Generative AI and its talent and change implications within FS – 2 of 3: what is the impact of Generative AI on work across banks and insurers?

For those who don't know I am a painter. A messy but enthusiastic one. Some of my favourite moments are painting with my daughter and mum quietly together. We went to the David Hockney production at the Lightroom last weekend and I was struck by someone who'd painted professionally for 60 years, but had continued to elevate his craft using new technology. I am fascinated by the interplay between technology and human ingenuity.

In my last blog, I explored generative AI and its potential impact on work in financial services.?Where we left off was the need to consider human, automated, augmented and emergent work in order to get the most from generative AI and use it to elevate human skills.

Using that lens, in this blog, I look at some of the potential applications of generative AI to work within banks and insurers – and some of the new work that’s likely to emerge as a result – and close with some thoughts on how we might elevate the best of humanity through AI. Let’s think about some of the implications ‘front to back’ across the bank or insurer:

Generative AI for Customers and Front Office Teams

Generative AI can transform customer and client interaction, providing more personalised experience within digital channels online, on-mobile and in-person. Interactions with self-service tools, chatbots and SMS will become more human-like – as will inbound and outbound correspondence – including between different languages. This is incredibly important for banks seeking the right tone and relevance in customers lives – and for those customers who want that level of personalisation and happy to talk knowingly with bot.

Beyond channel personalisation, generative AI can empower customers with personalised information, research and even basic advice that they may not have previously had access to. This can be helpful for clients who like to self-serve or may have been in segments where these services couldn’t be provided before – e.g. providing quality investment research advice to ‘mass wealth’ clients or business planning advice to startups and small business. Banks and insurers will need to be careful about how far this extends from a conduct risk perspective, especially with intense focus on fair treatment and customer outcomes (e.g. the upcoming Consumer Duty regulations in the UK).

More accessibly, generative AI can support the advice and service provided to customers and clients by relationship managers, coverage teams, brokers and advisors, working as their professional ‘co-pilot’. Imagine a relationship manager preparing for a meeting and being ‘coached’ on the client background and meeting attendees, maybe with support around understanding needs, researching solution options from across the bank and market, or even drafting proposals for the client. The AI could also help with the initial review of customer applications and business plans. Used well generative AI will help the RM with more time to think about and be with clients – and reduce the size of middle office teams. All this ‘generated’ content will still need a human to review, check and add to the advice provided (after all, the AI will not always get it right or see all the information/history with the customer). Equally this does not remove the need for human delivery of advice and support with emotional intelligence and adjustment within a discussion (after all, the AI cannot really listen or show true empathy).

Within investment banking and market facing roles, similar opportunities are present. Generative AI can provide research analysts or traders with summarised insights, analytics and evaluations on large and more diverse data sets – and the tailoring of investment and economic research into strategies. The previous lengthy process of drafting hundreds of pages of PowerPoint ‘information memorandums’ and supporting literature, research and financial information for mergers, acquisitions, divestments and investment rounds, can be compressed allowing deal makers to focus more on understanding the market, value, potential and relationships.

While there is much opportunity for augmentation of human work in the front office and some automation through client/customer self-service, the work and skills involved in solving more complex client problems and needs, developing empathy and human relationships, and providing judgement in more complex client and market transactions, will remain strongly human led. Each bank and insurer will need to decide how much they want to take saves from AI and reinvest back into frontline teams to grow the coverage and skills needed for market growth.

Middle and back office

There is already significant technology change in the middle and back office of banks and insurers. Simplification, analytics and straight-through-processing have reduced volumes – and work orchestration, robotics and earlier waves of AI have been applied to further automate residual workloads. Done well, this is about not only about efficiency, but also more responsive and intelligent operations that enable client experience and growth.

Generative AI adds to this opportunity. Process inputs can be improved, by checking for gaps and mistakes and converting unstructured data (e.g. email) into structured process inputs. It can support how to handle cases, searching internal procedures to find the correct operational steps and delivering these into workflow systems, or semi-automating operational risk and compliance controls. Generative AI can help personalise account or policy documentation and better handle, store and retrieve inbound documentation and correspondence. Some process steps can be significantly augmented, for instance basic research and fact checking in insurance claims through to assessment of video and image inputs sent by clients and assessors. Finally, much work is subject to a ‘4 eye check’ on process outputs – imagine being able to split this between the scrutiny and consistency that AI gives you and the insight and judgement that a human gives you.

