Dangote Refinery Adopts New Industrial AI Platform as Honeywell Projects Potential $60 Million Annual Operational Benefits

 

By Simpson Global Media News Desk

ANNOUNCEMENT DATE: Monday, 26 October 2026
IMPLEMENTATION LOCATION: Dangote Petroleum Refinery and Petrochemicals, Lekki, Lagos State, Nigeria.
TECHNOLOGY PARTNERS: Honeywell Technologies, NVIDIA and Dell Technologies.
PROJECTED FINANCIAL BENEFIT: More than US$60 million in annual gross operational benefits at full-capacity throughput, according to Honeywell's preliminary estimate.
REPORT PREPARATION DATE: Monday, 26 October 2026, West Africa Time.

AI enters Nigeria's large-scale industrial operations

Nigeria's Dangote Petroleum Refinery is set to deploy an industrial artificial intelligence platform developed by Honeywell Technologies, in a move intended to improve production efficiency, identify operational bottlenecks and strengthen the performance of one of the country's most important industrial investments.

Honeywell announced on Monday, 26 October 2026, that its Forge Plant Cognition platform would launch first at the refinery in Lekki, Lagos. The system combines Honeywell's industrial engineering and process-control expertise with AI technology and accelerated computing capabilities from NVIDIA and computing infrastructure from Dell Technologies.

According to Honeywell, preliminary estimates suggest the platform could deliver more than US$60 million in gross operational benefits annually when the refinery operates at full capacity. The company estimates a potential benefit exceeding US$0.29 per barrel processed, but cautions that the projection is not guaranteed and excludes implementation costs. (Honeywell, 26 October 2026; Reuters, 26 October 2026)

The announcement is significant because it represents an application of AI beyond the chatbots, image generators and office assistants that dominate public discussion of the technology. Industrial AI is intended to support complex physical operations where equipment performance, production conditions, energy consumption and engineering decisions can have substantial financial consequences.

At a refinery, small changes in operating conditions can influence the quantity of products obtained from processing crude oil, the reliability of equipment and the cost of maintaining production. Identifying those changes quickly may help operators make more informed decisions.

Honeywell says Forge Plant Cognition is designed to combine real-time operating data with engineering knowledge and physics-based models. Its intended functions include identifying production constraints, anticipating equipment problems and helping operators determine actions that could improve performance.

The deployment also has a workforce-development dimension. Dangote Refinery chief executive David Bird said the investment supports the company's commitment to operational excellence, technological development and the training of Nigerian engineers and operators. Honeywell said the platform is intended to make specialised process knowledge more accessible to local personnel as the refinery expands. (Honeywell; BusinessDay, 26 October 2026)

However, the projected financial gain remains a company estimate rather than a verified saving already achieved. The announcement does not establish how much the refinery has saved since deployment began, how quickly the full system will be operational or whether every projected benefit will be realised.

The important next stage will be determining whether the technology produces measurable improvements under actual operating conditions and whether those improvements translate into lasting operational and financial value.

What is Forge Plant Cognition?

Forge Plant Cognition is an industrial AI platform designed to help operators improve the performance of complex industrial facilities.

Unlike a conventional consumer AI assistant, which might answer questions or draft text, the platform is intended to work with industrial information and specialised engineering knowledge. It combines real-time operational data, process expertise, industrial automation and physics-based calculations to help teams understand how a facility is performing and what may be limiting its output.

At a refinery, this information can include operating conditions across connected processing units, equipment performance, process constraints and data associated with production quality and efficiency. The platform is designed to help engineers identify relationships across these systems rather than examining individual data points in isolation.

Honeywell says the technology is intended to serve three broad purposes: capturing otherwise hidden opportunities to improve production margins, protecting output by identifying emerging problems, and making specialised engineering knowledge more accessible through AI-supported workflows. (Honeywell)

The first purpose involves identifying constraints that prevent a facility from producing at its most efficient achievable level. If a particular unit limits the throughput of several connected processes, understanding that relationship can help engineers determine where an operational adjustment may deliver the greatest benefit.

The second purpose concerns reliability. Equipment problems that are identified early may be easier to address than failures discovered after they interrupt production. AI-supported analysis could help maintenance and operations teams recognise patterns that indicate a developing problem.

