The AI Summit London is pleased to present an exclusive interview with Mostafizur Rahman, Chief Technologist for AI at The Manufacturing Technology Centre (MTC), part of the UK’s High Value Manufacturing Catapult. With more than 20 years of experience in Artificial Intelligence and Data Science, Mostafizur is recognised for incubating and designing next‑generation industrial AI applications using cutting‑edge AI technologies, and for leading the system‑level architecture and real‑world validation needed to deploy them safely at scale. He combines strategic leadership with hands‑on technical delivery, working directly with industry to shape and implement trusted AI systems that deliver measurable operational impact across manufacturing, infrastructure, healthcare, defence, and enterprise technology.

In this interview, he shares practical insights on the barriers to scaling AI in industrial environments, the realities of data readiness, and the governance approaches needed for regulated sectors. He also discusses the future of industrial AI applications and emphasises the role of both leadership and frontline operators in embedding AI into day‑to‑day operations.

Q1. Challenges & Barriers to Embedding and Scaling AI

Most industrial AI initiatives struggle not because the technology is immature, but because organisations approach AI tactically rather than strategically. Too many pilots begin without clearly identifying the highest-impact use cases, the value drivers, or the system level constraints that determine whether AI can scale.  Many organisations still lack a coherent AI roadmap that aligns business outcomes with data architecture, infrastructure strategy and operational readiness, meaning early successes remain isolated rather than forming part of a scalable industrial AI programme.

Scaling often fails when AI development outruns data quality, engineering maturity and infrastructure capability; without data contracts, stable pipelines and governance, every deployment becomes a bespoke solution. Lifecycle costs such as MLOps, model monitoring, retraining, cybersecurity controls and AI safety assurance are rarely defined early, creating hidden overheads later. And accountability is frequently fragmented, with IT, OT and digital teams owning different parts of the AI system. True scaling requires alignment across strategy, data, systems and roles, not just strong pilot performance. This is the foundation of successful largescale industrial AI adoption.

Q2. How do you help organisations prove an AI system is truly production‑ready for mission‑critical environments, and what does real data readiness look like in complex industrial settings?

To help organisations demonstrate that an AI system is genuinely production ready for mission-critical environments, we take an end-to-end approach that begins with fixing the upstream issues that typically derail industrial AI at scale. We work closely with customers to build a clear AI strategy and roadmap, jointly identify the highest-impact use cases, and analyse system level constraints, data gaps, maturity levels, ROI potential and business case viability. Our AI specialists collaborate with engineering and industrial domain experts to ensure every concept is grounded in real operating conditions. We design the complete system architecture, support technology down selection and build versus buy decisions, and develop early Proof of Concepts (PoCs) and technology demonstrations in real manufacturing environments to derisk adoption. For vendors and solution providers, we offer independent testing and validation on our R&D production lines to assess performance, safety, robustness and integration in realistic industrial conditions.

In complex industrial environments, real data readiness goes far beyond simple data availability. It requires stable and well governed pipelines, consistent semantics and data models across sites, clear ownership, and the confidence that the data accurately reflects real production behaviour. It also involves addressing gaps in coverage, quality, calibration, traceability and lineage so that AI systems can operate reliably over time.

Q3. Trust is essential in regulated sectors like healthcare, defence, and critical infrastructure. What governance frameworks do you recommend, and how do you balance innovation with safety and accountability?

Trust depends on strong governance, transparent decision making and confidence that AI systems are designed and deployed with accountability built in. We start with an industrial AI governance framework we had developed and tailor it to each customer, aligning with recognised standards such as ISO/IEC 42001, the NIST AI Risk Management Framework and sector specific safety and quality practices. A core element is creating clear ownership across IT, OT, data and AI teams so responsibilities for data quality, model behaviour, operational monitoring and incident response are well defined. We guide organisations through a full end-to-end AI lifecycle from scoping and risk assessment through architecture design, validation, deployment and decommissioning ensuring transparency and cross functional oversight at every step.

We support early experimentation in controlled environments while ensuring that assurance considerations are captured from the outset, with formal assurance gates applied before anything enters real operations. This enables organisations to innovate at speed while maintaining robust, auditable and compliant AI systems that stakeholders can trust. Crucially, trust can only be fully achieved when AI is validated at the system level, not just in isolation. We recommend independent third-party testing and system-level validation to ensure AI solutions perform reliably and safely within real operational contexts, meeting regulatory and stakeholder requirements.

Q4. How do you work with operators and leadership teams to build confidence and embed AI into daily operations?

We take a dual approach that supports both leadership teams and frontline operators. With leadership, we begin by clarifying the business case, expected impact and ROI, as well as the lifetime cost of running, maintaining and evolving AI systems. We help leaders understand the capabilities and operating model required to support AI long term, including data ownership, model monitoring, change control and the skills needed to manage a dynamic, continuously improving system.

With operators, our priority is building trust and demonstrating value to their jobs. We run hands-on technology demonstrations and pilot solutions directly in real workflows to show how AI can augment tasks rather than replace roles. We encourage operators to challenge assumptions, shape the solution and understand its benefits. Training is practical and transparent, openly addressing concerns around job displacement, which in practice is almost never the case. This approach has proven highly effective. Our rollout of Copilot tools followed the same model, and user engagement was extremely positive. By bringing together strategic clarity for leaders and user confidence, AI becomes embedded into daily operations rather than existing as a separate initiative.

Q5. Looking ahead five to ten years, what emerging industrial AI applications excite you most—and what barriers need to be overcome to make them a reality?

Over the next decade, industrial AI will be shaped by the convergence of foundation models, agentic AI and edge intelligence, moving well beyond today’s business layer applications into the core of shopfloor and physical operations. We will see agentic AI systems coordinating quality, scheduling, energy management and throughput across production environments; world-model driven factories that learn machine and process behaviour in real time; advanced multimodal inspection using vision, acoustics, NDT and process signals; and autonomous maintenance systems capable of diagnosing issues, recommending interventions and supporting engineers. As foundation models increasingly underpin engineering knowledge, documentation and troubleshooting, and as edge AI becomes robust enough to run safely on equipment and cells, we will shift from decision support to true operational augmentation.

The barriers to achieving this are significant: industrial data readiness at scale, validated assurance methods for adaptive and agentic systems, interoperability across legacy IT/OT environments, and clarity in human–system teaming and operational ownership. To date, AI adoption has largely transformed the business layer; over the next decade, it will integrate with physical layer where industrial productivity can be truly transformed.

About:

Mostafizur Rahman, Chief Technologist for AI at The Manufacturing Technology Centre (MTC), part of the UK’s High Value Manufacturing Catapult. With more than 20 years of experience in Artificial Intelligence and Data Science, Mostafizur is recognised for incubating and designing next‑generation industrial AI applications using cutting‑edge AI technologies, and for leading the system‑level architecture and real‑world validation needed to deploy them safely at scale. He combines strategic leadership with hands‑on technical delivery, working directly with industry to shape and implement trusted AI systems that deliver measurable operational impact across manufacturing, infrastructure, healthcare, defence, and enterprise technology.

Discover actionable strategies and forward-looking perspectives from one of the industry’s foremost AI leaders, and don’t miss Mostafizur’s session, Panel: Scaling | Moving Beyond Pilots – Why Industrial AI Stalls, taking place on Thursday, 11 June, on the Industrial AI stage at The AI Summit London.

Secure your pass today and join the conversation!