Hertfordshire Community Conference

Artificial Intelligence has become a strategic priority for organisations across every industry, yet many AI initiatives continue to struggle in delivering meaningful business outcomes. While businesses are generating more intelligence than ever through predictive models, Generative AI, and advanced analytics, the expected improvements in operational performance, efficiency, and commercial value often remain elusive.

This session explores the critical gap between AI-generated insight and real-world execution. Rather than focusing solely on models and technology, it examines why successful AI adoption depends on strong foundations in data, enterprise architecture, governance, and operating models. Through a practical business case, the session demonstrates where AI initiatives typically break down, why intelligence alone is not enough, and how fragmented workflows, disconnected systems, and unclear decision ownership prevent organisations from realising value.

Attendees will gain a practical framework for designing AI systems that do more than generate recommendations—they enable timely decisions, integrate seamlessly into business processes, and drive measurable outcomes. The session concludes with actionable guidance on how organisations can move beyond AI experimentation to build scalable, trusted, and business-focused AI capabilities that deliver sustainable impact.

Building an AI agent is only the first step. A real agent needs a strong harness around it — the system that controls how the agent thinks, uses tools, manages memory, handles errors, tracks cost, and responds safely.

Without a good harness, even a powerful model can become expensive, slow, unreliable, or difficult to debug. This is why agent optimisation is so important.

In this session, we will explore how to design agents that are not just smart, but also practical for real-world use. We will look at how to reduce token usage, choose the right model for the right task, improve tool calling, manage context, add evaluation checks, and monitor cost before deploying agents into production.

The goal is to help developers move from “I built an agent” to “I built an agent that is reliable, efficient, and ready to use.”

This talk explores the reality of the digital divide in Africa—moving beyond the common narrative of limited internet access to uncover the deeper challenge: the gap in digital skills, opportunity, and inclusion. While connectivity is improving across the continent, millions of young people still lack the knowledge and pathways needed to fully participate in the digital economy.

Drawing from real-world experience in technology, community leadership, and youth engagement, Sobere Opusunju shares practical insights into how we can shift from simply providing access to creating meaningful opportunities. The session highlights the critical role of education, mentorship, and locally relevant digital skills in empowering the next generation.

Through examples and lived experiences, this talk will challenge assumptions, spark new thinking, and offer actionable strategies for individuals, organizations, and communities looking to make a tangible impact.

Attendees will leave with a clearer understanding of what it truly means to bridge the digital divide—and how they can contribute to building a more inclusive and opportunity-driven digital future for Africa.

AI agents can now read documents, match records, draft invoices and write meeting minutes, yet most organisations cannot say what their agent is allowed to see, what it may decide, or how a mistake would be caught. Drawing on earlier data quality experience in utilities and professional services and on systems built in the INNOVATION FORUM Foundry for Innovation Forum, this talk sets out the engineering patterns that make agentic work safe to adopt: sealing sensitive data away from the model, extraction contracts with provenance, reconciliation against a single source of truth, and human authorisation before anything leaves the building. It then asks who gets left out. Open source and small language models running on modest hardware put these patterns within reach of charities, councils and small firms that cannot afford enterprise platforms, and low-code validation lets non-programmers check the results. Smaller models and processing each item only once also cut the energy and carbon cost of every task. Finally, it treats bias as an engineering control: statistical bias in the data is measured, behavioural bias in the people using it is governed, and every decision can still be traced to a person.

Most people use a mainframe every day without knowing it – when you tap your card, book a flight, or take cash out of an ATM, a mainframe is often doing the work behind the scenes. This talk explains what a mainframe actually is, in plain language, and why people are wrong to think of it as old or dying. We'll look at how it's built to never go down, how it can run software from decades ago without rewriting it, and why banks and airlines still trust it with their most important systems. We'll also talk about why there's a real shortage of people who know how to work with mainframes, which is good news if you're early in your career. Right at the end, we'll touch on one surprising thing: even though mainframes are everywhere, almost nobody can clearly explain what they cost to run, and why that's starting to change.

Key takeaways: you'll understand what a mainframe really is, and why "legacy" is the wrong word for it; see why mainframes are still critical today, and why there's a growing need for new people to work on them; and learn why the cost of running a mainframe is surprisingly hard to see, and why that matters.