The third purpose involves knowledge transfer. Industrial facilities rely on experienced engineers who understand the relationships between equipment, operating conditions and process outcomes. Embedding appropriate engineering knowledge into digital tools may help other staff access relevant guidance when investigating problems.

The company has not published every technical parameter of the Dangote implementation, including the full deployment schedule, the detailed configuration of its systems or the precise production targets assigned to each component.

The platform should therefore be understood as an AI-assisted industrial optimisation system with stated performance objectives, rather than a fully autonomous refinery guaranteed to operate without human involvement.

How AI can help engineers identify production bottlenecks

A production bottleneck occurs when one part of an operating system limits the throughput or efficiency of the wider process.

Refineries consist of connected operations that separate crude oil into fractions and process those fractions into products with different characteristics and uses. The performance of one unit can influence the workload and efficiency of other units.

If a particular stage cannot process material at the required rate, the refinery may be unable to increase overall output even when other equipment has available capacity. In other circumstances, operating conditions may influence the quantities or quality of products obtained from a given volume of crude.

Finding the source of a bottleneck requires engineers to examine the interactions between process units, operating limits, equipment performance and other relevant variables.

Honeywell says Forge Plant Cognition is designed to combine real-time plant information with process knowledge and advanced calculations to help operators identify constraints and evaluate potential actions. (Honeywell)

An AI-supported system may be able to examine large quantities of operational information, identify patterns and flag conditions that deserve closer investigation. Engineers can then review those findings against established process knowledge and safety requirements.

For example, a system might identify that a process unit is approaching a constraint that could affect several connected operations. It could help operators investigate the cause and compare possible responses before deciding whether an adjustment is appropriate.

The value lies in supporting an earlier or more complete understanding of the problem. AI cannot guarantee that every constraint will be identified, and its recommendations must be assessed against the actual conditions of the facility.

The refinery's configuration, the quality of its operating data and the reliability of the underlying engineering models will influence how useful the analysis becomes.

If the system helps identify genuine constraints and engineers can safely address them, the refinery may be able to process crude more efficiently or reduce avoidable production losses. Whether those improvements occur, and how large they are, will require measurement after implementation.

Protecting production by anticipating equipment problems

Unplanned equipment downtime can interrupt industrial production and create additional maintenance costs.

When a critical unit fails unexpectedly, the resulting disruption may affect other parts of the facility. Maintenance teams may need to arrange replacement parts, undertake urgent repairs and restore operating conditions before normal production can resume.

The duration and cost of an interruption depend on the equipment involved, the cause of the failure, the availability of spares and the complexity of the repair.

Predictive maintenance uses information about equipment condition and operating behaviour to identify signs that a problem may be developing. It can help maintenance teams prioritise inspections or plan interventions before a failure occurs.

Forge Plant Cognition is designed to assist with the detection of emerging equipment and process issues, recommend corrective actions and help reduce unplanned downtime, according to Honeywell. (Honeywell)

In practice, this type of capability depends on reliable information. Equipment sensors, operating records, maintenance history and other relevant data must be sufficiently accurate for the system to distinguish meaningful warning signs from routine variation.

False alarms can also create costs. If a system repeatedly identifies ordinary operating conditions as potential failures, engineers may spend time investigating problems that do not require intervention.

Conversely, a system that fails to flag a genuine warning may provide limited protection against unexpected downtime.

This means that the usefulness of predictive maintenance must be assessed against actual operating outcomes. Relevant indicators could include unplanned shutdowns, maintenance response times, equipment availability and the cost of repairs.

The platform's announcement establishes its intended role, but it does not provide a verified before-and-after record of downtime at Dangote Refinery.

That evidence will be important for assessing whether the platform improves reliability under real operating conditions.

The technology behind the system

The platform combines technologies and services from several companies.

Honeywell provides the industrial process expertise, automation knowledge and Forge platform. NVIDIA contributes AI models and accelerated-computing technologies, while Dell Technologies provides AI infrastructure and related services.

The stated technical architecture includes several components with different functions.

