Who leads AI at Rio Tinto?

AI at Rio Tinto is led by Daniel Evans (Chief Information Officer); Scott Brown (Chief Information Security Officer). Ownership of AI on the org chart: No named AI executive; IS&T under CIO Daniel Evans, data science in the product groups.

What AI platform does Rio Tinto use?

Rio Tinto builds AI applications on Palantir Foundry, whose enterprise agreement Rio Tinto extended in Nov 2024 to include Palantir's AI Platform (AIP), and on AWS services such as Amazon Bedrock and SageMaker; ReconAI, Rio Tinto's rail fault classifier, reads data collected in Foundry and shows its results there (Jun 2025).

Which AI models does Rio Tinto use?

Rio Tinto's ReconAI calls Anthropic Claude models, mainly Sonnet, through Amazon Bedrock (Jun 2025); a Rio Tinto aluminium team fine-tuned Meta's Llama 3.1 8B on Amazon SageMaker (May 2026); and Rio Tinto finance teams use Microsoft Copilot Studio and Copilot in Power BI (May 2026).

How many Rio Tinto employees use AI?

Rio Tinto has not said how many of its roughly 61,000 employees use AI tools; a finance shared services manager said in May 2026 that teams use Copilot at an increasing rate, and ReconAI serves maintenance and reliability engineers on the Pilbara rail network.

What return has Rio Tinto reported from AI?

Rio Tinto has reported no business return it credits to AI since May 2025: its 2025 annual report on Form 20-F, its Dec 2025 Capital Markets Day and its Jul 2026 half-year results put no cost, headcount or revenue figure on its own use of AI, and the $870 million of productivity benefits banked in the first half of 2026 is credited to production uplifts, cost savings and cost efficiencies, not to AI. The nearest figures are a task speedup (engineers trace dependencies in a legacy aluminium system in minutes rather than days, May 2026), a classification accuracy of 96% for ReconAI (Jun 2025) and a BCG X case study claiming that Rio Tinto's Future Scheduling Platform more than doubled scheduler productivity, which is a vendor's claim. Rio Tinto has not reported a company-wide AI rollout either: it gives no count of employees using AI tools and names no executive who owns AI. Rio Tinto's AI work since May 2025 is on the public record, from ReconAI in production (Jun 2025) and a 5-year Ideon partnership (Oct 2025) to a memorandum with AI Singapore (Mar 2026), so Rio Tinto sits at Experimenting rather than Unknown. A cost, production or headcount return Rio Tinto credits to AI in one use case, or a stated count of employees using AI tools under a named owner, would move Rio Tinto to Piloting.

Verified Nov 5, 2026

Rio Tinto

Experimenting

Mining
~61,000 employees
Palantir Foundry and AIP
Claude and Llama on AWS
No AI-credited return reported

Rio Tinto deploys AI through its IS&T technology function and data science teams in the product groups, building tools on Palantir Foundry and AIP and on AWS, where ReconAI calls Anthropic's Claude through Amazon Bedrock to classify locomotive faults and an aluminium team runs a fine-tuned Llama model. Finance staff use Microsoft Copilot Studio, and a Mar 2026 agreement with AI Singapore targets freight invoices. Rio Tinto has reported no AI user count and no return it credits to AI since May 2025.

riotinto.com

Their approach to AI

The platform: Rio Tinto builds AI applications on Palantir Foundry, whose enterprise agreement Rio Tinto extended in Nov 2024 to include Palantir's AI Platform (AIP), and on AWS services such as Amazon Bedrock and SageMaker; ReconAI, Rio Tinto's rail fault classifier, reads data collected in Foundry and shows its results there (Jun 2025).
The models: Rio Tinto's ReconAI calls Anthropic Claude models, mainly Sonnet, through Amazon Bedrock (Jun 2025); a Rio Tinto aluminium team fine-tuned Meta's Llama 3.1 8B on Amazon SageMaker (May 2026); and Rio Tinto finance teams use Microsoft Copilot Studio and Copilot in Power BI (May 2026).
The reach: Rio Tinto has not said how many of its roughly 61,000 employees use AI tools; a finance shared services manager said in May 2026 that teams use Copilot at an increasing rate, and ReconAI serves maintenance and reliability engineers on the Pilbara rail network.
The catch: Rio Tinto reports no AI spending, no AI user count and no business return it credits to AI since May 2025; its productivity program, $870 million banked by Jun 2026, is credited to operational and cost improvements, and most detail on its AI tools comes from staff talks at vendor events.

Snapshot

Employees: ~61,000 (2025 average); 55,572 employees at managed operations on Dec 31, 2025, plus 4,092 contractors (2025 Form 20-F)
Technology budget: Not disclosed; Group-wide technology and systems spending sits inside $0.6 billion of 2025 restructuring, project and one-off costs, without its own figure
AI/ML headcount: Not disclosed; A central machine learning capability set up around 2017, with data scientists in Brisbane, Perth, Singapore, Montreal and India (iTnews, Jul 2024)
Who owns AI: No named AI executive; CIO Daniel Evans leads IS&T, Rio Tinto's technology function; data science teams sit in the product groups
Platform: Palantir Foundry and AIP, plus AWS; Foundry since about 2021, AIP added in Nov 2024; AWS Bedrock and SageMaker for in-house tools
Models: Claude, Llama, Microsoft Copilot; Claude Sonnet via Amazon Bedrock (Jun 2025), Llama 3.1 8B fine-tuned on SageMaker (May 2026), Copilot Studio in finance (May 2026)
Reach: Not disclosed; No count of employees using AI tools; named users are finance, rail maintenance and aluminium engineering teams
Use cases: Several in production; Rail fault classification, a legacy system assistant, rail and port scheduling, finance Copilot, IT service AI, employee survey analysis

