.

Let's talk About Me

The business, then the operator.

01 · Understanding the business

FTAI sells time on wing, not shop visits.

First, proof I read your machine the way you do. Then what I would build on it.

FTAI pre-builds serviceable engine modules and swaps them onto an airline's engine in weeks, not a 120 to 270 day shop visit. Parts are about 80% of a shop-visit bill, so FTAI goes straight at that line: it reuses good parts from retired engines and fits FAA-approved alternatives that cost 30 to 70% less than buying new from the manufacturer. That swap-not-repair model is its MRE business.

Scale compounds. More engines lower the cost per module, which wins more airlines, which fills more managed aircraft, which buys more engines. PMA parts pull the unit cost down at every turn.

  1. Own the iron: 1,000+ CFM56 and V2500 engines.
  2. Manufacture green time in the Module Factory.
  3. Exchange rather than repair, airlines swap in weeks.
  4. Captive demand: the managed fleet feeds MRE contracts.
  5. PMA parts run 30 to 70% below OEM, lowering unit cost and looping back to step one.
The core of the P&L

Aerospace Products (MRE)

The Module Factory. Module-level swaps measured in weeks. An owned, managed shop network across Montreal, Miami and Rome.

MRE · swap in weeks, not months
Asset ownership

Aviation Leasing

A portfolio of owned and managed assets that feeds the shops and the exchange pool with a steady supply of engines.

290 assets · 47 aircraft / 243 engines
Third-party capital

Asset Management (SCI)

FTAI's Strategic Capital Initiative (SCI) brings in institutional capital to buy aircraft. FTAI earns management fees, co-invests, and captures captive MRE demand from the fleet it services.

Fees, co-invest, and captive maintenance demand
1,050
2026 goal
Module output +39%
$1.625B
2026 goal
Segment EBITDA guidance
Aerospace $1.05B + Leasing $575M
12%→25%
Goal
MRE share of the ~$25B CFM56 market

Sources: FTAI 2026 segment EBITDA guidance, IR 8-K (Apr 29, 2026), reaffirmed Q1 2026; 1,050-module (+39%) and 25%-share targets per BTIG, Seeking Alpha and Yahoo Finance coverage of FTAI guidance.

02 · The MRO margin machine

The maintenance machine, made smarter.

FTAI's MRE business is what funds the company while everything else scales. It runs on the same CFM56 platform and the same shop network, and it is where AI on the shop floor turns directly into margin. FTAI's Power segment matters too, but the near-term margin win is here on the maintenance floor.

Running the shop smarter

The maintenance margin machine

Palantir's Artificial Intelligence Platform (AIP), in a partnership signed November 17, 2025, puts scheduling, inventory and parts procurement on one operating picture. That efficiency is how FTAI targets 1,050 modules in 2026 without adding 1,050 modules' worth of headcount, and how it presses its 12% share of the roughly $25B annual CFM56 market toward 25%.

Operating leverage. More output from the same bench, and the margin that funds everything else.

The shop floor is still part spreadsheet. Digitize it into live records, put AI agents on top, and the 2026 goals get hit without adding headcount.

AI Agents to Transform FTAI

Foundational, builds the base others depend on Leverage, more output without new heads Reusable, redeploys across shops and assets

Scope the engine before it lands

Data Ingestion AgentBorescope Vision AgentWork-Scope Orchestrator
Fixes work scope set too late: today the repair plan is only set once the engine lands, so parts get ordered late, stretching turnaround (TAT) and racking up grounded-aircraft (AOG) cost.
Moves: turnaround time → the 1,050-module ramp
Foundation & workflow

Stream config, utilization and borescope data into the engine ontology before off-wing. The work-scope planner moves from reacting at induction to planning ahead of arrival.

L2 DataL4 KnowledgeDigital twin
Agents, working together
Data Ingestion AgentStreams config, utilization and borescope feeds pre-arrival
Foundational
Borescope Vision AgentReads inspection imagery for damage and wear signals
Reusable
Work-Scope OrchestratorDrafts the scope and parts kit before the engine lands
Leverage

The orchestrator directs the two specialists, then hands a draft scope to the planner for sign-off.

