Financial services is a spearhead for ambitious AI adoption

The Nordic financial services industry has long been at the forefront of digitalisation and technology adoption. PA Consulting conducted a survey on AI in financial services on behalf of the Finance Sector Union of Norway (Finansforbundet). The survey finds the Norwegian financial services sector leads the way: its use of AI is high (85 percent) compared to numbers for the financial industry in Europe (40 percent according to EU’s European enterprise survey on the use of technologies based on AI). Many financial services organisations in the Nordics have already started with more advanced forms of AI and have increased their ambitions in the past three to five years.

PA’s study reveals an up-to-date picture of AI applications, capabilities, barriers to use, and views on the responsible future use of AI within the Nordics. Amongst other findings, it revealed 89 percent of survey respondents expect to increase the use of AI in the coming years – so in which areas are we likely to see this happen?

Increasing digital touchpoints to fuel AI decision making

Organisations with General Insurance offerings to retail customers have the highest AI adoption with 86 percent of firms using some form of AI. The industry has a long history of applying advanced analytics in the pricing of a relatively wide range of insurance products. This seems to have led to an increased maturity and support for adopting new technology such as AI.

Life and Pension (L&P) insurance, meanwhile, has an AI adoption rate of 67 percent. The gap between General Insurance (86 percent) and L&P insurance (67 percent) is likely explained by L&P’s traditional focus on the corporate market which constitutes fewer customers than retail. Large parts of the corporate market are characterised by more direct personal customer contact, leading to fewer digital touchpoints and less data to fuel AI recommendations and decision-making. Looking ahead, many organisations will therefore be considering how they can catch up their corporate markets to more closely align to the general market.

Creating bias-free and responsible AI

The vast majority of Norwegian banks use AI in some form ranging from anti-money laundering models, language models in chatbots, and image recognition for customer identification. However, banks consider AI ‘explainability’ a major challenge. For example, how do you explain an automated AI decision to a customer who’s rejected for a mortgage in a digital self-service application? And how do you ensure that AI models and decisions are bias-free so as not to discriminate due to personal characteristics and attributes?

In addition, societal trust remains a barrier to greater adoption. Banks and insurance companies are dependent on trust to succeed with everything they do. In this regard, the use of more complex AI models (deep learning and neural networks) that mimic how the human brain works bring new challenges and necessitate new considerations. Consequently, to maintain trust among stakeholders, financial services firms in our survey are already embracing a responsible AI approach – either utilising responsible AI frameworks from the EU or Organisation for Economic Co-operation and Development (OECD), or developing their own internal frameworks.

Greater AI sourcing options

opens up competitive advantage The readiness and adoption of off-the-shelf standardised solutions with AI capabilities has increased. Standard off-the-shelf solutions enable financial services firms to not only reduce the time it takes to develop and launch AI initiatives, but also to address existing AI competence and capacity gaps by using modern, best-of-breed platforms and functionality from external vendors.

Indeed, some of the financial services firms who first experimented with in-house AI development are now replacing in-house solutions with standard solutions. As one survey respondent said: “We have phased out in-house solutions for standard solutions. It is challenging to make better AI than what’s being delivered by dedicated technology companies.”

Rapid improvements in development tools and methods are reducing the barrier to in-house development too, however. This helps increase the pace of in-house solution development in the areas where unique custom solutions can create a competitive advantage. Many firms surveyed also complement their own AI workforce with external capacity and competence from consultancy firms.

Historically, data analysts and model developers have enjoyed significant freedoms when choosing analysis and development tools. Our study shows that firms are standardising the sourcing and procurement of AI tools. This means it has also become more important to understand the potential of open-source libraries and third-party tools than to build models from scratch.

Most of the firms surveyed currently run the relevant infrastructure for AI on-premise but aim to transition to cloud-based solutions. This is driven by the need for more scalable data processing, improving development tools, and reducing the investment needs of infrastructure to secure appropriate data quality and access. AI-as-a-Service offerings may lower the barrier to entry for financial services firms even further and level out the playing field – especially for those who are currently on the sidelines.

AI talent pools and embedding AI

As the role of global technology companies and standardised AI platforms grows, some organisations expect integration between competence and capability to be key. Finding candidates with both quantitative and technical skills creates the most challenges when filling AI roles. Additionally, many financial services firms find AI operations require different competences to traditional IT operations.

Beyond the technical roles, financial services firms need to enhance their AI competence within existing business roles to fully embed AI. Some firms chose to centralise AI development to ensure sufficient capacity and scale. However, this risks creating a ’distance’ between the AI team from the recipient organisations, business requirements, and the end-users. Consequently, these organisations are exploring alternative models for collaboration.

We also observed decentralised models where AI resources work closely with respective business teams. For this to succeed, a supportive architecture and collaboration tools are among important success criteria.

Hybrid models feature regularly within financial services alliances with not only centralised AI development but also decentralised resources close to the businesses. This resembles the centre of excellence model chosen by larger banks: a hub that stays updated on the latest technology and best practice, and also supports innovation and development in the business lines and product units.

Our survey found that successfully embedding AI requires sharing success stories and transparency of results to increase trust and buy-in. Organisations who can demonstrate increased sales by using AI will generate more interest and trust in AI as a complementary tool for employees. Similarly, organisations who experience cost benefits, increased simplification, or improvements in customer experience from AI show greater acceptance and tolerance for other opportunities enabled by the technology.

What does the future look like?

Firms who are using AI are already finding pockets of value today, and it is clear that Norwegian banks and insurance companies will invest more into AI going forward.

Today, financial services firms’ use of AI is mostly connected to core business processes and value chains. The firms surveyed also expect to use AI in HR functions for topics such as competence gap identification and individual development plans. Firms are also exploring ways to use AI to comply with regulations which are costly to comply with, for example counter-terrorist financing and fraud prevention. Moreover, AI in automated reporting can both reduce cost of compliance and reduce the risk of non-compliance fines. Technical solutions which facilitate sharing and training of sensitive data across entities is an interesting area being explored. Financial services firms have sensitive data that cannot be shared across entities, but training on models on different banks’ datasets can help uncover complex money-laundering schemes across banks even if the datasets aren’t shared.

Firms who successfully scale AI value creation will need to have a pragmatic approach to investments, operating models, and technology choices. As we’ve identified, more mature solutions are needed before firms want to fully invest. For example, sentiment analysis and image recognition technology need to be improved before the surveyed firms consider the benefits great enough in all use-cases. Nevertheless, pragmatic firms already deploy image recognition technology in confined use-cases such as bill scanning and customer verification.

Successful firms will also need to have an adaptable operating model which caters to the right combination of buying, building, and partnering as the most obvious choice today may not be the correct choice in a few years. Similarly, we expect successful firms to continuously optimise AI roles and internal organisation to ensure they combine good business understanding with technology expertise. The requirements for competence will consequently change, too.