🚆 Business Case · May 2026

The AI-native Rail Operating System

Rail delay cascades from one train to thousands of passengers. Unplanned track failures disrupt hundreds of services. Traction energy is 30-40% of operating cost. RailOS deploys 13 AI agents for train performance, infrastructure intelligence, and passenger service.

13 AI AgentsRSSB Safety StandardsORR CompliantFor TOCs · Network Managers · Rail Freight

Open Live Dashboard ARTlligence ↗

92.4%

On-time performance — AI delay prediction 20-40 minutes ahead enables control room intervention before cascades

+8pts

OTP improvement — AI disruption management plans recovery scenarios in minutes not hours

−47%

Infrastructure failures — AI track and signalling monitoring predicts faults weeks before service impact

−22%

Energy cost reduction — AI eco-driving advisory and regenerative braking optimisation

The Problem

Why this sector needs AI-native infrastructure

🚆 Performance: Delays Cascade Rapidly

A 5-minute delay at one station becomes 30 minutes at the terminus as subsequent trains queue. AI delay prediction 20-40 minutes ahead enables proactive intervention — letting late trains through, adjusting platform allocation, and managing connections.

🔧 Infrastructure: Unplanned Failures Cost Millions

A broken rail or signalling failure can disrupt an entire route for hours — affecting thousands of passengers and triggering significant compensation obligations. AI infrastructure monitoring detects failure precursors weeks ahead.

👥 Passenger: Information During Disruption

80% of passenger dissatisfaction during disruption relates to information quality — not the disruption itself. AI passenger intelligence provides real-time, accurate, and personalised disruption information across all channels.

⚡ Energy: 30-40% of Operating Cost

Rail traction energy is the largest controllable cost. AI eco-driving advisory and regenerative braking optimisation reduces energy consumption 22% — without affecting schedules.

🦺 Safety: SPAD and Near-Miss Learning

Safety Performance Assessment Diagrams (SPADs) are among rail's most serious safety events. AI safety intelligence monitors SPAD risk factors, analyses near-miss data, and identifies systemic risks before they result in incidents.

📊 Revenue: Yield Unoptimised

Rail fares are set seasonally and rarely optimised dynamically. AI yield management identifies revenue opportunities within regulatory constraints — improving revenue per train mile.

AI Agent Capabilities

Every function. A specialised agent.

Operations

🚆 Train Performance AI

Delay prediction, real-time performance, headway management.

Infrastructure

🔧 Infrastructure Intelligence AI

Track, signalling monitoring, failure prediction 2-4 weeks.

Passenger

👥 Passenger Intelligence AI

Real-time information, crowd management, complaint prediction.

Disruption

⚡ Disruption Management AI

Recovery scenarios, timetable adaptation, rebooking.

Environment

🌿 Energy Intelligence AI

Eco-driving advisory, regenerative braking, carbon reporting.

Safety

🦺 Safety Intelligence AI

SPAD monitoring, near-miss analysis, safety trend detection.

Commercial

📊 Revenue Intelligence AI

Yield management, pricing, season ticket analytics.

RailOS — advisory intelligence across every capability. Every recommendation requires human approval. Every decision is logged and explainable.

— Built by ARTlligence on the 10-component architecture · Temporal · RAGAS · Langfuse · NeMo

Financial Impact

Measurable value from Day 1

On-Time Performance

92.4% (+8pts)

AI prediction

Infrastructure Failures

−47%

Predictive maintenance

Energy Cost

−22%

AI eco-driving

Passenger Satisfaction

+1.1pts

AI service

Compensation Claims

−34%

OTP improvement

Responsible AI

Advisory intelligence — humans decide

🦺

Safety: SPAD absolute priority

All SPAD risk signals result in immediate operational review. Safety management system requirements are non-negotiable. ORR-notifiable events handled per regulatory process.

🚆

Operations: control room authority

All operational decisions remain with control room managers. AI provides intelligence and recommendations — controllers act.

📊

Revenue: ORR fare compliance

All pricing within ORR-regulated fare caps. Revenue intelligence optimises within regulatory constraints.