In the middle and back office this means potentially much smaller workforces, but doing significantly higher-value and higher-skilled work, especially exercising judgement on complex exception cases. Making this transformation of work and skills a compelling journey and professionalising the operations workforce is vital.

Functions

In most corporate functions, we have already seen the impact of platforms in reducing the processing workloads with basic support services and a corresponding move to planning and insight skills within functions professionals. Generative AI will accelerate this trend.

In Marketing and broader communications and investor relations functions, generative AI has the potential to create even greater personalisation at scale and augment creative capabilities. This will include generating options and tailoring draft marketing, digital and social written and media content – and support to manage, edit and demise content too. It will also support teams in condensing diverse internal and external sources (e.g. press articles, correspondence, social media, complaints cases etc) in order to create more insightful and dynamic ‘listening’ inputs on customer, colleague, shareholder and community sentiment, feedback and needs. In an increasingly digital world, financial institutions have become functionally complete, but sometimes emotionally detached and needing more personality. So truly strategic and deeply creative skillsets becoming even more highly valued, but will work alongside generative AI for scaled impact.

HR and people management is one of the more sensitive areas and opinion varies on what is an acceptable use of AI in people management, given the serious impacts of decisions for people and employment law. However, with the right parameters and management, there are some brilliant uses. As a service provider, HR operations typically handles many routine queries that can be served faster and more effectively through conversational chatbots and helping employees better search and retrieve answers from guidance and policies. As for the front office, correspondence and communications with employees can be improved and personalised through drafting support from AI. Within talent acquisition, the laborious tasks of drafting content for sourcing channels, job descriptions, interview guides and assessment scenarios can be taken by generative AI (albeit recruiters also need to contend with candidate plagarism) and candidate experience and onboarding can be better personalised. Within talent development, there are applications to improve skills insights and matching colleagues to opportunities. There’s opportunity to generate micro learning content, personalise learning and deliver more dynamic authoring on new subjects, especially in the space of collective intelligence and social learning inputs from crowds and networks. Leaders can be better supported too, for instance in interpreting people data or elevating the frequency and quality of coaching and performance feedback by giving the leader a starting point based on diverse data sets. In all these cases, generative AI will need to augment even stronger professional HR capability and people leaders across the business.

As Finance becomes more focused on planning, advice and insight, generative AI can support these tasks, such as summarising and interpreting financial information and reports – and preparing draft insights and advice for the business. Generative AI opens up new Risk and Compliance needs (covered in future blogs), but also has implications for detection of fraud, financial crime and cybersecurity. It can be used for synthetic data generation and risk factor modelling. Like finance it can be used to provide draft summaries and insights across large risk and compliance reports, allowing more concentrated leadership and second line attention on resolving issues and preventative measures. Finally, it provides opportunities for summarising and interpreting new regulations and supervisory notes – and preparations for response. In Audit, it has similar implications including summarising large volumes of evidence and the drafting of findings. In Legal and Supply Chain lots of manual work around searching and drafting contracts can be removed. Across all these control functions, the AI does not take away the vital role of professionals in strong advice, controllership and strategic planning.

Unsurprisingly, Technology, Data and Security have been quick to look at the application of generative AI to their own work. As for other support functions, the IT helpdesk can be significantly improved through better chatbots, guidance search, case enablement and improved correspondence. Adoption within engineering for code and configuration generation is already widespread, including code translation between programming languages, generating code variants (e.g. AB testing different digital and mobile front end variants) and creating better system notifications, help and documentation. For instance, Github Copilot and similar platforms can typically enable 30-40% of base code generation. Generative AI can support data cleaning and completion, as well as summarising large data sets for anomalies, patterns and insights, as well as generating synthetic data for testing.

In summary, from the front to back office and across all the corporate functions, there is significant opportunity to automate and augment human work, requiring smaller but more skilled workforces.

What is the emergent work associated with generative AI in financial services?