Key findings

Rio Tinto's ReconAI classifies faults on the autonomous locomotives that haul Pilbara iron ore to port: senior data scientist Milan Pareek said in Jun 2025 that ReconAI calls Anthropic Claude models, mainly Sonnet, through Amazon Bedrock, reads rail and SAP data collected in Palantir Foundry on AWS, and reached 96% accuracy in production, and that Rio Tinto is extending ReconAI to ore cars, signalling and port conveyors.
Rio Tinto renewed its Palantir enterprise agreement for 4 more years in Nov 2024 and added Palantir's AI Platform (AIP) to the Foundry deployment it uses for plant operations, geotechnical risk at Oyu Tolgoi and coordination of driverless iron ore trains.
Rio Tinto's Aluminium Pacific Operations built an assistant on a fine-tuned Llama 3.1 8B model, Amazon Bedrock Knowledge Bases and Amazon Bedrock AgentCore to document Metpro, a 30-year-old manufacturing execution system, and senior data science advisor Ke Shi said in May 2026 that engineers now trace the effect of a change in minutes rather than days.
Rio Tinto signed a 2-year memorandum with AI Singapore in Mar 2026 to build AI tools for its supply chains, starting from an AI-assisted process for the more than 10,000 freight invoices Rio Tinto handles each year, and its finance shared services use Microsoft Copilot Studio and Copilot in Power BI (May 2026).
Rio Tinto has reported no AI spending, no count of employees using AI tools and no business return it credits to AI since May 2025; the $870 million of productivity benefits it banked in the first half of 2026 is credited to production and cost improvements, and AI appears in Rio Tinto's 2025 Form 20-F mainly as a cyber and compliance risk and as a source of metals demand.

Where Rio Tinto sits

Stage: Experimenting
Rio Tinto has reported no business return it credits to AI since May 2025: its 2025 annual report on Form 20-F, its Dec 2025 Capital Markets Day and its Jul 2026 half-year results put no cost, headcount or revenue figure on its own use of AI, and the $870 million of productivity benefits banked in the first half of 2026 is credited to production uplifts, cost savings and cost efficiencies, not to AI. The nearest figures are a task speedup (engineers trace dependencies in a legacy aluminium system in minutes rather than days, May 2026), a classification accuracy of 96% for ReconAI (Jun 2025) and a BCG X case study claiming that Rio Tinto's Future Scheduling Platform more than doubled scheduler productivity, which is a vendor's claim. Rio Tinto has not reported a company-wide AI rollout either: it gives no count of employees using AI tools and names no executive who owns AI. Rio Tinto's AI work since May 2025 is on the public record, from ReconAI in production (Jun 2025) and a 5-year Ideon partnership (Oct 2025) to a memorandum with AI Singapore (Mar 2026), so Rio Tinto sits at Experimenting rather than Unknown. A cost, production or headcount return Rio Tinto credits to AI in one use case, or a stated count of employees using AI tools under a named owner, would move Rio Tinto to Piloting.
How the stages are defined
Ownership: No named AI executive; IS&T under CIO Daniel Evans, data science in the product groups
Platform: Hybrid: Palantir Foundry and AIP plus AWS services; ReconAI and other tools built in-house
Models: Anthropic Claude via Amazon Bedrock, Meta Llama 3.1 8B, Microsoft Copilot
Hosting: AWS (Bedrock, SageMaker, EC2, S3) and Palantir Foundry on AWS; Microsoft and ServiceNow cloud services; own data centers not disclosed
Reach: Not disclosed; finance, rail maintenance and aluminium engineering teams

How they deploy

AI inside a productivity drive

Rio Tinto's strategy under Simon Trott, Chief Executive since Aug 2025, is to become stronger, sharper and simpler: 3 product groups, a leaner centre and a productivity program that reached a $1.3 billion annualised run rate in the first half of 2026, with a target of $1.8 billion by the end of 2026. At the Dec 4, 2025 Capital Markets Day, Rio Tinto listed simplifying digital applications and support costs among the early productivity actions, and the Aluminium product group listed targeted digital and AI initiatives among the ways it plans to raise returns. Rio Tinto has not credited any part of the productivity program to AI, and Simon Trott's May 2026 address to shareholders did not mention AI.

Data platforms first, then AI

Rio Tinto's AI work rests on data platforms built over several years. Palantir said in Nov 2024 that Rio Tinto, an early Foundry adopter, had spent about 3 years building an ontology, a digital model of its operations, and that AIP would let Rio Tinto build, test and deploy AI use cases on top of it. Chief Commercial Officer Bold Baatar said then that the Foundry ontology had made Rio Tinto's structured data accessible and that AIP was doing the same for unstructured data. Separately, Rio Tinto runs machine learning on Amazon SageMaker, with a central machine learning capability set up around 2017 and work under way in 2024 on a multicloud data lakehouse.

Tools built by practitioners

Most of what is public about Rio Tinto's generative AI comes from engineers and data scientists describing tools they built: ReconAI for rail fault classification (Jun 2025), an assistant for the Metpro aluminium manufacturing system (May 2026) and Copilot Studio use in finance shared services (May 2026), each presented at AWS or Microsoft events and reported by iTnews. Rio Tinto has not described a sanctioned assistant for all employees or a company-wide generative AI program.

Partners for operations and the supply chain

Rio Tinto also brings in outside AI work for specific problems. Ideon Technologies announced a 5-year enterprise partnership in Oct 2025 to model the subsurface at 6 of Rio Tinto's largest operations with a platform that combines muon imaging, geophysical sensing and AI services, after a pilot at the Bingham Canyon copper mine. Rio Tinto's Commercial team in Singapore signed a 2-year memorandum with AI Singapore in Mar 2026, and BCG X says it helped Rio Tinto replace dated Pilbara rail and port scheduling systems with an AI scheduling platform.

