Artificial intelligence: the vision, the reality and the opportunities

Long exerting a powerful grip on our imagination, AI is now a reality and the opportunities are tremendous. What should companies focus on?

The vision

Nicolas Vayatis, Director of CMLA (Centre for Mathematical Studies and their Applications) at École Normale Supérieure Paris-Saclay, attributes the seemingly unlimited potential of artificial intelligence (AI) to, first, the proliferation of sensors (and hence, the data); second, the technological capabilities of High Performance Computing; and third, the ability of mathematics and computer science to better grasp complex data structures and design efficient algorithms.

In other words, the relevance of AI to businesses and its success in performing any given task depends on the capability to gather and organise information (both data and expertise; internal as well as external), the capacity to process this information efficiently, and also the ability to create real value by using it to inform our decision-making.

Essentially, AI concepts and methods have led to software that can imitate, accelerate and augment the main mechanisms underlying perception and decision-making. This has been possible because such systems are able to extrapolate data from past experiences.

According to Mr. Vayatis, the most exciting, disruptive developments in AI in the near future will be seen in the following three areas: fully predictive vs. expert augmentation; companies completely transformingthe way they use IT in their business operations; and the human factor inside and outside of the company.

Fully predictive vs. expert augmentation
According to Geometric and Functional Analysis (GAFA), a mathematical journal, the four most powerful tech companies in the world are introducing massive automation across their supply chains, from emotion expression to transaction and delivery. Mr. Vayatis thinks that this is not applicable to all situations. In his opinion, intuition, out-of-the-box thinking, and complex system analysis cannot be replaced by intelligent machines. However, it is possible to design intelligent knowledge management tools to assist experts and decision-makers in fast and targeted information retrieval, or making a comparison of the actual system and a prototype. Just think about the existing approach based on numerical simulation in industrial systems, or the use of virtual patients in digital medicine.

Companies will transform the way they use IT systems in their business operations
In the context of digital transformation, most companies will be editing software and utilising interfaces with data and AI tools. This is not surprising since refining IT is an essential task for companies that want to stay competitive. For instance, it is quite likely that, in many companies, product design and marketing departments will be sharing the same software in order to accelerate product development and mitigate risk.

Skills needed inside and outside of the company
Needless to say, the development of AI will have implications for a business’s strategy in the same way computers and office automation software did in the 1990s.

Every employee will have to know how to manage data to some extent, and also data handling and processing will be part of a well-established software development process.

However, specific skills and tools will have to be available because:
  • internally, companies will need to make sure their employees have the right skill sets and training 
  • externally, companies will want to observe and measure customer behaviour and detect global economic and social trends.

The reality today

At BNP Paribas, we try to detect and understand all the changes that impact our customers, be it in their personal or professional lives. Thanks to AI, we can use available data to create better sustainable solutions that fit the needs of our clients.

There are many fields in the banking industry AI has been applied to up to this point, for instance: translation, image or character recognition, and named-entity recognition.

We at BNP Paribas think that AI is truly a unique opportunity not just for our internal developments, but also in our ability to add value to our clients and propose more relevant solutions.

Currently, AI applications support our Corporate and Institutional business in three different ways: they help us be more efficient, enable us to perform large-scale analyses, and make us smarter in choosing between different options.

Efficiency 
AI tools help us be more efficient in performing otherwise long, difficult, and costly tasks like translation across a number of languages. Our recent AI developments have led to efficiency savings in some areas of at least 50%! And, that’s even before applying machine learning!

Large-scale analyses
Likewise, AI components enable us to perform fast, large-scale analyses – tasks that would otherwise not even be humanly possible, like payment controls or scanning news in a very comprehensive and deep manner. For example, we have developed a tool to systematically screen contracts for compliance purposes. It takes 15 seconds to screen 150 pages, and the tool makes it possible to identify the names of legal entities, people, locations, vessels, etc. This particular tool has increased efficiency and significantly enhanced the protection of our clients’ interests.

Smarter decision-making
Flaminem, a specialist in predicting rare events, deploys “data science to deliver a service solution for lead scoring, sales enablement and client retention”. It is an example of how AI can make us smarter in choosing between different options by providing a customer-facing workforce with actionable client insights quickly and efficiently. (A speech-to-text tool fitted to our data and vocabulary, for example, means we are able to link the content of a client call to web analytics, and then propose the most suitable solutions for our clients’ needs.)