There will be new ‘emergent’ work associated with generative AI. Many of these roles and new skills will be in technology and data:

AI and Data Science. Kind of obvious, but we will need more of the data science and coding skills needed to train and use AI well. As for most technologies there is a talent shortage at the start of the S curve. Demand picks up rapidly, skills will be in scarce supply and attracted to be best work and value propositions, and wages will peak until demand normalises and supply increases. So banks and insurers who want to be leaders in generative AI will need to become net creators of AI talent (rather than just consumers). ‘Strategic AI scaler’ companies are 1.5-2.5 times more likely to professionalise AI ahead of their industry with clear roles and accountabilities, dedicated multidisciplinary teams (architects, testers, data scientists etc.), formal education and deep training, defined processes and democratized AI literacy across the organisation. Investment in these skills and professionalisation journeys is a precursor to successfully scaling generative AI.

Data management. To use generative AI effectively, you need to be able to ‘train’ the AI with the right data and keep it refreshed. Banks will need to be able to explain and verify the source and validity of this data – the third parties they work with will need to help with verifying the training of the foundation models. As banks train the AI using domain specific data from inside their organisation, so data needs to be treated more as a curated asset, with investment on data architecture and data management roles. To scale up generative AI it will need to be integrated with core systems, so good integration and API development skill sets will be essential too.

Technology and Ecosystem Management. Scaling up generative AI will depend on having the right technical infrastructure, demanding strong technical architecture and cloud skills. The processing demands of generative AI will create the need for new sustainable IT and energy work within technology. Much of the progress made on generative AI will be made by working with the best of the ecosystem, demanding strong ventures management and ecosystem partnering skills, challenging traditional working practices within technology procurement.

However, it’s not just new technology and data work. There will also be ‘emergent’ work associated with generative AI in the business and functions:

Product or Journey Owners. The potential of generative AI will only be realised through business product or journey owners who can understand the application to work within their businesses and customer journeys. These leaders need to be able to lead the wider adjustments to value chains and business units, organisation structures and teams, organisational culture and human roles and skills, alongside wider changes in technology and data. These leaders will need strong client experience and proposition teams around them.

Responsible AI and Risk Management. In my next blog I will talk about some of the responsibilities and risks associated with generative AI, but there will be a lot of work in defining what responsible AI means for your business and operationalising it. This starts with defining responsible AI principles and risk appetite in line with emerging laws and regulations – and reflecting the purpose and values of your business. These principles need to be translated into effective governance structures, policies, and risk and control architecture, ownership and reporting. Then there is the work of helping teams consistently create use of generative AI that is ‘responsible by design’, for instance, checking models are working as intended, with valid and reliable results and making sure models are as free from bias as possible, especially in customer outcomes and treatment. This work extends into explaining the step by step working of the algorithm and the factual explanation of its decisions to customers, colleagues and regulators.

Checking and Judgement. OK, so not many of us set ‘checking things’ high up the list of ‘jobs I want to do when I grow up’, but checking things well is about to become an incredibly important job. This is particularly important for banks and insurers who rely on getting things right consistently.?When they don’t, the human, economic and reputational consequences can be dire. To use generative AI you need a ‘human in the loop’. The beginning might be a human asking the right question(s) – or a system may trigger that question from a customer interaction. However, checking the output of the AI must be done by a human before it is published or used with another human (at least for any generated information of importance, especially anything going externally). Generative AI can play an incredible role ‘drafting’, but we should be really careful not to give it the ‘send button’. To check work properly, we will need the sorts of skills that journalists, historians, anthropology, philosophy, literary and theology majors will have built in:

-?Editorial skills, checking and improving copy generated by AI

- Fact checking, making sure the AI has not misread information or made things up

-?Ethical decision making and judgement, understanding whether something is right or wrong, whether it reflects the spirit and letter of the law and if it reflects the company purpose and values

-?Critical analysis of the ‘thought pattern’ of the AI, challenging biases and assumptions, using counter-examples and arguments, changing the frame of reference and deep questioning

-?Research skills to bring new tangential data, especially behavioural science and anthropological techniques to understand human behaviours, augmenting the AI findings

As a sidebar, we will shortly be unlikely to be able to sort AI-generated from human created content in many fields, without the support of AI. There are lots of varied and amusing examples of teachers asking ChatGPT to assess whether an essay was written using itself and voting games on social about ‘which painting was AI generated – wrong!’. Boris Eldagsen recently won and then subsequently turned down a prize at the Sony World Photography Award with his haunting entry "Pseudomnesia: The Electrician", which was AI-generated, but convincing enough to fool the professional judges. While checking is an important area of ‘emergent’ human work, ironically we may well need to augment it with digital capabilities to be effective.