AI as a market for metals

Rio Tinto also treats AI as demand. Rio Tinto's 2025 Form 20-F lists AI data centres among new drivers of metals demand, Amazon Web Services became the first customer for copper from Rio Tinto's Nuton bioleaching process in Jan 2026 for use in its U.S. data centres, and Rio Tinto said in Jul 2026 that copper, aluminium and lithium contributed more than half of its first-half underlying EBITDA.

Deployment model: Rio Tinto runs technology through IS&T, its technology function, led by Chief Information Officer Daniel Evans, who was appointed in Dec 2021 after acting in the role from Apr 2021 and is based in Brisbane. Data scientists and citizen data users sit mainly in Brisbane for aluminium, Perth for iron ore, Singapore for the commercial business and Montreal, with a team in India for productionised machine learning (iTnews, Jul 2024). Rio Tinto buys platforms (Palantir Foundry and AIP, AWS, Microsoft Power Platform and Copilot Studio, ServiceNow) and builds its own tools on them, such as ReconAI and the Metpro assistant; in Oct 2022 Rio Tinto began building an in-house software development function that could grow to more than 500 people. The product groups (Iron Ore, Aluminium & Lithium, Copper) and Commercial each run AI work close to their operations, and finance shared services runs its own automation with IS&T.
Model providers: Rio Tinto uses Anthropic Claude models, mainly Sonnet and including Claude 3.5 Sonnet v2, through Amazon Bedrock for ReconAI (Jun 2025). A Rio Tinto aluminium team chose Meta's Llama 3.1 8B through Amazon SageMaker AI JumpStart as the initial model for its Metpro assistant and fine-tuned it on a curated dataset (May 2026). Rio Tinto finance teams use Microsoft Copilot Studio and Copilot inside Power BI (May 2026), and Rio Tinto's IT service desk runs ServiceNow's Now Assist for incident summarisation and AI search (May 2025). Palantir's AIP is part of Rio Tinto's Palantir agreement, but neither company has named the models behind it. Rio Tinto has not described a single model gateway or a policy for choosing models.
Data and infrastructure: Rio Tinto runs its named AI tools on AWS and on vendor clouds. ReconAI's data is collected in Palantir Foundry running on AWS, replicated to Amazon S3 and read with AWS Glue and Athena; ReconAI's backend runs on Amazon EC2 and calls Claude through Amazon Bedrock, with Bedrock knowledge bases synced to SharePoint (Jun 2025). The Metpro assistant uses Amazon Bedrock Knowledge Bases, Amazon Bedrock AgentCore and Amazon SageMaker (May 2026). Rio Tinto's machine learning environments run on Amazon SageMaker Studio and Canvas, and its production machine learning accounts were air-gapped with no internet access by default in Jul 2024. Finance automation runs on Microsoft Power Platform fed from Rio Tinto's SAP system, and IT service management runs on ServiceNow. Rio Tinto has not said whether any AI model runs in its own data centres, and Nuton uses AWS platforms for simulation and analytics at the Johnson Camp mine in the United States (Jan 2026).

Platforms and internal tools

Palantir Foundry and AIPFoundry about 2021; AIP added Nov 2024
Palantir Foundry holds Rio Tinto's ontology, a digital model of its operations used for plant operations, geotechnical risk monitoring at Oyu Tolgoi and coordination of driverless iron ore trains; Rio Tinto's Nov 2024 renewal for 4 more years added Palantir's AI Platform (AIP) to build and deploy AI use cases on the same data. ReconAI's users see its fault classifications through a Foundry front end.
Users
Not disclosed
Source
ReconAIProof of concept mid-2024; in production by Jun 2025
ReconAI classifies equipment faults on Rio Tinto's autonomous locomotives to a causing component and failure mode, using telemetry, SAP maintenance notes and free text from Rio Tinto's HOSDI application, so that engineers can prioritise defect elimination projects. ReconAI runs a retrieval-augmented, agent-based workflow on Amazon EC2 that calls Claude through Amazon Bedrock and shows engineers its full reasoning.
Users
Maintenance and reliability engineers on the Pilbara rail network
Models
Anthropic Claude, mainly Sonnet, via Amazon Bedrock
Source
Metpro domain assistantPresented May 2026
Rio Tinto Aluminium Pacific Operations built an assistant that captures the dependencies and decision logic of Metpro, a 30-year-old manufacturing execution system that runs the aluminium product lifecycle from tapping to shipment, so engineers can check the effect of a change before making it. Agents pull current information so the model does not need retraining for every change.
Users
Engineers who maintain Metpro
Models
Meta Llama 3.1 8B, fine-tuned on Amazon SageMaker
Source
Future Scheduling Platform
The Future Scheduling Platform supports the Integrated Scheduling team in Perth, about 50 schedulers who make rail and port decisions for Rio Tinto's Pilbara iron ore supply chain, with algorithms that combine machine learning, optimisation, simulation and heuristics, according to a BCG X case study. Rio Tinto has not published its own description or launch date.
Users
About 50 schedulers (BCG X)
Source
Amazon SageMaker machine learning environmentsDescribed Jul 2024
Rio Tinto's data scientists and citizen data users build and deploy machine learning models in Amazon SageMaker Studio and SageMaker Canvas, with a machine learning operations team standardising how models reach production; Rio Tinto said in Jul 2024 that models were used for rail track maintenance, smelter safety and mine planning.
Source
ServiceNow Employee Centre with Now AssistEmployee Centre Nov 2024; Now Assist for IT service management May 2025
Rio Tinto's ServiceNow Employee Centre brings HR services and IT support into one portal, and Rio Tinto switched on ServiceNow's Now Assist for incident summarisation, resolution notes and AI search in May 2025.
Source

