IA Fintech Member Insights: Limeglass
Artificial Intelligence in the business world – where are we now?
Everyone’s talking about the potential for artificial intelligence in the enterprise, but what’s actually been achieved to date? Are the majority of companies comfortable with the technology and how it can be used, and are many already using it? Or, are the majority of organisations still unclear as to exactly what it is and what it offers to the business? DW asked a range of experts for their thoughts, and there was no shortage of responses, and plenty of food for thought. Part 5.
AI-driven credit scoring – the benefits for SMEs and emerging market companies
The traditional credit scoring process is outdated, and remains reliant on a small number of accounting entries. Tradeteq’s Michael Boguslavsky believes technology is the key to achieving greater transparency and rigour in the credit scoring process, minimising the risks associated with global trade flows and most importantly, opening up access to trade finance for SMEs and emerging market companies:
Artificial intelligence (AI) is a very popular buzzword these days. Generally, it refers to the use of computers and computer aided systems to help people make decisions, or make decisions for them. It usually relies on large volumes of data or sophisticated models to help understand the best ways to make sense of all the information and draw intelligence.
In trade finance, AI is particularly helpful in analysing quantitative data; there are usually a large number of repetitive small transactions. The repetitive nature of trade finance means that there is a lot of non-traditional data at our disposal.
This means AI-driven models can be very efficient for data analysis and revealing intelligence relating to small companies that traditional trade analysis tools cannot cover.
AI-driven supply chain analysis for SMEs
Trade finance as a business is only beginning to be digitalised and inefficiencies therefore remain, but that is changing very rapidly. A group of institutions, including banks, pension funds, institutional investors, funders and trade associations, came together earlier this year to form the Trade Finance Distribution Initiative (TFD Initiative).
An important component of the Initiative is bringing new levels of transparency to a market that is opaque and paper-based. This allows institutions to identify attractive financing opportunities where it previously may not have been possible.
Beneficiaries of this include micro, small and medium-sized enterprises (MSMEs) and corporations based in emerging markets, where credit providers simply don’t have enough data, accounting or trade information to make sound decisions.
Tradeteq’s AI technology creates credit scoring models to analyse the history of a company and their transactions. It uses vast amounts of public and non-public data, including data on each company in the supply chain as well as each receivable. From this, we can create a sophisticated evidence-based, credit scoring model.
Tradeteq also allows a company to receive early warning signs when a supplier or counterparty is in distress or at risk of not fulfilling credit or trade requirements. Tradeteq’s algorithms predict the effect on each business to ensure the risk of interrupted trade flow is minimised.
For larger international banks, this can create more trust between them and smaller corporations, including those further down the supply chain where trade finance banks typically cannot provide coverage.
Overcoming the data challenges
Even though the process seems simple, substantial challenges remain due to the availability and reliability of the data. The traditional approach to credit scoring relies heavily on company accounts and the data within them, which can be out-of-date. This is a major barrier and prevents many small companies from accessing trade finance.
A lot of financial organisations are gradually collecting their historical data and merging it with current operations to create a single source. I don’t think that this has been very successful so far, but we welcome the strong effort to put data into a reasonably uniform format.
The second challenge is the legal aspect of accessing the data once it is available and retrievable. When looking at the cross-jurisdictional legal issues, there are still things that need to be resolved in order to ensure compliance with all local and international laws.
The third challenge, which probably the easiest to resolve, is modelling accurately when the data has been validated. This is where AI can become particularly useful.
Closing the trade finance gap for MSMEs
Once the data is available in its entirety, counterparties or funders can use AI to observe patterns in a small company’s trade and payment history, including non-payments, then look at the history of comparable companies in the supply chain to identify and assess undue risks.
As it stands, this cannot be done efficiently, at speed or at scale. AI can help funders make more confident trade predictions for companies, opening up trade finance access for many small companies who would have otherwise not had that access and reduce the trade finance gap.
How is artificial intelligence driving innovation in financial research? asks Rowland Park(right) and Simon Gregory(left), Co-Founders of Limeglass:
Despite innovation in many sectors of finance, the development of financial research has lagged far behind other areas. This seems counterintuitive when we consider that research forms the bedrock of any corporate or trading decision-making and has a huge impact on a company’s performance and profitability.
Yet many market participants rely on outdated processes such as scrolling through an email inbox or using ‘Control+F’ in documents to try and find relevant paragraphs. The inefficiency of these methods is exacerbated by the sheer volume of reports that financial institutions create and receive every day. The result is information overload, with key pieces of information missed, and ultimately, potential loss of opportunity in the market.
How can AI maximise financial research?
In order to realise the value in research assets, research producers must have a way of zeroing in on the relevant details in the library of documents. In the traditional provision of text documents, this is not possible, and therefore the reports are of limited value.