Collective Intelligence and Curation. Large language models can consume all digital data and content within a business and ‘know’ everything within that data set. Rather than relying on individual experts it can be used to bring together multiple experts with the power of the crowd. However, we already have too much data and information and not enough insight and knowledge inside banks and insurers. Curating content will become vital work. Over time this will extend to checking and verifying digital assets of customers. There is a financial and carbon cost of processing, maintaining and storing large amounts of content, so curating and demising generated content carefully will become important. Curating the ‘collective intelligence’ of the organisation will become a source of competitive advantage and an economic and sustainability imperative.

Commercial banking and insurance relationship management. Generative AI will accelerate innovation and new value streams in existing industries creating demand for financial services – for instance accelerating DNA or protein research in life sciences. It may even open up entirely new industries – for instance around digital assets. Startups and large corporates alike will need investment, lending and other financial services to grow. The data services and cloud industry – and underlying sectors like semiconductors – will see increased demand and the need for infrastructure investment. Companies may need to insurance against their AI risks. Savvy banks and insurers will see the growth and investment opportunities of generative AI, leading to emergent work in these proposition and segment teams.

Closing thoughts

In summary, the potential impact of generative AI in financial services is that there is extensive application to existing work in terms of automation and augmentation. This requires careful analysis and then restructuring of work to release these benefits and build better work for people. This also requires reinvestment in the skills to do uniquely human work better and skills needed to work ‘alongside’ your AI co-pilot. There are likely opportunities for ‘emergent’ work and skills, both technical skills and human skills for judgement, curation and responsible AI.

This is not a ‘amoral’ set of decisions and the mindset with which we approach this change matters greatly.?We need to embrace the opportunity and move quickly; we need to think commercially and strategically; and we need to elevate our people. Generative AI is too big an opportunity to ban it or ignore it – you have to start now. You need to analyse where it will add value and where it can open up new revenue streams. And you need to understand how AI isn’t a race to make humans obsolete, but an opportunity to enhance and elevate the things that make us truly human.

Generative AI is not truly creative like humans can be. It can generate new content that looks human-like. It cannot generate novel and non-intuitive insights or work – it tends to ‘reflect back’ answers in non-creative directions. As humans, we can truly imagine and dream, create new things and find new opportunities. We can look ahead and think about different futures, not just what they might be, but why they might be good or bad. So rather than trying to replace creative or strategic work, instead consider how generative AI can be a support for creativity and another brush for the artist. Think about how photography became an art form it itself, but was also a catalyst for impressionism and modernism – through to how David Hockney painted using photography and tricked perspective. Technology can enable greater creativity when used well.

Generative AI sounds more human, but it cannot offer true judgement or empathy. As humans, we have a conscious self and weigh potential outcomes against each other in order to make judgement-based decisions. We can empathise and think about the thoughts and emotions of other people – we can understand the consequences of our decisions and actions on others, even feeling their emotional pain in the same way we do our own. Why is this? We are social beings that build lasting relationships, cooperation and trust. We define our own identity, wanting to belong and be uniquely ourselves. We want better for ourselves to reach our own goals, but also to help others reach their goals too. Technology can enable greater empathy and humanity in work when used well.

You probably recognise and love some of these creative, empathetic and other truly human characteristics in your colleagues, friends, family and yourself. On most days you see the best of human nature in the cooperation and mutual trust that sits behind financial systems, markets and institutions – and in the bank collapses and market failures in history some of the worst. As we approach this next wave of AI, let’s use it to elevate humanity and let’s make sure it reflects the best of our human nature.

That’s where I want to go in my next blog, looking at what we can learn from past technology change and how we lead with responsibility through some of the risks and challenges with generative AI.

If you want to find out more about Accenture’s perspective on generative AI, please see our Technology Vision or our latest point of view on Generative AI.

Thanks for reading and please get in touch if you’d like to discuss further.

?

This content is provided for general information purposes and is not intended to be used in place of consultation with our professional advisors. This document may refer to marks owned by third parties. All such third-party marks are the property of their respective owners. No sponsorship, endorsement or approval of this content by the owners of such marks is intended, expressed or implied.