Vendors and partners

PalantirFoundry about 2021; AIP Nov 2024
Foundry data platform and AI Platform (AIP) · Palantir renewed its enterprise agreement with Rio Tinto for 4 more years in Nov 2024; no contract value was published.
Amazon Web ServicesMachine learning on SageMaker described by Jul 2024
Cloud for ReconAI, the Metpro assistant and machine learning (Bedrock, SageMaker, EC2, S3); Nuton analytics; copper customer · Rio Tinto signed a 2-year agreement in Jan 2026 to supply Nuton copper to AWS and to use AWS services to optimise the Nuton process.
AnthropicNamed Jun 2025
Claude models behind ReconAI, through Amazon Bedrock · Rio Tinto has no announced direct agreement with Anthropic; access runs through AWS.
MetaNamed May 2026
Llama 3.1 8B, the initial model of the Metpro assistant
MicrosoftNamed May 2026
Power Platform, Copilot Studio and Copilot in Power BI for finance shared services
ServiceNowNow Assist live May 2025
Employee Centre and Now Assist for IT service management
Ideon TechnologiesOct 2025
REVEAL subsurface platform combining muon imaging, geophysical sensing and AI services · 5-year enterprise partnership announced by Ideon, starting at 6 of Rio Tinto's largest operations after a pilot at Bingham Canyon.
AI SingaporeMar 2026
AI solutions for supply chain problems, starting with freight invoice processing · 2-year memorandum of understanding facilitated by Enterprise Singapore.
BCG X
Built the Future Scheduling Platform for Pilbara rail and port scheduling, per BCG X

Leadership

AI, data and platform leaders

Daniel Evans

Chief Information Officer

Daniel Evans leads IS&T, Rio Tinto's technology function, as Chief Information Officer, a role he held on an acting basis from Apr 2021 and permanently from Dec 2021 after serving as Rio Tinto's chief information security officer. The 2025 Form 20-F credits Evans with 15 years of cyber security leadership and says the CISO reports to him. Evans said in Oct 2022, when Rio Tinto began building an in-house software development function, that machine learning and AI were among the reasons Rio Tinto is also a technology company.

Scott Brown

Chief Information Security Officer

Scott Brown is Rio Tinto's Chief Information Security Officer, reporting to the CIO, according to the 2025 Form 20-F, which credits him with more than 17 years of cyber security experience. Brown was acting CISO from Apr 2021 until a permanent CISO was appointed in May 2022, and before that managed Rio Tinto's cyber security operations centre (iTnews, May 2022).

Executives

Simon Trott

Chief Executive

Simon Trott became Rio Tinto's Chief Executive in Aug 2025 after leading Iron Ore and, before that, serving as Rio Tinto's first Chief Commercial Officer; Trott joined Rio Tinto more than 25 years ago and has sat on the Executive Committee since 2018 (2025 Form 20-F). Trott's program is to make Rio Tinto stronger, sharper and simpler, with a productivity target of $1.8 billion annualised by the end of 2026. Trott has not spoken publicly about Rio Tinto's own use of AI in the sources reviewed; his Jul 2026 results comments tied AI to metals demand.

Peter Cunningham

Chief Financial Officer

Peter Cunningham joined Rio Tinto in 1993 and has been Chief Financial Officer and an executive director since Jun 2021, after roles including Group Controller and Head of Investor Relations. Cunningham chairs Rio Tinto's Risk Management Committee and is the accountable executive for cyber security risk, a risk the 2025 Form 20-F says is rising with the adoption of automation and AI.

Bold Baatar

Chief Commercial Officer

Bold Baatar has been Rio Tinto's Chief Commercial Officer since Sep 2024, responsible for sales and marketing, procurement, marine, logistics and business development, after leading Copper and Energy & Minerals (2025 Form 20-F). Baatar spoke for Rio Tinto on the Nov 2024 Palantir renewal, saying AIP was opening up Rio Tinto's unstructured data. Rio Tinto's Commercial hub in Singapore signed the Mar 2026 memorandum with AI Singapore.

Matthew Holcz

Chief Executive, Iron Ore

Matthew Holcz became Chief Executive, Iron Ore in Aug 2025 after serving as Managing Director, Pilbara Mines, where he led Rio Tinto's 18 iron ore operations in Western Australia; Holcz joined Rio Tinto in 2007 (2025 Form 20-F). Holcz's product group runs the autonomous trains and haulage, the ReconAI rail tool and the Pilbara scheduling platform.

Jérôme Pécresse

Chief Executive, Aluminium & Lithium

Jérôme Pécresse joined Rio Tinto as Chief Executive, Aluminium in 2023 from GE Renewable Energy, where he was President and CEO, and became Chief Executive, Aluminium & Lithium in Aug 2025 (2025 Form 20-F). Pécresse's product group presented targeted digital and AI initiatives as one way to raise returns at the Dec 2025 Capital Markets Day, and its Pacific Operations built the Metpro assistant.

Katie Jackson

Chief Executive, Copper

Katie Jackson has led Rio Tinto's Copper product group since Sep 2024, after serving as President of National Grid Ventures and holding roles at Shell, UBS, Anadarko, Equinor and BG Group (2025 Form 20-F). Jackson announced the Jan 2026 agreement under which Nuton uses AWS data and analytics services and AWS buys Nuton copper.

AI org map

Central technology function (IS&T) and platform contracts; data science and AI tools built inside the product groups, Commercial and Group Services

Rio Tinto's Chief Executive, Simon Trott, leads a 9-member Executive Committee that includes the heads of the 3 product groups (Iron Ore, Aluminium & Lithium, Copper), the Chief Commercial Officer and the Chief Financial Officer. IS&T, led by Chief Information Officer Daniel Evans, is Rio Tinto's technology function, and the Chief Information Security Officer reports to the CIO; CFO Peter Cunningham is the accountable executive for cyber risk. Rio Tinto has no executive with AI in the title, and the AI tools on the public record were built by data science teams in the product groups, by Group Services finance with IS&T, and by the Commercial team in Singapore.