NVIDIA Nemotron open models are used to power AI agents that interpret operational information. NVIDIA Omniverse libraries are intended to help create a digital twin of the refinery and its processes, supporting simulation-based training and planning.

Honeywell also identifies NVIDIA Cosmos open models and the NVIDIA Metropolis Video Search and Summarisation Blueprint as components enabling vision AI agents for real-time alerts and anomaly detection. NVIDIA PhysicsNeMo is intended to support the development, training and refinement of physics-based AI models used in simulation and plant optimisation workflows. (Honeywell, 26 October 2026)

These systems serve different purposes. Language models can help interpret and organise information, while physics-based models are designed to represent relationships governed by physical principles. Digital twins provide virtual representations of equipment or processes that can be used to explore scenarios.

A digital twin does not automatically guarantee an exact representation of every condition in a live refinery. Its usefulness depends on the accuracy of its models, the quality of the information feeding them and how closely its assumptions match the actual plant.

Similarly, an AI model's ability to identify patterns does not establish that every recommendation is safe or optimal. Operators must assess the findings against the refinery's engineering constraints and approved operating procedures.

The combination of AI, physical models and industrial expertise is intended to improve the relationship between data analysis and operational decisions.

For Dangote, the practical test will be whether this combination helps engineers solve problems more effectively and improve measurable aspects of production.

Why the projected $60 million benefit needs careful interpretation

Honeywell estimates that Forge Plant Cognition could generate more than $60 million in gross operational benefits per year at Dangote Refinery when the facility operates at full capacity.

The company's preliminary estimate exceeds $0.29 per barrel processed. Honeywell says the annual calculation is specific to the Lekki refinery and is based on the expected effect of AI-enabled optimisation. (Honeywell; Reuters)

The distinction between gross operational benefits and net financial gains is essential.

Gross operational benefits can describe the value of improvements before accounting for all the costs associated with achieving them. Honeywell explicitly states that the estimate excludes implementation costs.

The eventual financial outcome would depend on how often the refinery can use the platform effectively, the operating conditions, the quality of the data, the product mix, the facility's utilisation and the margin associated with the products produced.

The estimate is also tied to full-capacity throughput. If a refinery processes less crude than the level assumed in the calculation, the annual benefit may differ.

An improved production process does not automatically produce the same financial outcome under every market condition. Product prices, crude costs, energy expenses, maintenance requirements and other operating factors influence the amount of value retained by the business.

The projected $60 million should therefore not be reported as a saving already delivered, a guaranteed return or a confirmed reduction in fuel prices for Nigerian consumers.

It is a preliminary company estimate of possible gross benefits under stated conditions.

To verify the eventual result, Dangote and Honeywell would need to compare actual performance with an appropriate baseline, account for implementation and operating costs, and distinguish changes associated with the AI platform from other influences on production.

What this could mean for Dangote Refinery's expansion

The AI initiative comes as Dangote Refinery pursues a major increase in its planned processing capacity.

Honeywell said in its announcement that it is already supporting the refinery's expansion from a nameplate capacity of 650,000 barrels per day to 1.4 million barrels per day within three years. The company has described the work as part of a wider relationship with Dangote spanning nearly a decade. (Honeywell, 26 October 2026; Reuters)

Nameplate capacity is the facility's stated design capacity. Actual throughput can vary according to operating conditions, maintenance requirements, feedstock availability and other factors.

Increasing capacity is therefore not simply a question of installing equipment. A refinery must coordinate multiple processing units and related infrastructure while maintaining safe operating conditions and product quality.

As the facility expands, its engineering and operations teams may need to assess a greater volume of information and manage additional interactions among equipment and production processes.

An industrial AI platform could help with that work by identifying constraints, supporting troubleshooting and making relevant technical knowledge easier to access.

However, the relationship between the AI launch and the capacity-expansion programme should not be overstated. Honeywell has said that it is supporting the wider expansion and that the platform is intended to improve operational performance. The announcement does not establish that AI alone will deliver the entire capacity increase.

The expansion depends on engineering, equipment, project execution and operational readiness, alongside any benefits produced by digital technology.

The AI deployment is therefore one component of the refinery's broader effort to improve performance and expand production capability.

