The continuous drive of artificial intelligence led by Generative AI (GenAI) is set to be one of the most transformative forces shaping business and society. Unlike previous waves of digital innovation, this revolution is defined by intelligent, human-controlled AI assistants capable of interacting naturally through text, speech, images, and video. These AI tools will perform the groundwork– drafting content, analysing data, generating ideas, and responding to complex queries – while humans guide, verify, and refine the output. This powerful collaboration between people and AI will redefine productivity, creativity, and decision-making across industries. Beyond efficiency and innovation, the positive power of GenAI extends to social benefit – helping bridge knowledge gaps, support education, empower small businesses, and enhance accessibility for individuals with diverse needs. In essence, GenAI represents not just a technological leap but a chance to shape a more inclusive, equitable, and intelligent future.
AI training data biases
Training data are the large sets of information that teach AI models to find patterns and make predictions. The quality and variety of this data are crucial for the AI’s success. If the data is biased or lacks diversity, the AI may make poor decisions. For example, an AI trained mostly on images of young, light-skinned people might struggle to recognise darker-skinned or older faces. Gaps in training data can arise from limited information, inconsistent labelling, and a lack of diverse viewpoints. Regularly reviewing and improving training data is vital for creating fair and reliable AI systems. This requires collaboration among data scientists, experts, and affected communities. This is being addressed by retrieval-augmented generation (RAG), a technique for enhancing the accuracy and reliability of GenAI models with facts fetched from external ‘trusted’ sources.
Challenges with AI-generated information
At times, GenAI can generate inaccurate information, which includes:
- – Opaque information: Hard to understand and unclear.
- – Inaccurate information: Incomplete or inconsistent data not intended to mislead.
- – Misinformation: Deliberate falsehoods meant to deceive, often propaganda.
- – Inappropriate information: Content that can lead to unlawful or dangerous behaviour, such as advice on illegal activities.
- – AI hallucination: Outputs that sound plausible but are factually incorrect or unrelated to the context.
To tackle this, we have developed AI tools that enable professionals to create, monitor, and manage personalised support for the clients through controlled environment, called a walled garden. It limits users’ access to specific online content and services. This approach steers users toward helpful information while blocking access to harmful/fake material.
Importance of keeping a human in the loop
The primary reason for having a ‘professional in the loop’ with any AI platform is to create accurate educational content and monitor chatbot-generated information, preventing misinformation and deepfakes.
AI hallucinations pose a significant challenge, as they involve the generation of confident yet incorrect responses. These inaccuracies stem from limited or biased training data and a lack of information on certain topics, leading to overconfidence in outputs that may hold errors.
The GENAIE educational platform
The GENAIE platform was designed to provide AI assistants for public sector key workers and deliver eLearning content across various educational and vocational areas. Building on the Meganexus Community Campus software used in the UK Probation Service and Prisons, GENAIE offers secure access to tailored learning materials in multiple languages.
The system lets key-worker professionals to create personalised lessons and resources using GenAI, streamlining course development and supporting moderated delivery. It provides lesson plans, resource links, and extra-curricular activities like quizzes and flashcards, all customisable by teachers.
Content can be published in various formats, easing lesson preparation and allowing teachers to focus on individual learners. GENAIE operates independently or integrates with existing LMS (learning management system) platforms, delivering personalised education and automated content tailored to each user’s needs. Its microservices architecture ensures scalability and flexibility in service development.
Conclusion
Our ‘AI for Social Good’ mission is to empower people and businesses through knowledge transfer using AI tools. Our GENAIE tool is being successfully used to deliver training.
GENAIE created customised courses for the Community Campus supporting People on Probation. These courses have adopted an animated style and feature a diverse range of characters representing varied demographics, reflecting the diversity of today’s society. Moreover, all courses are available in 10 languages, aligning with HMPPS’s cohort. Community Campus is accessible across England and Wales, supporting over 53,400 users with over 596,600 hours of education, training and employment learning.