Rio Tinto has published no count of data scientists, machine learning engineers or AI staff. Rio Tinto said in Jul 2024 that it set up a central machine learning capability about 7 years earlier to serve the product groups, with data scientists in Brisbane, Perth, Singapore and Montreal and a team in India for productionised machine learning, and in Oct 2022 that it would build an in-house software development function of more than 500 people over 5 years. The people named on Rio Tinto's AI tools are practitioners: a senior data scientist in Iron Ore rail, a senior data science advisor in Aluminium, a finance digital strategy manager in Group Services and the Commercial CFO in Singapore.

Reporting lines

Chief Executive
Simon Trott
- Finance
  Peter Cunningham, CFO
  Chairs the Risk Management Committee; accountable executive for cyber security risk
- Iron Ore
  Matthew Holcz, Chief Executive
  Pilbara mines, rail and ports; ReconAI and the Future Scheduling Platform; Executive Committee member
- Aluminium & Lithium
  Jérôme Pécresse, Chief Executive
  Metpro assistant at Pacific Operations; Executive Committee member
- Copper
  Katie Jackson, Chief Executive
  Nuton analytics on AWS; Executive Committee member
- Commercial
  Bold Baatar, Chief Commercial Officer
  Singapore hub; AI Singapore memorandum and freight invoice work; Executive Committee member
IS&T
Daniel Evans, CIO
Rio Tinto's technology function, based in Brisbane; the CIO's reporting line is not published
- Cyber Security
  Scott Brown, CISO
  Reports monthly to the CFO and quarterly to the Risk Management Committee (2025 Form 20-F)
Group Services finance
Global shared services; month-end automation and Copilot Studio with IS&T; reporting line not published
Data science and machine learning
Central machine learning capability set up around 2017, with teams in Brisbane, Perth, Singapore, Montreal and India (Jul 2024); reporting line not published

Rio Tinto's 2025 Form 20-F names the CIO and CISO in its cyber security disclosure and says the CISO reports to the CIO and the CFO is the accountable executive for cyber risk; it does not say to whom the CIO reports. The product group chief executives, the Chief Commercial Officer and the CFO sit on the Executive Committee, which supports the Chief Executive; their direct lines to the Chief Executive are inferred from that. Rio Tinto's executive committee was cut from 11 to 9 members in the second half of 2025.

Where the AI workforce sits

IS&T, including machine learning operations and in-house software development
Rio Tinto planned in Oct 2022 to build a software development function of more than 500 people over 5 years; current size not published.
Not disclosed
Data science teams in Iron Ore, Aluminium & Lithium, Copper and Commercial
Teams sit mainly in Perth, Brisbane, Singapore and Montreal (Jul 2024).
Not disclosed

Named people

Executive sponsors

Simon Trott
Chief Executive · Chief Executive since Aug 2025
Peter Cunningham
Chief Financial Officer · Accountable executive for cyber security risk
Bold Baatar
Chief Commercial Officer · Spoke for Rio Tinto on the Palantir AIP renewal (Nov 2024)

IS&T and cyber security

Not disclosed

Daniel Evans
Chief Information Officer · Leads IS&T, Rio Tinto's technology function
Scott Brown
Chief Information Security Officer · Reports to the CIO (2025 Form 20-F)

Product groups

Matthew Holcz
Chief Executive, Iron Ore · Runs the Pilbara rail, scheduling and autonomous operations where ReconAI works
Jérôme Pécresse
Chief Executive, Aluminium & Lithium · Product group plans targeted digital and AI initiatives (Dec 2025)
Katie Jackson
Chief Executive, Copper · Announced the Nuton analytics agreement with AWS (Jan 2026)

Practitioners named in public talks

Milan Pareek
Senior data scientist and superintendent engineering, Iron Ore rail · Presented ReconAI at AWS Summit Sydney (Jun 2025)
Ke Shi
Data science senior advisor, Aluminium · Presented the Metpro assistant at AWS Summit Sydney (May 2026)
Brandon Hogan
Senior manager, digital strategy, automation and analytics, finance and accounting services · Described Copilot Studio and Power Platform use in finance (May 2026)
Will Millsteed
Commercial Chief Financial Officer · Spoke for Rio Tinto on the AI Singapore memorandum (Mar 2026)

Use cases

Rio Tinto has 8 AI use cases in production on the public record as of Nov 2026, plus 2 pilots and one announced deployment: ReconAI for rail fault classification (Jun 2025), the Metpro assistant in aluminium (May 2026), Copilot in finance shared services (May 2026), ServiceNow Now Assist for IT support (May 2025), AI sentiment analysis of employee surveys (2025), AI scheduling for Pilbara rail and ports (per BCG X), and Palantir Foundry workflows such as geotechnical risk monitoring at Oyu Tolgoi, plus machine learning for rail track maintenance (last described Jul 2024); the pilots are AI-assisted freight invoice processing and AI-driven mine closure tracking, and the announced deployment is Ideon's AI-assisted subsurface modelling. Rio Tinto has reported no business return it credits to AI for any of them since May 2025 and no firm-level return from AI; the figures it has given are accuracy rates and task speedups.