Why now: most people, even in tech, don't realise how much of everyday life still depends on mainframes, and a real gap is opening up as experienced staff retire. At the same time, organisations are starting to pay more attention to what these systems cost to run.

This talk explores the development and real-world application of Artificial Intelligence–driven robotic and humanoid automation in laboratory research, healthcare environments, diagnostic systems, and industrial manufacturing. The session will discuss how intelligent automation platforms are transforming modern laboratories, hospitals, and production facilities by improving precision, safety, efficiency, and scalability.

The presentation will also cover the use of AI-controlled robotic systems in agriculture, warehouse and supply-chain operations, food production, and retail automation, demonstrating how autonomous and semi-autonomous technologies are reshaping multiple sectors. Practical examples of laboratory automation, robotic assistants, smart monitoring systems, and intelligent control platforms will be presented, together with discussion on future humanoid automation and fully autonomous research environments.

Organisations generate enormous quantities of information, yet the evidence needed to understand decisions is often fragmented across emails, meetings, systems and records. Drawing on Keith Brown's MBA research across 154 UK charities and Community Interest Companies, this talk explores how AI may reduce the effort required to find, connect, compare and challenge organisational evidence. It distinguishes model, system and governance risk while considering provenance, auditability, behavioural bias, privacy and human accountability across RPA, IPA and agentic systems, ultimately asking whether cheaper scrutiny will make governance more demonstrable or create stronger ways of controlling the organisational record.

Security Operations Centers are facing growing volumes of alerts, increasingly sophisticated attacks and a shortage of skilled security professionals. Traditional SOCs still rely heavily on analysts to investigate alerts, correlate events and determine the appropriate response across multiple security tools.

This session explores the evolution towards the AI Autonomous Cyber SOC — where AI can continuously analyse security telemetry, correlate events, identify anomalies and attack patterns, support investigations, prioritise threats and accelerate response.

The session will examine how AI can move the SOC beyond simple automation towards an intelligent operating model in which AI assists with detection, investigation, threat hunting and response, while humans retain appropriate oversight and accountability.

It will also explore the practical challenges of autonomous security operations, including explainability, AI accuracy, access controls, governance, human-in-the-loop decision-making and determining which actions can safely be automated.

Attendees will be introduced to a practical maturity model showing the journey from traditional SOC → automated SOC → AI-assisted SOC → AI-driven SOC → governed autonomous SOC, providing a framework for organizations considering how to introduce AI into their security operations.

University research can create value long before it generates revenue, yet investors often lack ways to assess that value or identify who benefits. Dr. Proselkov will present a proposed world model for tracing how research generates practical, societal, and financial impact via the stakeholder network, following a normative-empirical approach to determining objectivity through stakeholder networks. The model connects universities, investors, practitioners, and beneficiaries, linking evidence of progress to each stakeholder’s goals and capacity to act. It explores how the same research can produce different outcomes depending on who provides funding, access, expertise, or routes to adoption. The aim is to support earlier investment and more effective coordination, with impact assessed through evidence gathered as projects develop.

Modern organisations generate and use data spread across multiple applications, databases, teams and data products. While organisations can move data between these systems, business concepts may be defined and represented differently. A customer, asset, product or transaction, for example, may have different definitions across systems or business functions. This can create ambiguity, duplication and considerable effort in reconciling data before it can be trusted and reused.

Ontologies and knowledge graphs provide a way to create a common enterprise language without replacing existing systems or forcing diverse enterprise data into a single, monolithic data model. Semantic mappings connect local data structures to shared concepts, allowing information from different systems to be integrated and kept semantically aligned. Controlled vocabularies using the Simple Knowledge Organization System (SKOS) can align different terms, synonyms and related concepts. Ontologies make business concepts, relationships and rules explicit, while constraints define what valid and consistent data should look like.

Building a common language starts with the questions an organisation needs its data to answer. These can be translated into competency questions that guide ontology design, semantic mappings, knowledge graph construction, constraint validation and inferential reasoning. The resulting knowledge can then be made available to applications, analytics and AI. This enables people, systems and AI applications to interpret and use enterprise information more consistently across organisational boundaries.

The impact can be measured through fewer conflicting or duplicated definitions, reduced manual data reconciliation and greater reuse of agreed concepts. Other measures include improved cross-system query capability, faster integration of new data sources and improved data quality. Greater traceability can also support more explainable enterprise decisions. A common enterprise language therefore goes beyond data standardisation. It provides a foundation for interoperability, governance, analytics, AI and decision intelligence, turning fragmented enterprise data into connected, contextualised and reusable knowledge.