Could industrial AI affect fuel availability and prices?

The announcement has implications for discussions about Nigeria's domestic refining capacity, but its direct effect on fuel prices remains uncertain.

A refinery that operates more efficiently may be able to reduce some avoidable costs, improve product yields or limit disruptions to production. Such improvements can strengthen the economics of the facility.

However, retail fuel prices are influenced by multiple factors beyond the efficiency of a single refinery. These can include crude-oil prices, the cost of transporting and processing crude, exchange-rate movements, taxes and charges, distribution expenses, competition and prevailing market conditions.

Even if the AI platform delivers the full projected gross operational benefit, the announcement does not establish how much of that value would be reflected in the prices paid by consumers.

The benefits could affect the refinery's operating margins, its ability to invest in production, the reliability of supply or other business decisions. Their distribution depends on commercial arrangements and wider market conditions.

It is therefore premature to conclude that the platform will cause petrol or diesel prices to fall.

The more defensible expectation is that a successful optimisation system could help the refinery manage certain operating constraints and improve its internal performance.

Whether those gains translate into broader consumer benefits will depend on additional factors and should be assessed through evidence as the deployment progresses.

Developing Nigerian engineers and technical expertise

Dangote has said the deployment is intended to support the development of Nigerian engineering and operations personnel.

David Bird, chief executive and managing director of Dangote Petroleum Refinery and Petrochemicals FZE, said the company selected the platform to improve operational performance while advancing technology and developing Nigerian talent. (Honeywell)

The announcement links this ambition to Dangote's goal of building a workforce that is 70 per cent Nigerian as the company expands. Honeywell says the AI platform is intended to make specialised process knowledge more accessible to engineers and operators and to reduce reliance on external expertise over time.

This is an important part of the deployment because industrial technology requires skilled people who can understand the systems, interpret their outputs and make appropriate decisions.

Refinery operations depend on expertise across process engineering, instrumentation, control systems, maintenance, safety and other technical disciplines.

AI tools may help employees retrieve relevant information, understand operating conditions and investigate problems, but workers still need the knowledge required to evaluate recommendations.

Training must therefore cover both the technical platform and the industrial processes in which it operates.

Staff need to understand how data is collected, what the system's outputs mean, when recommendations require further investigation and which decisions require formal engineering approval.

The workforce objective should also be assessed through practical results. Relevant measures could include training completed, the number of Nigerian engineers able to use the system independently, the transfer of technical responsibilities and the development of local expertise.

The announcement establishes an intention to build local capacity, but it does not yet provide verified figures showing how many workers have been trained through this specific deployment.

Why local skills matter to Nigeria's technology economy

Large industrial projects create demand for a range of technical skills, including software integration, data analysis, process engineering and the maintenance of specialised equipment.

Developing local capability in these areas can help Nigerian companies operate complex facilities while creating opportunities for technical careers.

The potential benefit extends beyond the refinery itself. Employees who gain experience with industrial AI, digital twins, process optimisation and data-driven maintenance may be able to apply related skills in other industrial settings.

However, those outcomes depend on the nature of the training, the opportunities available to employees and the extent to which knowledge is retained locally.

Buying sophisticated technology does not automatically create domestic technical expertise. Organisations must provide training, allow staff to work directly with the systems and establish processes through which they can assume responsibility for their operation.

The approach also matters for long-term resilience. If critical knowledge remains concentrated among a small number of external specialists, the company may continue to depend heavily on outside support for certain tasks.

Embedding appropriate process knowledge into digital workflows may help widen access to expertise, but AI cannot replace every aspect of professional judgement or practical experience.

For Nigeria, the broader objective is to combine advanced technology with sustained skill development so that local personnel can operate, maintain and improve complex industrial systems.

Industrial AI is different from ordinary chatbots

The Dangote deployment illustrates why AI should not be treated as a single category of technology.

Consumer-facing systems often generate text, answer questions, create images or assist with administrative work. Industrial AI is applied to physical systems where operational constraints, equipment behaviour and process safety must be taken into account.

At a refinery, an AI system cannot be evaluated only by whether its answer sounds plausible. Its output must be relevant to the engineering problem, consistent with the physical conditions and appropriate for the operating environment.