Use case	Area	Status
ReconAI fault classification for defect elimination ReconAI takes over 3 of the 6 steps of defect elimination on Rio Tinto's autonomous locomotives, classifying each fault to a causing component and failure mode from telemetry, SAP and free-text records and then measuring whether corrective actions worked, so engineers spend more time in the field. ReconAI reached 96% accuracy for locomotive faults in production, up from 80% in a proof of concept a year earlier, and Rio Tinto is extending it to ore cars, signalling and port conveyors; Rio Tinto tracks its cost per classification but has not published a saving.	Iron Ore rail	In production
Metpro domain assistant for a legacy manufacturing system Rio Tinto Aluminium Pacific Operations built an assistant on a fine-tuned Llama 3.1 8B model and Amazon Bedrock agents that documents the code, dependencies and decision logic of Metpro, a 30-year-old manufacturing execution system, instead of rewriting it. Data science senior advisor Ke Shi said in May 2026 that engineers can now see how a proposed change will spread in minutes rather than days, a task speedup without a cost equivalent.	Aluminium	In production
Copilot and Power Platform in month-end close Rio Tinto's global shared services moved journal calculations for the month-end close from hundreds of spreadsheets to Power BI and Power Automate fed by SAP, and use Copilot Studio and Copilot in Power BI to analyse and summarise data in natural language. Brandon Hogan said in May 2026 that the automation had cut month-end effort significantly and given Rio Tinto's largest product group its fastest close on record, without figures.	Group Services finance	In production
Future Scheduling Platform for Pilbara rail and ports BCG X says it built Rio Tinto's Future Scheduling Platform to replace dated scheduling systems for about 50 schedulers in Perth, using machine learning, optimisation and simulation, and claims the platform more than doubled scheduler productivity and paid back its cost in under 3 months. The claims are the vendor's, undated, and Rio Tinto has not published them.	Iron Ore supply chain	In production
AI-assisted freight invoice processing Rio Tinto handles more than 10,000 freight invoices a year, a manual task that means cross-checking several data sources; an earlier AI-assisted processing tool is the starting point for the 2-year memorandum Rio Tinto signed with AI Singapore in Mar 2026 to speed up validation, improve accuracy and cut turnaround times.	Commercial	Pilot
Geotechnical risk monitoring and operational workflows in Palantir Foundry Rio Tinto's Palantir Foundry ontology combines data from thousands of sensors at the Oyu Tolgoi block cave mine in Mongolia into one view of cave health, instrumentation and risk, and also supports plant operations and the coordination of driverless iron ore trains; Palantir said in Nov 2024 that AIP would extend these workflows with AI.	Copper and operations	In production
ServiceNow Now Assist for IT service management Rio Tinto went live with ServiceNow's Now Assist in May 2025 for incident summarisation, resolution notes and AI search, as it retired its legacy ticket management system.	Information technology	In production
AI sentiment analysis of employee surveys Rio Tinto gives leaders AI sentiment analysis of employee survey feedback, according to its 2025 Form 20-F; almost 40,000 employees took part in the Q4 2025 survey.	People	In production
AI-driven workflows for mine closure tracking Rio Tinto has been trialling AI-driven workflows and remotely operated platforms that combine site and remote sensing to track progress and predict trajectories toward rehabilitation and closure criteria on mine sites (2025 Form 20-F).	Closure and rehabilitation	Pilot
AI-assisted subsurface modelling with Ideon Ideon Technologies' REVEAL platform, which combines cosmic-ray muon imaging, geophysical sensing and AI services into 3D models of the subsurface, is to be deployed first at 6 of Rio Tinto's largest operations under a 5-year partnership announced in Oct 2025, after a pilot at the Bingham Canyon copper mine.	Exploration and mining	Announced
Machine learning for rail track maintenance A Rio Tinto machine learning model runs on a sliding window to predict about 7 weeks ahead where track on the Pilbara rail network needs maintenance, so planners can schedule around it; Rio Tinto described the model in Jul 2024, along with an hourly model in Canada that rates the risk of water leaking into a smelter.	Iron Ore rail	In production

Adoption

Rio Tinto has published no count of employees who use AI tools and has not described a company-wide generative AI assistant or rollout. The tools on the public record serve defined groups: maintenance and reliability engineers on the Pilbara rail network use ReconAI (Jun 2025), engineers who maintain the Metpro system use its assistant (May 2026), and finance shared services teams use Copilot at an increasing rate, according to senior manager Brandon Hogan (May 2026). Rio Tinto's IT service desk runs ServiceNow Now Assist (May 2025), and leaders receive AI sentiment analysis of employee surveys (2025). Rio Tinto's machine learning environments in Amazon SageMaker are open to citizen data users as well as data scientists (Jul 2024).

By the numbers

ReconAI accuracy classifying locomotive faults: 96%; In production on the Pilbara rail network, up from 80% in the proof of concept; Milan Pareek at AWS Summit Sydney (iTnews, Jun 10, 2025)

Return credited to AI since May 2025: None; No cost, headcount or revenue figure credited to AI in the 2025 Form 20-F, the Dec 2025 Capital Markets Day or the Jul 2026 half-year results

Palantir enterprise agreement extension: 4 years; Renewed in Nov 2024, adding access to Palantir's AI Platform (AIP) on top of Foundry

Operations in Ideon's first AI-assisted subsurface deployment: 6; 5-year enterprise partnership announced by Ideon Technologies, Oct 2025

Employees: ~61,000; 2025 average including Rio Tinto's share of joint ventures; 55,572 at managed operations on Dec 31, 2025 (2025 Form 20-F)

Rollout, first pilot to today

About 2017
Rio Tinto set up a central, dedicated machine learning capability to work across its product groups, according to its machine learning operations principal in 2024.
iTnews
Oct 2022
Rio Tinto began building an in-house software development function in Brisbane that could grow to more than 500 people over 5 years; CIO Daniel Evans cited machine learning and AI among Rio Tinto's technologies.
iTnews
Jul 2024
Rio Tinto described standardising machine learning delivery on Amazon SageMaker, with models for rail track maintenance, smelter safety and mine planning.
iTnews
Nov 2024
Palantir renewed its enterprise agreement with Rio Tinto for 4 more years and added access to Palantir's AI Platform (AIP).
Palantir release
May 2025
Rio Tinto went live with ServiceNow's Now Assist for IT service management.
iTnews
Jun 2025
Rio Tinto said ReconAI, built on Claude through Amazon Bedrock, was in production for locomotive fault classification with 96% accuracy.
iTnews
Oct 2025
Ideon Technologies announced a 5-year enterprise partnership to deploy its AI-assisted subsurface platform at 6 Rio Tinto operations.
Canadian Mining Journal
Dec 2025
At its Capital Markets Day, Rio Tinto's Aluminium product group listed targeted digital and AI initiatives among its levers for higher returns.
Capital Markets Day presentation
Mar 2026
Rio Tinto signed a 2-year memorandum with AI Singapore to develop AI solutions for its supply chains.
TechNode Global
May 2026
Rio Tinto staff described Copilot Studio in finance shared services and a fine-tuned Llama assistant for the Metpro aluminium system.
iTnews

