This post was originally published by Raj Shroff at Medium [AI]
At the height of the pandemic, a CNN report explained how two tech firms, X-Mode and Tectonix, tracked phone location data of spring break visitors to a Florida beach in March 2020. The firms were able to track where these phones (and their owners) went after leaving the beach. A location map showed where these people ended up across the United States.
When the world opens up, anonymized geolocation data can track human activity in shopping districts, tourist areas, and economic hubs. This data can serve as an early signal of economic activity before official data is released. These insights can be used to make investment bets on the hospitality industry, for example. Machine learning techniques can predict where crowds will move based on past movement patterns.
Automating Investment Analysis
In the good old days (a few short years ago), analysts would spend countless hours poring over annual reports, industry news, and earnings calls to understand how a company was doing.
Nowadays, a branch of AI called Natural Language Processing (NLP) is capable of ‘reading’ these reports, articles, and call transcripts. These tools can then extract insights from annual reports and summarize key findings. Sentiment analysis tools can analyze earnings call transcripts and determine the extent to which management feels positive or negative about the company’s prospects. AI startups such as Alpha Sense provide these tools to institutional investors.
Similar NLP tools can also be applied to news and social media data, processing massive volumes of data that human analysts can’t hope to match. The good news for analysts is that they are now free to focus on more value added, alpha generating analysis.
Tailored Client Reports & On-Demand Information
Natural Language Generation (NLG), a technique related to NLP, can automatically generate text-based content from underlying data. Investment managers can use this technique to automate periodic client reports and even serve clients market insights on-demand.
Bloomberg has been using automated reporting to write up to one-third of its news stories, according to a 2019 New York Times report. Bloomberg is not alone — the NYT report points out that hedge funds also use automated reporting to serve their clients market info.
Investment managers can automated reporting to cut costs and save time internally. More importantly, providing timely reporting and value added insights to clients will improve client satisfaction and the firm’s reputation.
Investment management firms stand to gain considerably from AI adoption. In order to maximize benefits from AI, firms must consider the following:
- AI Use Case Identification: What are the most rewarding business opportunities that we can solve with AI? Which AI techniques can we use?
- AI Prioritization: How should prioritize our AI projects, by use case and time horizon?
- Acquiring Data: What types of data do we need? Where can we get it?
- AI Vendor Partnerships: How can we partner with AI startups to co-create AI solutions that are designed for our unique needs?
Use Case Identification
Identifying AI use cases comes from gaining AI awareness. Investment management senior executives will understand their industry and their client’s needs, but may need an introduction to what AI is capable of. This foundational AI awareness will help executives conceptualize how AI will help the firm and its clients.
AI awareness & education can come from online courses, corporate training, AI consultants, or AI vendors that have a close relationship with the firm.
It is especially effective when a trusted consultant or AI vendor comes in and works with employees to identify use cases. By working with the firm’s management and domain experts, AI vendors and consultants can help identify AI use cases for the firm’s most rewarding business opportunities, whether it is risk management, alternative data insights, or automated analysis.
Another benefit to leveraging AI vendors and consultants for use case identification is that they can explain (and maybe provide) the AI methods and tools required to deploy a solution.
Companies can’t roll out multiple AI solutions at once due to time and resource constraints.
A good rule of thumb is to knock out ‘quick win’ projects first — smaller projects that will quickly have a measurable impact. For instance, an investment management firm can acquire and deploy an intelligent Robotic Process Automation (RPA) tool that uses AI to automate a routine administrative workflow.
After gaining momentum and progressing on the AI learning curve, firms can shift to the higher-value AI use cases they identified. It makes sense to split an AI initiative into short/medium/long term projects. That way, costs are controlled, benefits are stacked, and the AI initiative is continuously validated over time.
Much of the data for an investment management firm’s AI projects will be internal data. Still, there is the question of unifying data sources and data cleaning — a non-trivial activity to say the least. Other data will come from financial data vendors such as Bloomberg and Reuters.
Alternative data, such as satellite imagery and anonymized geolocation data mentioned above, comes from specialized alternative data vendors. When vetting these vendors, firms must:
- Ensure that the vendor does not use material non-public information. This could put investment managers at risk of insider trading accusations.
- Ensure that you can easily integrate vendor datasets into your AI models
- Check whether the vendor data is tagged using machine learning only, or if humans perform a secondary check to improve tagging accuracy
- Consider whether the data vendor will still be in business next year (competition is tough)!
AI Vendor Partnerships
Building AI is not easy. Only the biggest investment managers can afford to build everything in-house using an internal AI team. While firms should aim for this in the long run (greater knowledge effects, data security, IP protection), many firms still have to get to this level of AI maturity.
At the same time, just going out any buying a bunch of AI tools won’t work in the long run. Off-the-shelf AI products may not be tailored to your business needs and may not integrate well with your data.
A more robust strategy is to partner with trusted AI startups and vendors to co-create AI solutions. Vendors can work with your employees (end users) and your technical & data team to develop a suite of AI tools that work well together, both with each other and with the company’s data.
This strategic partnership helps ensure that AI solutions are built to last. A vendor that is a partner will understand your business objectives over the long term and can upgrade your AI tools as objectives change.
Finding high-quality AI startups and vendors to partner with is not easy. A prospective startup or vendor should have:
- A diverse team of machine learning developers, product managers, software engineers, data specialists, and business specialists
- Deep knowledge about your industry
- A track record of deploying AI tools across functions and value chains
Deploying AI at scale — so that it improves all major corporate functions — can lead to sustained competitive advantage and investment out-performance.
Scale matters. Companies don’t invest in AI to look cool. They invest in AI to improve business outcomes and solve problems. These problems, such as investment risk management, are large scale. Therefore, AI solutions must scale accordingly.
Partnerships will help companies scale their AI initiatives. Partnering with competent AI startups and vendors will help shorten the learning and development curve and result in faster implementation.
In the end, each firm’s AI journey is unique. Identify high-value use cases, lay out a plan, pick your partners, and build something that lasts.
This post was originally published by Raj Shroff at Medium [AI]