Guillaume Prunier, CEO of Flaminem, explains that the firm’s Know Your Customer platform allows websites to be personalised to a much greater degree than before: by collecting behavioural data from difference sources and matching identities between these sources, Flaminem is able to create datasets, which in turn are used to learn patterns via AI algorithms. It therefore provides users invaluable insights into end customers’ behaviours and probabilities of actions based on standard web user profiles (or even on specific profiles), while fully complying with rules such as the European Union’s General Data Protection Regulation (GDPR).

These insights gleaned from data underline the vast potential of technology like Flaminem’s: deployed correctly, users are already able to understand their end-customers better than before. And using such technology is relatively straightforward – even while the data crunching process is complex.

The opportunities

It is clear that the potential of AI to transform industries and introduce new sources of growth is significant. Accenture research at the end of 2016 indicates that, by 2035, AI could double economic growth rates in 20 countries, and boost labour productivity by up to 40%. 

Since the application of AI is undoubtedly going to accelerate, here are some recommendations to companies on how to best prepare for this AI revolution.

Accept it!
Many managers can still not accept a world with AI, either because they fear it, or simply because they believe that humans are smarter. At BNP Paribas, we believe that AI is a key business asset to be more efficient in administrative tasks, and to augment but not replace human judgment. AI can help us be more useful to our clients, and protect their interests by detecting abnormalities in a more comprehensive and accurate way, or by suggesting solutions with better outcomes. However, humans are needed to be the trusted partner our clients need, especially when things are neither easy nor mainstream. Also, it takes humans to create financial innovation in partnership with our clients in order to foster responsible, sustainable growth.

Adopt it fast!
The world is accelerating! AI is everywhere. AI is not just for GAFAs! Companies in traditional industries should also take action, guided by the rigorous regulation in Europe.

Screen and standardise potential applications of AI!
There are probably dozens of cases where AI can bring enhanced security, greater efficiency or increased revenues. Get your business managers with basic AI understanding to screen them all with this one key objective in mind:  “Stop thinking specific! Think big and standardised functional needs!” In other words, aim at large-scale solutions that can be applied to most of your operational value chain. And, do it in full partnership with your IT architects and compliance teams.

Reinvent the Business–IT partnership
Data scientist teams need to work in close partnership with both the business and IT. Here are some guidelines:
          • Forget remote garages! Position data scientists in diverse project teams mixing clients, end users, business managers, and technical teams (UX, designers, developers) to co-create solutions together.
        • Favour team members who are producers (over supervisors).
        • Make sure they get trained to use agile methodologies!

        • Define new ways of working with startups and academia!
          Many big companies have developed a collaborative approach with startups since they can bring a lot of value to organisations due to their agility and ability to recruit very talented people.

          But with AI, we would suggest that you apply different ways of collaboration than you might do with digital apps. The reason being that in AI the value comes from data – not algorithms. Nowadays the most recent, state-of-the-art algorithms are often open-sourced and so an algorithm is no longer a differentiator. The value will come from the mix of internal and external data. Just as an example, by using open-source algorithms and the BNP Paribas data, two of our developers were able to create a translation bot in less than a month. It performs better, for financial documents, than the best generalist translation systems currently on the market.

          It is important to be pragmatic! Start by identifying a startup with the right technical and entrepreneurial competencies. Then, establish a close, mutually beneficial relationship between the startup and your company. If you think it’s necessary, negotiate an exclusive partnership.

          Innovation also comes from academia. To be successful, R&D projects should definitely be familiar with the latest scientific state-of-the-art thinking right from the start. Please note that there are initiatives in the scientific community for promoting reproducible research not only through open codes, but also through online experimentation platforms corresponding to Technology Readiness Levels (TRL) 3-5. Likewise, research communities and companies are currently exploring new modes of cooperation between themselves that go beyond the standard bilateral collaboration.

          Prepare each and every one in your company!
          While working on AI, you need to conduct a major transformation programme throughout your company in order to ensure it is accepted when ready.

          When rolling out your first data-enabled solutions, start with those that will first and foremost benefit:
          • the employees, alleviating challenges of their most painful administrative tasks, thus making them feel smarter, and
          • the managers, giving them enhanced control and helping them make better decisions.
          The earlier your staff start using AI, the easier it will be for you to identify the best applications of AI and implement it efficiently. Generally speaking, developing a technical solution is only a quarter of the journey; changing behaviours and fully leveraging on AI are the rest of the journey – the hard part!