Artificial intelligence (AI) offers a solution for the publishers of financial research, and consequently for their clients who read and use the reports.
An AI platform can scan documents quickly, tagging words and phrases, and identify paragraphs for cross-referencing. Rich Natural Language Processing complements this by analysing text to identify where synonyms are being used and can draw connections between associated topics or phrases. This can then be organised into an asset specific taxonomy, enabling users to very quickly access the right level of detail on any given topic in their entire library of research.
For the best results, the technology should be expanded upon with human expertise. When it comes to identifying what subjects and terms are linked together, an algorithm can produce a statistical answer, but it takes human expertise to be able to add the additional layer of subtlety which makes a technology like this as incisive as possible.
This means human experts should be involved in initially defining the taxonomy of tagged terms, while also having regular input to grow and develop the taxonomy as additional market issues emerge. New topics are appearing constantly with a raft of phrases and associated synonyms. For example, the general term for a viral infection, ‘Coronavirus’, has been given a specific name ‘COVID-19’, for the global outbreak which began in 2019 and this in turn has many variants as to how it is written.
In any scenario, there may be multiple synonyms that need to be linked in a taxonomy for which a research team provides invaluable knowledge.
Personalisation through AI
For financial research creators, the customisation of reports by surfacing key paragraphs and providing relevant information will lead to a better service to clients. The ability to break down a report into its component parts, via AI, enables the analyst to provide not only specific information that the reader is interested in, but also other paragraphs with associated information.
This personalises research with both a macro view of the topic as well as a granular level of detail of specific issues. Moreover, such personalisation means that the recipient no longer has to wade through a mass of long documents to find what they want or risk missing out on key pieces of information.
The oil and gas industry is in the middle of a digital revolution, says Tim Bisley, Managing Director, Digital Products at Lloyd’s Register:
Following the recent resurgence of the market and the development of advanced AI solutions, organisations are faced with a unique opportunity to reimagine the way they manage their physical assets.
However, despite this revolution, it seems that many major oil and gas companies are not making the most out of these new advances in AI, nor the multiple sources of data available to them across their operations.
Although organisations often collect vast amounts of data, it’s clear that in the oil and gas industry at least, many remain challenged to be able to use it. To put it into context, one industry study found that less than 1% of the data being generated by 30,000 sensors on an offshore oil rig was actually used for decision making.
Our own research revealed that 71% of oil and gas asset managers still rely on just a single source of data to analyse their asset performance and risk management. This is a missed opportunity to maximise production, reduce waste, eliminate unnecessary downtime, and reduce the risk of a containment or safety incident.
The way companies with big infrastructure (e.g. chemicals, oil and gas, power, transport) typically inspect their assets to avoid failures and breakdowns has, until recently, followed a time-based approach. However, thanks to AI, 3D digital twins and machine learning, actionable insights can now be derived from multiple sources of data with speed.
Using AI in this way has been proven to lead to cost savings of 10% to 40%, yet only 18% of companies have adopted this approach according to those we surveyed. The slow pace of adoption, in spite of the efficiencies these innovations bring, sits somewhat at odds with the commonly held view that industry needs to pursue operational excellence to maintain growth.
If organisations can begin to draw on multiple sources of data, they’ll be able to derive previously unprecedented levels of insight from their assets. In time this would allow businesses to reinvigorate their maintenance approach, improving productivity and uptime.
While the majority of organisations are yet to transform their asset performance management systems, interest in emerging technologies is growing rapidly. Our research revealed the greatest interest is in big data (25%) and AI and machine learning (23%), while many are also excited by the possibility of 3D digital twins (19%).
Anecdotally, however, many organisations are reluctant to adopt new systems due to the required training and skill sets they can demand. Yet, there are now platforms available, including Lloyd Register’s AllAssets, that have been built with ease of implementation in mind, allowing oil and gas companies to rapidly modernise their approach to managing physical assets.
Advanced technology solutions are giving organisations a unique opportunity to reimagine their asset environment. It’s our view that it is time for organisations to move on from their legacy systems and enjoy a new age of operational success.
The gap between the realisation of value from deployments of AI based technologies and the hype around what today’s AI’s can achieve, is still very discernible across both the public and private sectors, according to Edward Charvet, Chief Strategy Officer, Logicalis:
At Logicalis we wanted to understand the extent to which AI is being used by businesses across the globe, how significant this value gap currently is and how fast it is closing. Our 2019 Global CIO study questioned over 800 CIOs from around the world on the extent to which they see potential in this technology, if they are embracing it today and if so, are they realising benefits as a result.
