Bringing AI to life

05 December 2019

This article looks at how artificial intelligence (AI) is set to transform the professional services sector.

  • Introduction to AI

    It’s now commonplace to hear that artificial intelligence (AI) is set to transform professional services – yet many firms have yet to feel its impact. However, significant advances in AI are likely to make it increasingly relevant to you and your firm. 

    We’re all becoming more and more familiar with various types of AI. 

    For example, some companies are already using AI in the form of a front desk ‘receptionist’ that combines robot and chatbot technologies. Pepper the Robot, manufactured by SoftBank Robotics, is perhaps the best-known example of this kind of ‘front-of-house’ application, and is being widely used in Japan by banks, medical facilities and restaurants.

    There is growing interest in the vast array of AI uses across many sectors. Its capacity for quickly processing information might have application in correctly prioritising people who are waiting to access various services. 

    In healthcare, work is being done to use chatbots trained in cognitive behavioural therapy to manage the early stages of depression, while research by the Massachusetts Institute of Technology involves AI to listen to the way people talk to predict the onset of depression. In the UK, DeepMind is applying AI to the NHS database of case records to try to predict age-related macular degeneration.

    AI covers a broad range of technologies, but in essence it can be defined as any computer program or system that does something that we would normally think of as intelligent in humans. 

    The benefit of AI is that it can improve on what humans can do on their own because of the complexity of the task or volume of data, the speed to insight needed or because manual or other traditional approaches are inefficient.

    Artificial Intelligence employs technology and algorithms to automatically extract concepts and relationships from data, ‘understand’ their meaning, learn independently from data patterns and prior experience, and, crucially, interact with humans in a natural way.

  • Types of AI

    The range of different AI technologies using advanced data science includes speech to text, text to voice, image recognition and classification of language. The basis of all AI is cognitive computing, a system that mimics but doesn’t replicate the human brain – the brain of course is so extraordinary that, in many respects, we’re still not really sure how it works. 

    However, what’s important about these technologies is that they're like Lego bricks – they can be stuck together, and in a variety of different ways. For example, a smart speaker product like Alexa hears what you say (speech to text), extracts key information (language classification) and then plays a tune or tells you where the nearest book shop is. So, another way to think of AI is the stringing together of technologies to get something done.

    Alexa is of course produced by Amazon, just one of many global players in the AI ecosystem, and there is fierce competition in this sector. However, rather than being proprietary, much of the AI technology is open source, which means that it’s relatively easy to access and get started.

    In practical terms, AI can currently be split into three main types.

    One of the core strengths of AI is its ability to classify text and images. These tasks are carried out by deep learning computing systems known as neural networks – so called because they look and behave a bit like a brain’s neural network. 

    To give a practical application in professional services, a neural network can be trained to automatically categorise client emails. Each email could be as different as the person who wrote it – unstructured, rambling even – and about different subjects, from, say, a late report to a fee dispute. Simply by being exposed to these complaints, the AI system can learn to recognise and classify the emails with 80% accuracy. As a safeguard, such a system can trigger the intervention of a human if it’s unsure how to classify an email. 

    The whole process represents a daisy chain of different technologies that can save precious time. For example, NHS Scotland is testing automatic assessment of GP referrals for urgency of care, freeing up clinicians to focus on other things. 

    The use of cognitive conversational search with voice-activated chatbots is now quite common in dealing with a range of basic queries – everything from lost credit cards to IT support – and means call centre staff can focus on more complex client questions. 

    This area of AI is developing incredibly quickly. Products like Google Duplex offer the possibility of chatbots that can book, for example, a customer’s hair salon appointment and sound so human that they’re unaware they’re not talking to a person. Google is now using such AI advances to move into the call centre sector.

    So-called expert systems can make use of AI to help with anything from avoiding a parking ticket to making PPI claims, or even helping a physician – through examination of a medical evidence database – to make an informed diagnosis. 

    A great example of insight-gathering AI that’s already operating on a large scale can be found on board many Tesla cars. Their autopilot function runs continuously in the background, whether engaged or not, collecting the data that is helping the company move towards its target of a fully autonomous car by the end of 2019.

  • Managing the risks of AI

    In this fast-developing space, the application of AI won’t always go to plan. 

    An example of the risks of unsupervised machine learning is Microsoft Tay, a chatbot released via Twitter in 2016, which was supposed to respond intelligently to tweets from its intended audience of teenage girls. Instead, when Tay was targeted by internet trolls, it began to send out offensive messages, which forced Microsoft to close it down. 

    This demonstrates the importance of supervised AI learning. Just like teaching a person, the objective is for AI to become skilled at something, but not to learn ‘bad habits’. 

    So, it’s often not simply a question of training AI how to do something and launching, but monitoring how the system is performing post-launch and then retraining it periodically. 

    The data on which AI is trained in the first place is also critical – if the data is biased or misleading in some way, this will inevitably impact its performance. To mitigate this, publicly available curated data sets, certified as bias-free, have been created. 

    Effective data governance is therefore paramount in the application of any AI.

  • Key takeaways
    • Start with the business problem you want to fix rather than the technology.
    • Identify whether your focus is on the use of AI to transform your firm’s strategy or to optimise ‘back office’ processes. 
    • If the business case is to save money, evaluate whether automating a task is cheaper than simply outsourcing it.
    • Training AI with the right data is critical – it’s essential to involve people who understand your business and your customers in development. 
    • Innovate, experiment and combine datasets – one proof of concept may lead to spin-off uses in the business.
    • Consider how team roles and profiles will change – you may need to rethink how to attract data science experts and offer appropriate incentivisation and career progression.
    • Stay focused on data governance issues like GDPR, and always ensure you have proper consent around personal data.
    • Remember the pace of digital change is accelerating all the time – sticking your head in the sand and pretending new technologies will go away isn’t an option.

Read related insights


Tech Forum

Insights on the latest innovative technology for forward thinking corporates.


Legal Sector Lowdown

A series of articles discussing key issues and opportunities for the legal sector, from expanding overseas to investing in innovative technologies.

Contact us

Get in touch

To discuss your business requirements and how Barclays can support you, contact us today.