Product teams can deliver a solution on time, within budget and against an agreed specification, yet still fail to create meaningful value. This happens when teams become more confident about the solution than they are curious about the user. A user mindset provides a practical way to correct that imbalance. It asks product professionals to understand the task people are trying to complete, the context in which they act, the constraints they face and the outcome that matters to them.

This session explores how a user mindset changes product delivery from discovery through to benefit realisation. It shows how user evidence can expose costly assumptions, improve prioritisation, reduce unnecessary scope and create stronger measures of success. The talk also distinguishes what users say from what they do and what changes as a result, explaining why interviews, observation and product data work best when used together.

Attendees will leave with a simple LENS framework: Listen to the user’s task and context, gather Evidence, Narrow the solution to the smallest useful change and See what happens after release. The framework can be used in backlog refinement, sprint planning, reviews and roadmap discussions to keep delivery connected to real user outcomes.

Generative AI has transformed how we search, create and interact with information. The next evolution is Enterprise Agentic AI — where AI moves beyond answering questions to understanding goals, reasoning over context, using enterprise knowledge, interacting with tools and APIs, and taking actions within business workflows.

This session explores how organisations can move from traditional AI assistants and copilots towards intelligent agent-driven business applications, covering the evolution from AI Assistants to Enterprise Agentic AI, what makes a system truly agentic, using RAG and enterprise knowledge (MCP, integration layers) to ground agents in business context, single- and multi-agent architectures, human-in-the-loop governance and auditability, and practical enterprise use cases across recruitment, financial services, customer operations and workflow automation.

Drawing on her work architecting Derbee Agentic Studio and applying the Derbee Agentic layer within Stynt.ai, Sujatha will share a practical perspective on embedding intelligent agents into applications as a reusable data, intelligence and workflow orchestration layer.

Key takeaway: Beyond Answers → Into Reasoning → Into Decisions → Into Action.

This session explores the gap between academic learning and workplace readiness — covering practical experience, confidence, communication, mentoring, professional communities, and stronger collaboration between education and industry, to help students move from qualification to real workplace capability.

This talk explores the development and real-world application of Artificial Intelligence–driven robotic and humanoid automation in laboratory research, healthcare environments, diagnostic systems, and industrial manufacturing. The session will discuss how intelligent automation platforms are transforming modern laboratories, hospitals, and production facilities by improving precision, safety, efficiency, and scalability.

The presentation will also cover the use of AI-controlled robotic systems in agriculture, warehouse and supply-chain operations, food production, and retail automation, demonstrating how autonomous and semi-autonomous technologies are reshaping multiple sectors. Practical examples of laboratory automation, robotic assistants, smart monitoring systems, and intelligent control platforms will be presented, together with discussion on future humanoid automation and fully autonomous research environments.

Generative AI has transformed how we search, create and interact with information. The next evolution is Enterprise Agentic AI — where AI moves beyond answering questions to understanding goals, reasoning over context, using enterprise knowledge, interacting with tools and APIs, and taking actions within business workflows.

This session explores how organisations can move from traditional AI assistants and copilots towards intelligent agent-driven business applications, covering the evolution from AI Assistants to Enterprise Agentic AI, what makes a system truly agentic, using RAG and enterprise knowledge (MCP, integration layers) to ground agents in business context, single- and multi-agent architectures, human-in-the-loop governance and auditability, and practical enterprise use cases across recruitment, financial services, customer operations and workflow automation.

Drawing on her work architecting Derbee Agentic Studio and applying the Derbee Agentic layer within Stynt.ai, Sujatha will share a practical perspective on embedding intelligent agents into applications as a reusable data, intelligence and workflow orchestration layer.

Key takeaway: Beyond Answers → Into Reasoning → Into Decisions → Into Action.

Organisations generate enormous quantities of information, yet the evidence needed to understand decisions is often fragmented across emails, meetings, systems and records. Drawing on Keith Brown's MBA research across 154 UK charities and Community Interest Companies, this talk explores how AI may reduce the effort required to find, connect, compare and challenge organisational evidence. It distinguishes model, system and governance risk while considering provenance, auditability, behavioural bias, privacy and human accountability across RPA, IPA and agentic systems, ultimately asking whether cheaper scrutiny will make governance more demonstrable or create stronger ways of controlling the organisational record.

