ISSUE #67 – Can AI perform due diligence, risk-profiling, investment-focused research and the like? Raphaël Bouzy (pictured below), CEO of Datarama, shares with us how AI technology has evolved in these domains in its activities in Singapore and the region, in this innovation spotlight on Artificial Intelligence.
How has AI technology evolved for you in risk-profiling and due diligence? What are some new risk factors that you are able to account for in your systems?
Datarama leverages technology to help process the vast amounts of structured and unstructured data we encounter in our research. It has allowed us to sift through high volumes of data and information fast, and translate the information into useful analytics and business insights for our clients.
Our webcrawlers and data protocols not only trawl major proprietary databases, news sites, and business directories, but we actively seek out alternative sources of data such as ICIJ’s Offshore Leaks Database, the Panama Papers, blogs with strong predictive track records on political and economic matters, and even databases containing information on religious, club, and grassroot affiliations. Other traditional or “pen-and-paper” risk consultancies claim to provide analysis on content gleaned from these sources as well but our value proposition is that we are cheaper, better, faster, and more comprehensive – as our functions and protocols are tech-enabled and therefore automated to significant degrees.
For instance, we have a technology-enhanced tool for measuring a company’s reputation, by conducting a comprehensive multi-lingual web scrape against media sources available in the public domain. This allows us to be fair and objective in our approach of ranking companies, mitigating the risk of human subjectivity when analysing sentiment on a company, and without having to manually read all the news available on a company in order to get a sense of its reputation.
What are the risks & challenges in introducing AI into this domain of activity?
One of the biggest challenges we continue to face is that technology is not presently at a state where it can independently perform complex tasks without substantial human intervention. Some analytical tasks still require human researchers to plug the gaps, whether it is understanding the bias in social media postings, performing entity resolution, filtering out fake news, or situating issues in the appropriate cultural and political contexts.
Some analytical tasks still require human researchers to plug the gaps, and there is also much work yet to do in terms of machine learning and natural language processing.
There is also much work yet to do in terms of machine learning and natural language processing when it comes to analysing large amounts of data in less dominant or rare languages such as Burmese or, to take more extreme examples – Sri-Lankan and Nigerian dialects. This is a challenge facing all tech companies that need to analyse and vet content in numerous languages, including giants like Facebook and Google. The key challenges that Facebook faces today, in terms of identifying hateful or controversial content in countries such as Myanmar or Sri Lanka, stems from the fact that there are just not enough people working to enable AI to identify such content in local and rarer languages. We share this challenge since part of the raison d’être of Datarama is to scan for ‘red flags’ or serious issues of concern for clients, as quickly as possible, across a plethora of jurisdictions.
Working with hard-to-find information also means that a lot of the data that we deal with are not available online, or are not in a data format that can be easily processed without manual input. Use of technology will only be as effective as the access we have to data that these tools can process.
Working with hard-to-find information also means that a lot of the data that we deal with are not available online.
We also have to always keep up with changes in technology so we are kept aware of what innovations and new ideas can help make our processes better. To keep our edge, we have to think out of the box and apply technology that is not typically used in this field, or seek out alternative data sources and adapt these to risk consulting. These include watch lists maintained by small, local-level NGOs that are not commonly accessed as well as data sources that keep track of political risk events as they happen, which is not something typically featured in a static risk consulting report. This means having to always be creative and looking at alternative data sets or technology in new and novel ways.
How do you see AI technology further transform your activity?
A lot of our processes can be further automated. AI technology can further replace manual functions such as maintaining good data hygiene, quicker entity resolution within our vast database, and perhaps even write a bulk of our reports in the future (with advances of applications in Natural Language Processing/Generation). The fundamental objective is to free up the time and resources of our experts and analysts to perform higher-order analytical tasks and to seek out even more obscure but powerful sources of information. We are not satisfied with merely acquiring and understanding the data but aspire to provide our clients with the predictive analyses that they need, whether it concerns the reputational profile of a target, a political event, or a country’s evolving regulatory framework, based on a rigorous and transparent methodology.
Of course, the technology will only be as effective as the data and analysis we have to power it. Ultimately, data and analysis are still pivotal in delivering the results our clients need. We are closely monitoring the exponential progress of AI and will leverage on relevant technologies and processes, combining these with localised knowledge and a human touch to provide our clients with the superior service that they deserve.
Interview with Raphaël Bouzy, CEO, Datarama
Published in FOCUS Magazine — Issue #3 2018 “The Innovation Issue”