According to our survey, 41% of CIOs have deployed an AI based technology solution in their business. This number has doubled since our survey just one year ago. This supports the wider market commentary around AI deployment and underwrites the inroads that AI based technologies are making in aligning to business needs within the enterprise. The survey goes on to reinforce that this trend will continue, as 60% of respondents believe AI is going to be one of the technologies to have a significant impact on their businesses over the next two years.
And yet today the value gap is clear, as the results of our survey show. Innovative technology leaders remain vocal about the potential of AI. Many embrace the importance of investing ahead of the value realisation curve, to ensure the maximisation of the efficient advantages on offer. Yet, it would be naïve to ignore that a high proportion of respondent’s are struggling to realise the business benefits that they were promised. In our survey almost half (47%) of respondents who have deployed an AI based technology say they are still waiting to see significant value across different business departments; this is material.
Students of the history of AI development will recognise the value gap as a precursor to the AI winters of the past. However, Logicalis believes that some key metrics that sit around the current vocalisation of the value gap, indicate that the likelihood of an AI fall from grace is small.
The first significant point is that where AI deployments are working, they are working well. Many forget that todays AI’s are narrowly defined, goal centric applications and, by definition, if the context of the requirement is not extremely well defined, then the potential to deliver value will be limited. Logicalis has helped clients realise significant value within the pattern rich data environments, such as in security managed services, through the judicial application of AI based technology solutions supporting an improved understanding of threat behaviours.
Secondly the AI’s of today are carving out a reputation for success within the field of optimisation, where the complexity within business need is so significant, only an elegantly defined, goal centric AI, can offer any form of solution. Multisite distribution or collection services, such as waste bin collection or home delivery, is not only a reality today as a result of AI, but the efficiency of the solution is improving all the time as the underlying neural networks evolve in response to the growing volume of data inputs.
Combine these realities with the continued progress made around compute capacity and data processing speeds, and every business embracing AI today can have both the confidence of knowing that value extraction at a material level is achievable and the base case for deployment, the context and deployment rationale, is rapidly becoming clearer. This is allowing businesses to specify needs that can be addressed by AI with ever more assurance.
The survey results infer that today organisations aren’t yet aligning a tight requirements’ specification to the practical constraints of today’s AI based technologies, so that the business benefits show themselves quickly. This is the true gap that underpins the perception of ‘lack of value’. As Architects of Change TM Logicalis sees its role as the trusted business and technology advisory partner to its clients, to ensure this foundation understanding is in place before our clients move forward with an AI-based technology solution.
AI is positively impacting cloud management and cloud services, says Ramprakash Ramamoorthy, product manager at ManageEngine:
As cloud computing continues to grow, cloud management systems keep on generating huge amounts of data. To provide the best possible customer experience and to manage cloud costs effectively, in-depth analysis of this data needs to take place. Yet, clear insights cannot be drawn from these huge volumes of data using traditional rule-based systems. This is where AI is being utilised to analyse these enormous data sets in order to help enterprises get valuable and in-depth insights about their systems.
Used in this way, AI facilitates better alerting, stronger monitoring of availability and greater identification of the root cause of failure events. AI systems can predict outages, help provide proactive infrastructure management, and ensure better service availability.
Another area where AI is having an impact is in IT service desks. Unlike cloud management where the bulk of data is machine generated, most service desk data is generated by people. The manual processing of this data can slow down productivity, especially when it comes to responding to service desk tickets. Enterprises are benefitting from a powerful form of AI called natural language processing (NLP). When used in chatbots, NLP is boosting service desk productivity and response times.
NLP has the ability to recognise and understand the context of each service desk ticket as they are submitted. It can then automatically assign the ticket to the most suitable technician for resolving the ticket based on similar interactions that have taken place. This not only accelerates ticket response and resolution times but makes it easier for service desk agents who can focus their energies on other tasks.
As AI becomes increasingly important across different aspects of managing IT infrastructure, enterprises must work to adapt to the natural changes that occur when implementing any new technology. While AI is enabling decision automation across the stack, it is important that businesses ensure their existing process hierarchy also factors in the decisions reached from the AI system.
Decision-making has traditionally been a question of yes or no for enterprises. Yet AI has introduced an element of probability to decision-making processes. Today, many enterprise monitoring systems issue alerts when there is trouble with a particular server but an AI-powered monitoring system will advise that there will be a 60 percent chance a particular service will fail in the next hour. The impact of these advisories can be factored into the organisation’s IT workflows.
Ensuring successful adoption of AI across the enterprise comes down to designing and modifying existing hierarchies to make them flexible enough to accommodate probabilistic decisions across various workflows in the organisation. Enterprises adopting AI will need to be ready to quickly adjust existing workflows in order to reap its benefits.