Most people use a mainframe every day without knowing it – when you tap your card, book a flight, or take cash out of an ATM, a mainframe is often doing the work behind the scenes. This talk explains what a mainframe actually is, in plain language, and why people are wrong to think of it as old or dying. We'll look at how it's built to never go down, how it can run software from decades ago without rewriting it, and why banks and airlines still trust it with their most important systems. We'll also talk about why there's a real shortage of people who know how to work with mainframes, which is good news if you're early in your career. Right at the end, we'll touch on one surprising thing: even though mainframes are everywhere, almost nobody can clearly explain what they cost to run, and why that's starting to change.

Key takeaways: you'll understand what a mainframe really is, and why "legacy" is the wrong word for it; see why mainframes are still critical today, and why there's a growing need for new people to work on them; and learn why the cost of running a mainframe is surprisingly hard to see, and why that matters.

Why now: most people, even in tech, don't realise how much of everyday life still depends on mainframes, and a real gap is opening up as experienced staff retire. At the same time, organisations are starting to pay more attention to what these systems cost to run.

Product teams can deliver a solution on time, within budget and against an agreed specification, yet still fail to create meaningful value. This happens when teams become more confident about the solution than they are curious about the user. A user mindset provides a practical way to correct that imbalance. It asks product professionals to understand the task people are trying to complete, the context in which they act, the constraints they face and the outcome that matters to them.

This session explores how a user mindset changes product delivery from discovery through to benefit realisation. It shows how user evidence can expose costly assumptions, improve prioritisation, reduce unnecessary scope and create stronger measures of success. The talk also distinguishes what users say from what they do and what changes as a result, explaining why interviews, observation and product data work best when used together.

Attendees will leave with a simple LENS framework: Listen to the user’s task and context, gather Evidence, Narrow the solution to the smallest useful change and See what happens after release. The framework can be used in backlog refinement, sprint planning, reviews and roadmap discussions to keep delivery connected to real user outcomes.

Building an AI agent is only the first step. A real agent needs a strong harness around it — the system that controls how the agent thinks, uses tools, manages memory, handles errors, tracks cost, and responds safely.

Without a good harness, even a powerful model can become expensive, slow, unreliable, or difficult to debug. This is why agent optimisation is so important.

In this session, we will explore how to design agents that are not just smart, but also practical for real-world use. We will look at how to reduce token usage, choose the right model for the right task, improve tool calling, manage context, add evaluation checks, and monitor cost before deploying agents into production.

The goal is to help developers move from “I built an agent” to “I built an agent that is reliable, efficient, and ready to use.”

Modern organisations generate and use data spread across multiple applications, databases, teams and data products. While organisations can move data between these systems, business concepts may be defined and represented differently. A customer, asset, product or transaction, for example, may have different definitions across systems or business functions. This can create ambiguity, duplication and considerable effort in reconciling data before it can be trusted and reused.

Ontologies and knowledge graphs provide a way to create a common enterprise language without replacing existing systems or forcing diverse enterprise data into a single, monolithic data model. Semantic mappings connect local data structures to shared concepts, allowing information from different systems to be integrated and kept semantically aligned. Controlled vocabularies using the Simple Knowledge Organization System (SKOS) can align different terms, synonyms and related concepts. Ontologies make business concepts, relationships and rules explicit, while constraints define what valid and consistent data should look like.

Building a common language starts with the questions an organisation needs its data to answer. These can be translated into competency questions that guide ontology design, semantic mappings, knowledge graph construction, constraint validation and inferential reasoning. The resulting knowledge can then be made available to applications, analytics and AI. This enables people, systems and AI applications to interpret and use enterprise information more consistently across organisational boundaries.

The impact can be measured through fewer conflicting or duplicated definitions, reduced manual data reconciliation and greater reuse of agreed concepts. Other measures include improved cross-system query capability, faster integration of new data sources and improved data quality. Greater traceability can also support more explainable enterprise decisions. A common enterprise language therefore goes beyond data standardisation. It provides a foundation for interoperability, governance, analytics, AI and decision intelligence, turning fragmented enterprise data into connected, contextualised and reusable knowledge.

This talk explores the reality of the digital divide in Africa—moving beyond the common narrative of limited internet access to uncover the deeper challenge: the gap in digital skills, opportunity, and inclusion. While connectivity is improving across the continent, millions of young people still lack the knowledge and pathways needed to fully participate in the digital economy.

