The art of the possible: How to balance opportunity with feasibility when implementing AI

Rohit Dhawan, Chief AI Executive and Group Head of AI

The art of the possible: How to balance opportunity with feasibility when implementing AI

^{Financing UK infrastructure}^|^{Embracing blockchain}^|^{Longevity risk}^|^{Accelerating digital trade}^|^{Implementing AI}

Artificial Intelligence (AI) offers unprecedented possibilities for businesses. With most firms now taking their first steps on what will inevitably be a lengthy AI implementation journey, questions are arising over where to start, what to prioritise, and what’s tangibly achievable. Rohit Dhawan, Chief AI Executive and Group Head of AI, Lloyds Banking Group, explores how firms can successfully plot a feasible course towards AI adoption in a landscape defined by rapidly changing technology.

Rohit Dhawan,
Chief AI Executive and Group Head of AI,
Lloyds Banking Group

Prioritising possibilities

While many organisations are now seeing their investment in data and analytics bear fruit, they have yet to plant the seed when it comes to establishing dedicated AI functions. Many are still in the very early stages of understanding not only the technological landscape, but also how different eventualities may play a role in the future of their business.

The emphasis an organisation should place on AI therefore depends entirely on how their industry will be affected by it. While some may be relatively unchallenged – at least for now – others will face game-changing disruption as AI transforms traditional practices. For organisations that will be impacted, a significant amount of investment may be required in order to take the right steps to ensure that both the business and its technology can be reimagined.

A crucial part of this process will involve practitioners taking the time to educate management on the art of the possible. If senior executives fully and accurately understand the role, cost, and potential of new AI models, then the business will be better placed to make effective decisions and embrace the technology.

^{Financing UK infrastructure}^|^{Embracing blockchain}^|^{Longevity risk}^|^{Accelerating digital trade}^|^{Implementing AI}

Once firms have identified the biggest potential opportunities, it’s key to assess the feasibility of undertaking them. Certain AI use cases demand large-scale implementation projects and significant investment, so firms should analyse their business to identify areas for implementation that are high value and low risk.

Building on solid foundations

A significant challenge faced by practitioners looking to adopt AI is ensuring cohesive integration with existing systems. Generative AI models are designed to directly interact with different end users – be they colleagues or clients – and this requires a delivery mechanism or an interface through which it can operate. For most companies, this will be a pre-existing platform – it’s highly unlikely that a firm will request that customers use an additional channel just for AI interactions. As a result, AI projects must be implemented with seamless integration into existing systems in mind.

System integration, in itself, instigates further challenges. General purpose Generative AI models rely on large datasets sourced from the internet, which can sometimes be inaccurate, lack context, or produce misleading or incorrect outputs, known as ‘hallucinations’. The AI system must therefore have a means to access recent and relevant company data to ensure a high level of accuracy when it comes to serving end users.

On top of this, businesses must consider regulatory and compliance issues, especially for firms working in highly regulated industries like financial services. AI tools can create opaque ‘black box’ spots, making it difficult to understand how a model arrived at a decision or recommendation.

Sourcing relevant and timely skills and maintaining learning and development to match the pace of change must therefore form one of the key pillars when weighing up possibilities.

Rohit Dhawan,
Chief AI Executive and Group Head of AI,
Lloyds Banking Group

The use of AI must be transparent, accurate, auditable, and compliant with regulations – especially under frameworks like the EU AI Act, which requires disclosure, traceability, and explainability for high-risk and general purpose systems. This presents significant integration challenges, as practitioners must construct robust operating models that document, expose, and track decision making for audits and regulatory review. Lloyds Banking Group addresses these challenges through a dedicated and industry-recognised AI Assurance Framework, ensuring AI is responsibly governed, explainable, and aligned with both internal and external regulations. The Group’s assurance system operationalises seven Responsible AI Ethics Principles, via clear guardrails, a central AI risk repository, and embedded oversight structures, strategically aligned with international standards.

^{Financing UK infrastructure}^|^{Embracing blockchain}^|^{Longevity risk}^|^{Accelerating digital trade}^|^{Implementing AI}

Running to stand still

Vital to artificial intelligence is human intelligence. Another core challenge faced by practitioners when implementing AI is sourcing the correct skill sets, particularly against a backdrop of rapidly evolving technology where new systems can be obsolete within months. Whereas older technologies have reached a level of maturity, AI is an evolving technology and methods or practices are becoming irrelevant quicker than we’ve ever seen before.

The pace of change means that experts are continually having to consume new technologies just to do their day job. With so many nuances – such as using the correct models, cost optimisation, or building the right interaction pathways – this can represent a real challenge for organisations looking to remain ahead of the curve. Sourcing relevant, timely skills and maintaining learning and development to match the pace of change must therefore form one of the key pillars when weighing up possibilities.

Think big, start small

These challenges point to the initial strategic deployment of AI on a relatively small scale. That is, practitioners should identify large areas for opportunity, but build up incrementally through small-scale projects. This approach is both more likely to be feasible and successful, considering that it is easier to implement from a technical, regulatory and skills perspective.

It’s also easier to measure the success or impact of an AI project when implemented on a small scale. Firms can not only measure the return on investment, but also look at other tangible metrics such as productivity gains or even the impact on colleague engagement as AI increasingly forms part of their day-to-day role. It becomes easier to measure these factors when working on a smaller scale.

^{Financing UK infrastructure}^|^{Embracing blockchain}^|^{Longevity risk}^|^{Accelerating digital trade}^|^{Implementing AI}

From algorithms to agents

There are currently over 800 advanced analytics and AI models in production across Lloyds Banking Group, many of which have been implemented on a small scale to build up to a radical transformation of our business. And this transformation is well underway. For example, customers use the Group's large language model-powered search feature for more than 4 million searches monthly, reflecting rapid adoption at scale.

Over the next year, we expect to witness another transformative shift as businesses – including our own – embrace Agentic AI. This class of AI enables autonomous decision making in AI agents, unlocking the potential for AI to complete specific tasks independently. We expect to see more and more use cases where AI can be used as agents to make autonomous decisions or take action, initially in straightforward ways, against a defined set of criteria and ethical principles we have put in place. We’re excited to see the opportunities this will begin to unlock across the entire economy.

Given the unprecedented pace of change, AI adoption should be approached with small, strategic, and carefully planned steps to ensure successful integration and long-term impact. It is clear that nearly every organisation has already started engaging with AI, and the choice to embrace these technologies proactively will determine which businesses remain competitive and which risk being left behind.

^{Financing UK infrastructure}^|^{Embracing blockchain}^|^{Longevity risk}^|^{Accelerating digital trade}^|^{Implementing AI}

Key takeaways

From a technological point of view, we are still in the foothills in terms of AI’s potential to revolutionise business operations.
Practitioners should start with small-scale AI projects, which allow for manageable implementation, measurable outcomes, and easier adaptation.
Within the next year, we expect the widespread adoption of autonomous Agentic AI models which will make a strategic approach to AI adoption essential.
Lloyds is actively leveraging AI across the Group, with over 800 advanced analytics and AI models currently in production.