Economics & controls

Spend and ROI

Technology and AI spending

Rio Tinto discloses no technology budget and no AI spending. Rio Tinto's 2025 Form 20-F includes continued investment in Group-wide technology and systems to drive productivity within $0.6 billion of restructuring, project and one-off costs, without a separate figure, and says Rio Tinto has become sharper with its investment in digital solutions. Rio Tinto has not disclosed the value of its Palantir agreement, its AWS agreement or its Ideon partnership.

Reported ROI

Rio Tinto has reported no business return it credits to AI since May 2025, and no firm-level figure for AI. Rio Tinto's productivity program, $650 million annualised by Dec 2025, a $1.3 billion run rate in the first half of 2026 and $870 million banked by Jun 2026, is credited to production uplifts, cost savings and cost efficiencies from a simpler organisation, operational discipline, labour productivity and contractor management; Rio Tinto does not credit any of it to AI. The figures Rio Tinto staff have given for AI tools are an accuracy rate (ReconAI, 96%) and a task speedup (Metpro, minutes rather than days). BCG X's claim that the Future Scheduling Platform more than doubled scheduler productivity and paid back in under 3 months comes from the vendor and is undated.

Headcount

Rio Tinto employed an average of about 61,000 people in 2025, including its share of joint ventures, and 55,572 at managed operations at the end of 2025. In the second half of 2025 Rio Tinto cut its Executive Committee from 11 to 9 members and senior management roles by 22% (2025 Form 20-F). Rio Tinto has not linked these changes, or any headcount change, to AI.

AI as a revenue market

Rio Tinto's AI-related revenue is in metals. Rio Tinto's 2025 Form 20-F names AI data centres among new drivers of metals demand, Amazon Web Services agreed in Jan 2026 to buy copper from Rio Tinto's Nuton process for components of its U.S. data centres, and copper, aluminium and lithium contributed more than half of Rio Tinto's underlying EBITDA in the first half of 2026. These are sales to AI infrastructure builders, not a return on Rio Tinto's own AI use.

Governance and risk

AI as a material risk

Rio Tinto's 2025 Form 20-F says AI brings risks including data privacy breaches, intellectual property challenges, misinformation, cyber security weaknesses and regulatory uncertainty, that generative AI vendors often put feature releases ahead of security, and that Rio Tinto responds with monitoring, internal upskilling and governance processes. Rio Tinto rates cyber risk as trending upward against 2024, driven partly by the adoption of automation and AI, and notes attackers using AI to bypass security controls.

Who oversees technology risk

Rio Tinto's Board oversees material risk, and the executive Risk Management Committee, chaired by CFO Peter Cunningham, oversees cyber security. The Cyber Security function, led by the CISO, who reports to the CIO, reports monthly to the CFO, quarterly to the Risk Management Committee and twice a year to the Board's Audit & Risk Committee, measured against the NIST Cybersecurity Framework 2.0 (2025 Form 20-F). Rio Tinto says it has had no significant cyber security incident since Dec 2014. Rio Tinto's filings describe no board committee or executive role dedicated to AI.

Controls in the tools

Rio Tinto's production machine learning accounts were air-gapped, with no internet access by default and outside access whitelisted through a central firewall (Jul 2024). ReconAI shows engineers the full reasoning of its agent workflow, including what it retrieved, which Milan Pareek said in Jun 2025 is how sceptical engineers came to trust it, and Rio Tinto monitors ReconAI's cost per classification. Palantir said in Nov 2024 that Rio Tinto's ambition was the secure use of AI.

Employee rules

Rio Tinto has not published an AI use policy for employees in the sources reviewed. Rio Tinto's updated mandatory Code of Conduct training was completed by 21,693 digitally connected and 17,182 digitally disconnected employees in 2025 (2025 Form 20-F), which does not say whether the Code covers AI.

Recent signals

Jul 29, 2026
Rio Tinto's half-year results reported $870 million of productivity benefits banked and a $1.3 billion annualised run rate, credited to production and cost improvements, with no mention of AI in Rio Tinto's own operations.
2026 half-year results
May 26, 2026
Rio Tinto data science senior advisor Ke Shi presented an assistant built on a fine-tuned Llama 3.1 8B model and Amazon Bedrock that documents Metpro, a 30-year-old aluminium manufacturing execution system.
iTnews
May 19, 2026
Rio Tinto finance shared services described automating month-end journals with Microsoft Power Platform and using Copilot Studio and Copilot in Power BI, with more AI agents expected.
iTnews
May 6, 2026
Simon Trott's address to Rio Tinto's annual general meetings covered safety, productivity and growth and did not mention AI.
AGM address, Form 6-K
Mar 2026
Rio Tinto signed a 2-year memorandum with AI Singapore, facilitated by Enterprise Singapore, to build AI tools for its supply chains, starting with freight invoice processing.
TechNode Global
Feb 2026
Rio Tinto's 2025 annual report on Form 20-F described AI as a material risk, named CIO Daniel Evans and CISO Scott Brown, and disclosed AI sentiment analysis of employee surveys and trials of AI-driven closure workflows.
2025 Form 20-F
Jan 16, 2026
Rio Tinto signed a 2-year agreement under which AWS buys Nuton copper for its U.S. data centres and Nuton uses AWS platforms for simulation and analytics.
iTnews
Dec 4, 2025
At its Capital Markets Day, Rio Tinto reported $650 million of annualised productivity benefits, including simpler digital applications and lower support costs, and its Aluminium product group listed targeted digital and AI initiatives.
Capital Markets Day presentation