Make every part a live, traceable record

Serialized-Parts TrackerTraceability Validator
Fixes life-limited parts that live in documentation, not databases: each life-limited part (LLP) needs back-to-birth traceability and cycles-remaining, and scattered documents cost hours and risk findings.
Moves: turnaround time → fewer findings delays
Foundation & workflow

Make every serial number a live record: history, location, cycles, next event. The records and traceability workflow stops being a manual hunt across systems.

L2 DataL4 Knowledge
Agents, working together
Serialized-Parts TrackerTurns each serial into a live back-to-birth object
Foundational
Traceability ValidatorFlags missing paperwork and cycle gaps before they become findings
Leverage

Put three shops on one system

Work-Order NormalizerCross-Site Capacity AgentReconciliation Orchestrator
Fixes three shops on three systems: Montreal, Miami and Rome each came in by acquisition with their own work-order conventions, so capacity and TAT are not comparable.
Moves: cross-site capacity → 25% market share
Foundation & workflow

One standard for how a shop visit is recorded across all three sites. The multi-site operations workflow gets a single, comparable picture of capacity and turnaround.

L1 InfraL2 DataL4 Knowledge
Agents, working together
Work-Order NormalizerMaps each site's conventions to one standard schema
Foundational
Cross-Site Capacity AgentCompares TAT and capacity across sites in real time
Leverage
Reconciliation OrchestratorResolves entity mismatches across the three systems
Reusable

Pre-position parts before induction

Findings Forecast AgentUSM Sourcing Agent
Fixes reactive parts buying: without pre-condition signals, used serviceable material (USM) gets bought late and dear, and scope creep adds labor.
Moves: parts cost → $1.625B segment EBITDA
Foundation & workflow

Forecast findings from engine history and source material ahead of induction. The procurement workflow shifts from reactive to pre-positioned.

L2 DataL3 Model
Agents, working together
Findings Forecast AgentPredicts likely findings from engine history and imagery
Leverage
USM Sourcing AgentPre-positions used serviceable material before arrival
Reusable

Give Power a recurring aftermarket

Turbine Uptime MonitorService-Trigger Agent
Fixes FTAI Power having no aftermarket yet: a turbine sold once, with no recurring service, leaves margin on the table.
Moves: a new recurring-service margin on Power
Foundation & workflow

Stand up the service-and-parts and uptime model alongside first delivery. The aftermarket workflow applies the module-maintenance model to power assets from day one.

L4 Digital twinL7 Domain agents
Agents, working together
Turbine Uptime MonitorTracks fielded-unit telemetry against a health model
Reusable
Service-Trigger AgentTriggers module maintenance the way it would for an engine
Leverage

The 8-level AI landing architecture (Levels 1 to 8)
Layer 7 to 8Experience

What the floor uses every day: copilots, embedded agents, decision tools.

L8AI Applications: approval inbox, exception dashboardL7Domain agents: work-scope, procurement, asset lifecycle
Layer 5 to 6Intelligence

The working brain: where models reason, retrieve, and stay safe.

L6Guardian & compliance (airworthiness, audit)L5Agentic runtime & orchestration
Layer 1 to 4Foundation

The structural base. Get this wrong and nothing above scales.

L4Knowledge layer: engine ontology + digital twinL3Model foundationL2Data foundation: engine, parts, borescopeL1Infrastructure & security

Reference architecture adapted from an end-to-end AI landing stack (Levels 1 to 8). Own the foundation, own the outcome. Agents at the top are only as good as the data and digital twin beneath them.

I am that operator.

Rene Charbonneau · Senior AI & Digital Transformation Executive · Montreal

03 · How I fit

The rare operator for this exact work.

I combine deep airline operations, Fortune 50 systems transformation, FAA-regulated delivery, and hands-on AI architecture. I can turn FTAI's shop floor digital faster and more completely than a conventional hire. I am in Montreal, where the shops are.