The consequences of an incorrect recommendation can be different from those of an incorrect response in a general-purpose chatbot.

A poor draft email may require editing. A flawed recommendation involving industrial equipment could potentially lead to unnecessary downtime, ineffective adjustments or safety risks if not properly assessed.

This is why engineering knowledge, verified data and established operating procedures matter.

Honeywell says Forge Plant Cognition combines AI with trusted calculations and physics-based models. The stated design is intended to bridge data analysis and engineering decision-making, rather than relying solely on text-based responses. (Honeywell)

The system's actual performance will still need to be evaluated in use. Combining AI with engineering models may support more useful decisions, but it does not guarantee that every output will be correct or that every identified opportunity can be implemented.

The technology should therefore be understood as a specialised decision-support and optimisation capability whose value depends on the quality of its design, data, validation and operational integration.

The role of NVIDIA and accelerated computing

NVIDIA's contribution is focused on AI models and computing technologies that support the platform's analytical and simulation capabilities.

Industrial AI tasks can involve large volumes of data, complex model calculations and repeated simulations. Accelerated computing can help execute certain types of workloads more efficiently than systems designed for more general-purpose tasks, depending on the application and configuration.

The technology identified by Honeywell includes NVIDIA Nemotron models for interpreting operational data, Omniverse libraries for digital-twin-related simulation and planning, Cosmos models with vision AI capabilities, and PhysicsNeMo for physics-based AI model development. (Honeywell)

These components do not all perform the same role. They form part of a wider technical architecture intended to connect information from industrial operations with models and tools that can support analysis.

The quality of the overall system will depend on how those components are integrated and whether they provide information that engineers can use in the context of the refinery's operating requirements.

The availability of advanced computing does not automatically improve industrial performance. The system must be configured to address specific problems, receive relevant data and operate within the wider technical environment.

The deployment reflects a growing interest in AI applications that involve physical assets rather than only digital content.

For Nigeria, its significance will depend on the outcomes achieved at the refinery and the degree to which the experience can inform future industrial technology projects.

Dell Technologies' role in the infrastructure

Dell Technologies is supplying the AI infrastructure and related support associated with the platform.

Industrial AI requires computing resources capable of processing operational data and supporting the relevant models and applications. Depending on the system design, the infrastructure may include computing equipment, data storage, networking and associated management services.

These components have to operate reliably because industrial users need consistent access to information during the tasks for which the system is intended.

Data storage and management are also important because operating records may be needed for analysis, troubleshooting, model evaluation and the reconstruction of events.

The infrastructure must be suitable for the workload and integrated with the relevant systems under appropriate security and access controls.

Honeywell identified Dell AI Factory with NVIDIA and Dell Services as part of the platform's architecture. The companies' roles are complementary: Honeywell brings industrial process expertise, NVIDIA contributes AI and accelerated-computing technologies, and Dell supplies computing infrastructure and services. (Honeywell)

The practical outcome will depend on whether the combined system meets the requirements of the refinery and provides sufficient reliability for its intended applications.

The public announcement does not disclose the detailed hardware configuration, total computing capacity, procurement value or full operating costs for the installation.

Those details would help analysts understand the scale of the infrastructure investment, but they are not necessary to establish the principal point of the announcement: the refinery is adopting an industrial AI platform developed through a partnership among several technology providers.

Data quality and system integration will be critical

AI systems rely on the information they receive, and industrial data can be difficult to interpret without sufficient context.

Refineries contain complex processes in which readings from sensors and control systems must be understood alongside equipment specifications, engineering calculations and other operating information.

Data may be incomplete, delayed, inconsistent or affected by changes in operating conditions. Some problems may be difficult to identify from one measurement alone.

If an AI system relies on poor-quality or inappropriate data, its findings may be misleading or insufficiently useful.

System integration is therefore one of the central challenges in industrial AI deployment.

The platform must be able to obtain the information required for its intended tasks, interpret it in the context of the refinery's processes and present the results in a form that staff can understand and review.

This may require attention to data formats, equipment interfaces, calibration, operating history and the quality-control processes used to maintain records.

