AI is more than just software
The traditional role of generative AI has been to respond to a question (a “prompt”) and return a response. In the same way that an AI can provide general information (“what are the ingredients of bread?”), it can also create a plan (“what are the steps to make bread?”).
By combining the plan with access to “tools” and memory, agents can undertake complicated multi- step tasks. Each step of the plan can be fed to another agent to execute and further agents are able to determine whether the task has been completed or needs to continue.
For example, a group of agents can respond to a wide-ranging customer enquiry, seek out a product or build a complex research report.
Selecting the right problem
Generative AI is tremendously powerful. It works by taking an input and extending it, based on its massive training data. But it does not “understand” in the conventional sense and so can get things wrong.
Steps can be taken to minimise errors – using clean data, breaking the problem into unambiguous steps, introducing reasoning that can be validated and human oversight all play a key role.
But choosing the right application is key – those that add value and are tolerant to small errors. For example, a tool that classifies the sentiment in social media can afford to mis-classify the odd post. Or a tool that suggests products for a particular application provides a valuable service and the small number of answers can be easily checked by a human.
Building a Resilient Application
Key to any AI application is “Agency” – in other words, what can the AI actually influence. The greater the agency, the stronger the controls that are needed to ensure the behaviour is as intended.
For applications with high degrees of agency, good project design is essential. This starts with basic principles (processes and ethics), though to system design (monitoring and guardrails) to ensure that from the outset, the AI system has been built with reliability in mind and that there are well established processes in place to handle any unexpected deviations in a defined and controlled manner.
01
Proof of Concept
AI is relatively new to most organisations and there are often varying levels of awareness about its capabilities and the impact on the rest of the business. A proof-of-concept application is a rapid and low-cost way to help demonstrate the capabilities and value of AI and to allow all stakeholders to understand and manage any changes that may be needed should the application be rolled out at scale (for example, staff training).
A proof of concept application will typically be developed in a “sandbox”, disconnected from live systems but use representative data to enable business decisions to be made on how to proceed.
02
Data Security
Once past the proof of concept stage, like any software application, data security is a key consideration. This is a combination of cybersecurity, personal privacy, protection of confidential information and regulatory compliance. Many jurisdictions will have guidelines on the data that can be exported outside national boundaries and companies will wish to be confident that proprietary information is not exposed. This leads into choices about what AI models to use, where they operate and how the data is presented. For example, in some cases it is sufficient to anonymise personally identifiable information and use services hosted in another country. In other cases it may be necessary to host the solution locally, either in-house or at a reputable data centre.
03
Consider AI as a Service
Cloud computing has become a way of life for banks in the last decade. It eliminates much of the complexity of software service hosting, including many of the considerations around security, reliability, backup and fail-over. AI is more complicated than conventional software because the underlying models and platforms are changing at a rapid rate (quarterly) and AI has additional regulatory and ethical components that need to be managed. “AI as a Service” goes one step beyond cloud computing, providing a consistent interface to existing banking systems while handling the complexities of continuous upgrade and performance monitoring. Particularly in the early phases of AI adoption, AIaaS can offer a faster, lower cost and simpler approach to adoption of AI applications in banking.
The AI landscape is moving very rapidly and new technologies are being announced all the time. For some, this is a reason to “wait and see”. But AI is more than software, rather like the Internet, it involves organisational change, that touches on management, processes, training, products and marketing.
AI will be transformational for the banking sector, removing repetitive activity, freeing staff to focus on high value tasks and creating customer delight with improved responsiveness and personalised services.
By taking a pragmatic and opportunity-driven approach to AI adoption, businesses can take advantage of the many benefits of AI today and position the organisation to be able to rapidly adopt the new advances as they evolve.
Akili AI is here to help. We bring a combination of market-leading skills, foresight and banking knowledge to support banks in the AI adoption journey to deliver a productive AI future while managing the transition to the new way of working.
Delivering Digital Transformation with Responsible AI
It is easy to get started with a simple clear approach:
- AI audit – What AI does the bank use today and how are they managed
- AI Governance – Put in place a simple but effective AI governance process
- AI Opportunity Discovery – Map the places where AI can have most impact
Getting started with AI is about choosing the right problem. One with clear business benefit, that can be readily integrated into your existing systems. And one that aligns with the ethos and brand to enhance the customer experience.
Talk to us to start your AI Transformation Journey!