Quick answers

How does Rio Tinto deploy AI?: Rio Tinto deploys AI through its IS&T technology function and data science teams in the product groups, building tools on Palantir Foundry and AIP and on AWS, where ReconAI calls Anthropic's Claude through Amazon Bedrock to classify locomotive faults and an aluminium team runs a fine-tuned Llama model. Finance staff use Microsoft Copilot Studio, and a Mar 2026 agreement with AI Singapore targets freight invoices. Rio Tinto has reported no AI user count and no return it credits to AI since May 2025.
What does Rio Tinto use AI for?: Rio Tinto has 8 AI use cases in production on the public record as of Nov 2026, plus 2 pilots and one announced deployment: ReconAI for rail fault classification (Jun 2025), the Metpro assistant in aluminium (May 2026), Copilot in finance shared services (May 2026), ServiceNow Now Assist for IT support (May 2025), AI sentiment analysis of employee surveys (2025), AI scheduling for Pilbara rail and ports (per BCG X), and Palantir Foundry workflows such as geotechnical risk monitoring at Oyu Tolgoi, plus machine learning for rail track maintenance (last described Jul 2024); the pilots are AI-assisted freight invoice processing and AI-driven mine closure tracking, and the announced deployment is Ideon's AI-assisted subsurface modelling. Rio Tinto has reported no business return it credits to AI for any of them since May 2025 and no firm-level return from AI; the figures it has given are accuracy rates and task speedups.
Who leads AI at Rio Tinto?: AI at Rio Tinto is led by Daniel Evans (Chief Information Officer); Scott Brown (Chief Information Security Officer). Ownership of AI on the org chart: No named AI executive; IS&T under CIO Daniel Evans, data science in the product groups.
What AI platform does Rio Tinto use?: Rio Tinto builds AI applications on Palantir Foundry, whose enterprise agreement Rio Tinto extended in Nov 2024 to include Palantir's AI Platform (AIP), and on AWS services such as Amazon Bedrock and SageMaker; ReconAI, Rio Tinto's rail fault classifier, reads data collected in Foundry and shows its results there (Jun 2025).
Which AI models does Rio Tinto use?: Rio Tinto's ReconAI calls Anthropic Claude models, mainly Sonnet, through Amazon Bedrock (Jun 2025); a Rio Tinto aluminium team fine-tuned Meta's Llama 3.1 8B on Amazon SageMaker (May 2026); and Rio Tinto finance teams use Microsoft Copilot Studio and Copilot in Power BI (May 2026).
How many Rio Tinto employees use AI?: Rio Tinto has not said how many of its roughly 61,000 employees use AI tools; a finance shared services manager said in May 2026 that teams use Copilot at an increasing rate, and ReconAI serves maintenance and reliability engineers on the Pilbara rail network.
What return has Rio Tinto reported from AI?: Rio Tinto has reported no business return it credits to AI since May 2025: its 2025 annual report on Form 20-F, its Dec 2025 Capital Markets Day and its Jul 2026 half-year results put no cost, headcount or revenue figure on its own use of AI, and the $870 million of productivity benefits banked in the first half of 2026 is credited to production uplifts, cost savings and cost efficiencies, not to AI. The nearest figures are a task speedup (engineers trace dependencies in a legacy aluminium system in minutes rather than days, May 2026), a classification accuracy of 96% for ReconAI (Jun 2025) and a BCG X case study claiming that Rio Tinto's Future Scheduling Platform more than doubled scheduler productivity, which is a vendor's claim. Rio Tinto has not reported a company-wide AI rollout either: it gives no count of employees using AI tools and names no executive who owns AI. Rio Tinto's AI work since May 2025 is on the public record, from ReconAI in production (Jun 2025) and a 5-year Ideon partnership (Oct 2025) to a memorandum with AI Singapore (Mar 2026), so Rio Tinto sits at Experimenting rather than Unknown. A cost, production or headcount return Rio Tinto credits to AI in one use case, or a stated count of employees using AI tools under a named owner, would move Rio Tinto to Piloting.

Open questions

Does Rio Tinto give employees a sanctioned generative AI assistant, which one, and how many of its roughly 61,000 employees use AI tools each week?
Who owns AI at Rio Tinto, and does Rio Tinto still have a Chief Digital Officer? An org-chart site lists Shawn Lyndon in that role since Nov 2023, but Rio Tinto's 2025 Form 20-F names no digital officer.
What has ReconAI saved in locomotive downtime, maintenance cost or engineer hours, and has it reached ore cars, signalling and port conveyors?
Does Rio Tinto confirm BCG X's claims that the Future Scheduling Platform more than doubled scheduler productivity and paid back in under 3 months, and when did the platform go live?
Which AI use cases has Rio Tinto built on Palantir's AIP since Nov 2024, which models run behind them, and what have they returned?
Does any part of Rio Tinto's productivity program, $870 million banked by Jun 2026, come from AI, and will Rio Tinto report an AI share?
What do the targeted digital and AI initiatives in Rio Tinto's Aluminium product group consist of, and what returns are they expected to deliver?
What does Rio Tinto spend on technology and on AI, including its Palantir, AWS and Ideon agreements?
Has Rio Tinto adopted an employee AI use policy, and how does it govern the use of Claude, Llama and Copilot on company data?
How many data scientists, machine learning engineers and software developers does Rio Tinto employ, and did the Oct 2022 plan for a 500-person development function go ahead?

Sources

Rio Tinto

Filings and investor materials

Partners

News and other

Data confidence

Confirmed: Stated in a primary or reputable public source.
Estimated: Derived or approximated from public reporting.
Unconfirmed: Reported, but not yet corroborated.

Rio Tinto profile · Enterprise AI diffusion report by Imagine · facts checked against their sources on Nov 5, 2026

More company profiles

All profiles