Honeywell's stated approach combines real-time operating information with engineering knowledge and physics-based models, but the public announcement does not describe all the data sources or validation procedures used in the Dangote implementation. (Honeywell)

The system's performance will therefore need to be evaluated through testing and operational monitoring.

If the data is reliable and the models reflect the relevant processes, the system may be able to identify useful patterns and support better decisions. If important inputs are missing or the model is poorly matched to the operating conditions, the results may be less reliable.

The quality of the implementation will be as important as the capabilities of the underlying AI technology.

Safety and human oversight

Industrial AI must operate within a broader framework of engineering safety and accountability.

A refinery is a complex facility where operational decisions can affect equipment, workers, product quality and the surrounding environment. Changes to process conditions must comply with the facility's approved operating procedures and safety requirements.

AI may help identify a potential issue or suggest a corrective action, but the decision to implement that action needs to reflect the responsibilities assigned to qualified personnel and the controls governing the system.

The public announcement describes Forge Plant Cognition as a platform that identifies constraints, anticipates problems and supports actions to improve performance. It does not establish that the Dangote installation will autonomously control every refinery process or independently make every operational decision.

The distinction matters because decision-support capabilities and autonomous control are not identical.

Decision support can help an engineer understand a situation and compare available options. Automated control involves a different level of authority and requires separate design, validation and safety arrangements.

The platform should not be assumed to have authority beyond the functions documented for the implementation.

Operators and engineers will continue to need appropriate training, clearly assigned responsibilities and procedures for reviewing AI-generated recommendations.

Monitoring should also account for unexpected outputs, changes in operating conditions and cases where the system's advice conflicts with established engineering knowledge.

The objective is to use AI to improve decision-making while retaining the safety controls and accountability required for a high-risk industrial environment.

Cybersecurity and the protection of operational data

Connecting industrial information systems to advanced analytical platforms can create cybersecurity considerations.

Industrial facilities may have operational technology networks that control or monitor equipment alongside corporate IT systems used for communication, planning and administration.

The relationship between these environments needs careful management because unauthorised access or disruption could affect the availability, integrity or confidentiality of important systems.

An AI deployment may involve data transfers, user accounts, computing resources and interfaces between software components. Each connection needs appropriate security controls.

The relevant measures can include access restrictions, authentication, network segmentation, monitoring, software maintenance and procedures for responding to security incidents.

The exact requirements depend on the system's design, the sensitivity of the information and the environment in which it operates.

The Honeywell announcement describes the platform's technical capabilities and intended industrial benefits but does not publish the full cybersecurity architecture for the Dangote deployment. It would therefore be inappropriate to claim that all security risks have been eliminated or that a particular security incident has occurred.

The important point is that cybersecurity must be considered part of the deployment, not an issue to address only after the technology is running.

Reliable access to the AI system is useful only if the underlying infrastructure and operational data are adequately protected.

How success should be measured

The projected financial benefits provide one possible measure of success, but a comprehensive evaluation would need to consider several operational indicators.

The first is throughput: whether the system helps the refinery process more crude within its safe and practical operating limits.

The second is product yield and value. Improving the mix or quantity of usable products obtained from a given amount of crude may affect the economics of the operation, depending on market conditions and processing costs.

The third is reliability. The refinery can assess whether the technology contributes to fewer unplanned interruptions, better maintenance decisions or shorter troubleshooting periods.

The fourth is the quality and speed of engineering decisions. This may include how quickly teams identify constraints, investigate unusual conditions and implement appropriate corrective action.

The fifth is workforce development. The company can assess whether local engineers and operators gain practical expertise and whether the platform helps them perform tasks more effectively.

The sixth is total cost. The evaluation must include the cost of implementation, computing infrastructure, support, maintenance, training and other expenses needed to operate the system.

These measures should be compared with an appropriate baseline to establish what has changed since the technology was introduced.

The evaluation should also consider other factors that affect refinery performance, including crude quality, utilisation, maintenance activity, changes in product demand and the progress of the capacity-expansion programme.

A higher throughput figure alone would not prove that the AI platform caused the entire improvement. Similarly, an increase in operating costs would not necessarily establish that the system failed if other changes affected the facility.