Implementation

Operational in 10 weeks

Phase 1 · Week 1–2

Foundation

Train telemetry integration

Infrastructure sensor feeds

Passenger system connection

Safety management system

Phase 2 · Week 3–4

Operations

Train Performance AI live

Infrastructure Intelligence active

Safety Intelligence deployed

Disruption management

Phase 3 · Week 5–7

Passenger & Energy

Passenger Intelligence live

Energy Intelligence active

Revenue Intelligence deployed

Real-time information

Phase 4 · Week 8–10

Full Platform

ORR compliance dashboard

Safety reporting

Energy reporting

Executive dashboard

Market Opportunity

A sector under transformation — now

$3.4B

market size 2025

24.6%

annual growth rate (CAGR)

UK rail: Great British Railways transformation, ETCS/ERTMS digital signalling, £96B Network Rail investment programme. Every TOC, Network Rail, and HS2 faces triple pressure: punctuality (92% OTP target), safety (ORR zero tolerance), and decarbonisation (net zero traction by 2040).

Compliance Framework

Every regulation built in — not retrofitted

Railways Act 1993 / RTSA 2003

ORR regulatory framework. Safety performance reporting and safety case obligations for TOCs.

ROGS 2006 — Safety Verification

Any change to the railway system including AI must be safety-assessed under Common Safety Method.

CSM-RA EU 1158/2010

Common Safety Method for Risk Assessment. Changes assessed via hazard analysis.

RSSB TSRS

UK rail safety standards. RailOS generates SPAD risk data aligned with RSSB TPWS and AWS.

ORR Safety Reporting

SPAD events, near-misses, and precursor indicators reported. AI automates data collection.

Full ROI Model

Financial impact — line by line

Value Driver	Financial Model
OTP +8 points to 92.4%	1 point OTP = £3.5M/yr per 250-train operation. +8 points = £28M/yr.
Infrastructure Failure −47%	100 failures/yr × £150K = £15M. AI reduces 47% = £7M/yr saved.
Energy −22%	Traction energy 35% of TOC OPEX. £70M energy: 22% = £15.4M/yr.
Delay-Repay Reduction	£45M/yr Delay-Repay payments. OTP improvement reduces 34% = £15.3M/yr.
3-Year NPV (250-train UK TOC)	Year 1: −£500K. Year 2: +£45M. Year 3: +£52M. NPV @ 3.5%: £90M. Payback: 5 months.

Competitive Landscape

Why not the alternatives?

Alternative	Limitation	Gap vs ARTlligence
Network Rail Intelligent Infrastructure	NR-only infrastructure data — no TOC operations, no revenue, no passenger intelligence.	NR-only
Siemens Railigent	Siemens rolling stock only — no cross-fleet, no OTP, no passenger, no energy AI.	Fleet-specific
Thales IntelliTRAIN	ETCS/signalling focus only — no operational intelligence, no revenue, no energy.	Signalling only

Integration Map

Connects to your existing stack

TRUST (NR performance system)DARWIN (real-time train running)ATOS Genus (rolling stock health)Sentinel (workforce competency)Darwin API (passenger information)Tyrell (crew management)PARS (passenger assistance)Smart ticketing APIs (Trainline/ATOC)

Risk Register

Top implementation risks — and mitigations

Risk	Level	Mitigation
CSM-RA safety assessment	High	RailOS advisory design minimises change scope. CSM-RA template assessment provided.
ORR regulatory engagement	High	ORR engagement strategy provided. Phased deployment starting with non-safety-critical functions.
DAS regulatory overlap	Medium	Energy eco-driving positioned as planning advisory, not real-time control.
TRUST data quality and delays	Medium	TRUST data has known latency. RailOS buffers and validates before operational recommendations.

Lowest-risk way to start: PoV Sprint

4-week PoV Sprint: Deploy Train Performance AI + Energy Intelligence against 4 weeks TRUST/DARWIN data for one route. Investment: £40,000.

4 weeks

to measurable results

£30–60K

PoV investment

Go/No-Go

before full commitment