Drawing from real-world experience in technology, community leadership, and youth engagement, Sobere Opusunju shares practical insights into how we can shift from simply providing access to creating meaningful opportunities. The session highlights the critical role of education, mentorship, and locally relevant digital skills in empowering the next generation.

Through examples and lived experiences, this talk will challenge assumptions, spark new thinking, and offer actionable strategies for individuals, organizations, and communities looking to make a tangible impact.

Attendees will leave with a clearer understanding of what it truly means to bridge the digital divide—and how they can contribute to building a more inclusive and opportunity-driven digital future for Africa.

This session explores the gap between academic learning and workplace readiness — covering practical experience, confidence, communication, mentoring, professional communities, and stronger collaboration between education and industry, to help students move from qualification to real workplace capability.

AI agents can now read documents, match records, draft invoices and write meeting minutes, yet most organisations cannot say what their agent is allowed to see, what it may decide, or how a mistake would be caught. Drawing on earlier data quality experience in utilities and professional services and on systems built in the INNOVATION FORUM Foundry for Innovation Forum, this talk sets out the engineering patterns that make agentic work safe to adopt: sealing sensitive data away from the model, extraction contracts with provenance, reconciliation against a single source of truth, and human authorisation before anything leaves the building. It then asks who gets left out. Open source and small language models running on modest hardware put these patterns within reach of charities, councils and small firms that cannot afford enterprise platforms, and low-code validation lets non-programmers check the results. Smaller models and processing each item only once also cut the energy and carbon cost of every task. Finally, it treats bias as an engineering control: statistical bias in the data is measured, behavioural bias in the people using it is governed, and every decision can still be traced to a person.

Security Operations Centers are facing growing volumes of alerts, increasingly sophisticated attacks and a shortage of skilled security professionals. Traditional SOCs still rely heavily on analysts to investigate alerts, correlate events and determine the appropriate response across multiple security tools.

This session explores the evolution towards the AI Autonomous Cyber SOC — where AI can continuously analyse security telemetry, correlate events, identify anomalies and attack patterns, support investigations, prioritise threats and accelerate response.

The session will examine how AI can move the SOC beyond simple automation towards an intelligent operating model in which AI assists with detection, investigation, threat hunting and response, while humans retain appropriate oversight and accountability.

It will also explore the practical challenges of autonomous security operations, including explainability, AI accuracy, access controls, governance, human-in-the-loop decision-making and determining which actions can safely be automated.

Attendees will be introduced to a practical maturity model showing the journey from traditional SOC → automated SOC → AI-assisted SOC → AI-driven SOC → governed autonomous SOC, providing a framework for organizations considering how to introduce AI into their security operations.

University research can create value long before it generates revenue, yet investors often lack ways to assess that value or identify who benefits. Dr. Proselkov will present a proposed world model for tracing how research generates practical, societal, and financial impact via the stakeholder network, following a normative-empirical approach to determining objectivity through stakeholder networks. The model connects universities, investors, practitioners, and beneficiaries, linking evidence of progress to each stakeholder’s goals and capacity to act. It explores how the same research can produce different outcomes depending on who provides funding, access, expertise, or routes to adoption. The aim is to support earlier investment and more effective coordination, with impact assessed through evidence gathered as projects develop.

Artificial Intelligence has become a strategic priority for organisations across every industry, yet many AI initiatives continue to struggle in delivering meaningful business outcomes. While businesses are generating more intelligence than ever through predictive models, Generative AI, and advanced analytics, the expected improvements in operational performance, efficiency, and commercial value often remain elusive.

This session explores the critical gap between AI-generated insight and real-world execution. Rather than focusing solely on models and technology, it examines why successful AI adoption depends on strong foundations in data, enterprise architecture, governance, and operating models. Through a practical business case, the session demonstrates where AI initiatives typically break down, why intelligence alone is not enough, and how fragmented workflows, disconnected systems, and unclear decision ownership prevent organisations from realising value.

Attendees will gain a practical framework for designing AI systems that do more than generate recommendations—they enable timely decisions, integrate seamlessly into business processes, and drive measurable outcomes. The session concludes with actionable guidance on how organisations can move beyond AI experimentation to build scalable, trusted, and business-focused AI capabilities that deliver sustainable impact.