Airline & regulated systems
  • 10+ years in the airline industry
  • Built maintenance and in-flight aircraft applications at Air Canada that required FAA sign-off before use
  • Fluent in on-time performance, turnaround margins and MRO pressure
Enterprise transformation
  • Fortune 50 ERP (enterprise resource planning) rewrite plus full change management
  • Digital transformations for Fortune 500 companies; cross-vendor data migration
  • $18M+ P&L, 8 units, ~150 people, 6 countries; multi-site scale across acquired shops
  • Air Canada: migrated 60+ applications to Amadeus across 10+ vendors. Built teams and systems across all areas of operations: maintenance, in-air, ramp and airport. Grew the team 6 to 75 in five months
AI builder
  • Claude Certified Architect, one of the first in the world
  • Delivered AI workshops to Microsoft and multiple client business units
  • Stood up data foundations for production AI systems and enterprise data platforms consolidated from multiple acquisitions
  • Audit-grade delivery: CISA certified; SOC 2 and regulatory reporting at a broker-dealer; 25% timeline compression

Where I've already done each piece
Each thing FTAI needs, what I bring, and where I have already done it.
What FTAI needsWhat I bringTrack record
Audit-grade discipline. Federal Aviation Administration (FAA)-regulated shops, investor reporting on SCI Regulated delivery where the books get inspected Certified Information Systems Auditor (CISA); built SOC 2 and regulatory reporting at a broker-dealer
Run multi-site scale to a P&L Operator accountability, not advice $18M+ P&L, 8 units, ~150 people, 6 countries
One standard across acquired shops; move data between vendors Large-scale aviation systems work Air Canada: migrated 60+ applications to Amadeus across 10+ vendors. Built teams and systems across all areas of operations: maintenance, in-air, ramp and airport
Take cost and time out of the shop, without adding heads Delivery that compounds, priced and measured Cut delivery timelines 25%; pricing models with auditable assumptions
Digitize the floor and stand up the data foundation I build it, so it runs rather than stays theoretical Data foundations for production AI; Databricks, Snowflake, Fabric, Purview, FinOps
Someone on the ground Already here Montreal-based
Proof I build, not just advise

Code is conducted, not written.

I run a personal multi-agent build system at home that ships real, tested software. An orchestrator directs specialist agents, an independent agent does QA, and a governed harness handles tests, parallel work and deployment. I do not talk about this work from the outside. I operate it daily.

How my multi-agent build system is wired, and what each part does
One orchestrator, a fleet of specialist agents, and a governed harness underneath.

See how it's wired: orchestrator, agents, governed harness
Human operator Sets goals, approves plans, breaks ties
Orchestrator Routes every request and dispatches all agents, then sequences QA after the architects Claude Code · Sonnet
spawns specialist agents
Strategy

Cross-project strategy and the scope-to-retro delivery pipeline

Claude Code · Opus
Advisors

Technical, product, security, knowledge and growth briefs

Claude Code · Opus
Project architects

End-to-end delivery in-fork: plan, build, verify, open the PR

Claude Code
Independent QA

A fresh fork per PR that reviews the diff with no inherited context

Claude Code · Sonnet
Infrastructure

Home network, container media stack, DNS and device onboarding

Claude Code + MCP
Release engineering

Deploy pipeline, service supervision and the runtime substrate

Claude Code + launchd
Always-on ops

Read-only fleet health, log tails and ticketing from the host

OpenClaw gateway
Session analyzer

Classifies sessions and tracks cost and process compliance

Python CLI
Governed harnessHook guardrails · parallel git worktrees · tests · deploy · audit trail
Connector planeGitHub · Jira · Playwright · CI/CD · MCP servers
Home infrastructureAlways-on host · OrbStack containers · AI runtime (Hermes)

My first 90 days
First 90 days · 01

Walk the floor

Sit with Montreal, Miami and Rome. Find where turnaround and parts cost actually leak.

First 90 days · 02

Digitize one shop visit

Take a single work-scope or procurement step, put data ahead of induction, and measure turnaround and parts cost against baseline.

First 90 days · 03

Make it repeatable

Turn the win into a standard play with the math that shows it pays for itself, then sequence it against the ramps.

Self-funding: the play pays for itself before it scales.

Let's talk

Bring the operator in before the ramp, not after.

Rene Charbonneau Let's talk