5 min read
Learn how delphai challenges the status quo of market intelligence with machine learning to automatically collect all relevant information.
The status quo in market intelligence looks like this: 500 analysts, manually updating spreadsheets, guestimating data points, and M&A and sales teams try to interpret these data points into actionable insights.
delphai challenges this status quo with machine learning and algorithms that automatically collect and aggregate information from a number of public sources. But how does it actually work?
In this article we want to dive deeper into artificial intelligence for market intelligence, different kinds of AI and the future of market intelligence.
The article is complemented with excerpts from an interview between delphai founder and CEO Dr. Robin Tech and Tricolore Marketing founder Gunnar Brune in Brune’s forthcoming book Artificial Intelligence Today: Applications from Business, Medicine and Science, Springer Vieweg (Brune, 2022).
So what does it mean to challenge the status quo of market intelligence through artificial intelligence?
First of all, most systems that want to revolutionize market intelligence by automatically collecting and analyzing business data are not AI-based at all. They merely scrape websites and clean them up (de-boilerplating is used to preserve only the content text and not the banners or template elements). When content is structured, this method already works well. However, since most online data is not stored in tables in a structured way, most business-related data is highly unstructured.
delphai initially acts similarly as the previously mentioned web scrapers:
Robin Tech: We collect data about organizations globally. We use what are called scraper and crawler programs to do that. This is very similar to how Google collects its data. I.e., these programs go “with big nets” through the Internet and collect data. Google is of course interested in everything, we are “only” interested in organizations. So when one of our crawlers goes to an event website, it identifies all the participating companies that attended the conference or were at the trade show. We do the same with newsletters and news articles in general.
But delphai goes further: We have developed algorithms that can automatically read and process texts to determine what a certain company does. An example would be a new product, an acquisition, or a financing round. Unstructured online data is then structured to be displayed in tabular form. These results then lead to the next stage of delphai’s AI: the classification of organizations and companies according to industries, technologies, and other topics. For that, individual neural networks have been trained to do the classification.
Robin Tech: Most of the AIs we’re talking about are “narrow AIs” that can do one thing really well. Most of the time at least, or ideally! But as soon as you give them any other task, then they are completely lost. For text analysis, we have a neural network that is specifically trained to identify products. Another neural network is specifically trained for Mergers & Acquisitions. If you applied one to the other, you would not get a meaningful answer.
We use our own specific artificial intelligence for each of these steps.
Some of our artificial intelligence has been trained through supervised learning. This works by annotating data and text to teach the AI what to look for and what not to look for.
But when companies provide only a few text data on an attribute, the unsupervised approach is used: So the AI has to learn itself what is relevant and what is not. Different models are applied unsupervised and the results are compared with what is already known. The models that then give the correct unsupervised results are kept and improved.
Robin Tech: The good thing about the unsupervised approach is that we don’t tell the AI what to look for, it looks for the anomalies itself. In fact, we can use this very AI to determine where we have a head start. A good example is natural language processing as a subcategory of artificial intelligence. This has gained tremendous momentum in recent years. We see this in the importance that chatbots have today. There is a lot of scientific attention here, of course. We compare our results with those of current research as found in preprint archives such as “arXiv”. In particular, for the unsupervised approach to identifying business attributes, we are 6.5 percent above the results of very recent preprints for the criteria of accuracy, recall, etc. So we know that we have a relevant lead here.
Customers can access firmographic data via a search engine. The interaction layer is a software, with a search bar into which they can type all the questions relevant to market intelligence and receive actionable answers. So delphai takes advantage of the full-text search we know from Google and accesses various sources of contextual data, including company websites, job ads, news, patents, and more. Based on the matches the search engine finds, delphai creates a list of companies in a convenient list format.
We enable M&A and sales departments to make efficient, data-driven decisions by transforming unstructured, global company data into structured, actionable information.
Robin Tech: At the macro level, clusters of companies are displayed, for example: Where are the fleet management software companies on the map? Or where are the quantum computing companies located? Then it shows how the companies are connected to each other. Are there clusters or are they very scattered? And looking technologically: Are there perhaps predominant industries where a technology is already more widely used today? Or are there “weak signals” because there are only three or four organizations that are already active in a niche? In this way, you can identify the big trends and the weak signals that are just starting to make themselves felt as small seedlings.
On a deeper level, each individual company profile can be viewed in detail. In the company’s profile view, you can find various vertical data, such as a proprietary, automatically generated business description of this company or automatically generated industry classifications and future technology classifications. Users can also view basic data about the company, including the year it was founded, its headquarters, the number of employees, and recent revenue.
Furthermore, peer lists can be created of companies that the delphai similarity search engine has identified as most similar to the selected company. The similarity was determined based on the companies’ full data profile, including their website, job ads, patents, and news. While this list is global, it can also be given a geographic filter.
Users can also view financials, product news, job news, and patents, all collected and displayed in a tabbed view. Lists of potential customers, competitors, and acquisition targets can be created here.
But delphai does not stop innovating there:
Robin Tech: We have also developed a module that allows you to feed in your own internal company data, which is then used as additional information in the search. So you can match your own company with other ecosystems. And you can search for specialists. Or you can match your own business units or competencies with the clusters of an ecosystem.
delphai is a horizontal product, which makes it extremely flexible – and where we have been able to build this well-modulated and micro-managed system, it has given us the foundation to easily modify and add new verticals – this is what sets us apart from our competitors. At delphai, we don’t believe that end-to-end vertical software has a future. These applications create silos that hinder cooperation and collaboration. Instead, we see the emergence of low-code/no-code-based systems that are tailored to each company’s unique needs across business units and functional areas. Examples include Microsoft Power BI-based MI/BI dashboards and teams that cross these boundaries.
For delphai, this means the following:
Robin Tech: This means that API calls are made for delphai. In this case, we define the interfaces together with the customers. For our similarity search example, there is then an API call for similar organizations and our machines only provide the data as output, but no longer the interface.
Additionally, CRM enrichment will increase. delphai offers customers the possibility to feed in internal company data. This enables the machine to establish links between existing competencies and potential new business areas.
Robin Tech: I think there’s a lot of potential there. It also disrupts the approach of management consultants like McKinsey, who get paid a million to show you where else you could do new business. Instead, you give the machine the data, and it already suggests the business areas that relate to your existing capabilities.
If you want to read the full version of the interview or purchase the book visit Springer Fachmedien.