A credible assessment needs to distinguish the platform's contribution from other influences.

Implications for Nigeria's wider industrial sector

The deployment provides an example of how AI could be used in other industrial environments where complex equipment, process data and specialised engineering knowledge interact.

Potential applications include power generation, petrochemicals, manufacturing, mining, logistics and other operations that depend on the performance of physical assets.

AI-assisted monitoring may help identify equipment problems, while simulation tools can support planning and the evaluation of possible operating scenarios.

However, the suitability of a platform for one facility does not establish that it will deliver the same benefits in every industry.

Different facilities have different operating conditions, data systems, equipment, safety requirements and cost structures. The technology must be adapted and evaluated for the specific problem being addressed.

The broader lesson for Nigerian businesses is that AI adoption should begin with a defined operational need rather than a general desire to use new technology.

A company should identify the problem it wants to solve, determine what data and expertise are available, assess the costs and risks, and establish how the results will be measured.

For some companies, the appropriate application may be relatively simple data analysis. For others, it may involve predictive maintenance, simulation or more specialised engineering models.

The objective is to identify the technology that addresses the problem effectively, rather than assume that the most advanced platform is necessarily the most suitable.

The Dangote project will provide a potentially important case for assessing the practical value of industrial AI in a major Nigerian facility.

What happens next?

The next stage is the implementation and operational evaluation of Forge Plant Cognition at Dangote Refinery.

Honeywell has announced the launch and described its intended capabilities, but the public material reviewed does not establish a complete deployment timetable, the full implementation cost or the actual savings achieved since the project began.

The companies will need to integrate the platform with the relevant operating information and ensure that engineers and operators understand how to use it.

They will also need to evaluate the quality of the system's outputs, establish appropriate oversight procedures and monitor whether the technology improves the refinery's performance under real operating conditions.

Further reporting should examine whether the refinery publishes measured results, how the platform affects downtime and throughput, and what progress is made towards developing the capabilities of Nigerian engineers and operators.

The capacity-expansion programme will also remain relevant. Honeywell has said it is helping the refinery work towards increasing nameplate capacity from 650,000 to 1.4 million barrels per day within three years.

The AI platform is intended to support operational improvements, but the expansion itself will depend on the engineering, infrastructure and execution requirements of the wider project. (Honeywell; Reuters)

The most important milestone will be the publication of evidence showing how the technology performs in practice, rather than further descriptions of what it is intended to achieve.

Conclusion: industrial AI moves from promise to a real-world Nigerian test

The launch of Honeywell's Forge Plant Cognition at Dangote Petroleum Refinery marks a significant new application of artificial intelligence in Nigeria's industrial sector.

The platform combines industrial engineering expertise, AI models and computing infrastructure from Honeywell Technologies, NVIDIA and Dell Technologies. Its stated purpose is to help engineers identify production bottlenecks, anticipate equipment problems, make better operational decisions and improve refinery performance.

Honeywell estimates that the platform could deliver more than US$60 million in gross operational benefits annually at full-capacity throughput, based on preliminary calculations exceeding US$0.29 per barrel. The company has cautioned that the figure is not guaranteed, excludes implementation costs and depends on actual operating conditions.

The deployment also connects with Dangote Refinery's planned capacity expansion and its stated ambition to build a workforce comprising 70 per cent Nigerians as it grows. The system is intended to make specialised process knowledge more accessible to local engineers and operators.

Yet the announcement is the beginning of a performance test, not proof that the projected savings have already been achieved. The actual outcome will depend on data quality, system integration, operational conditions, staff training, safety procedures and the cost of implementation.

For Nigeria's wider technology economy, the project demonstrates that AI can be applied to complex industrial processes rather than only to consumer applications. Its longer-term significance will depend on whether the technology produces measurable benefits and whether the associated expertise becomes embedded in the local workforce.

The central question now is not merely whether the refinery can deploy AI, but whether it can use the technology safely and effectively to improve production, reliability and commercial performance.

Simpson Global Media will continue to follow the Dangote refinery's AI deployment, the verified results of industrial automation and developments in Nigeria's technology and digital economy.