AI and Machine Learning for Asset Management Firms: Part 2 – Applications for Managing Institutional Pipelines and RFPs

AI and machine learning for asset management firms sound like marketing hype—with no practical application. But before you dismiss these concepts, read on.AI Asset Management

AI and machine learning are real, relevant concepts to asset management. But traditional CRM and point solutions like Salesforce, Protrack, or Satuit cannot deliver this technology.

A technology platform for asset management with these capabilities, implemented by qualified experts who understand your industry, can deliver amazing benefits to your firm. Regardless of your industry, AI and machine learning give you new and exciting ways to improve productivity and efficiency while strengthening relationships and delivering more value.

This blog series discusses how AI and machine learning can address challenges in three scenarios:  

In Part 1, Managing Wholesaler/Retail Line of Business, we discussed how sales people operating in a vacuum don’t have insight into how their interactions with advisors, branch offices, or broker dealers are affecting sales outcomes, which means mistakes and missed opportunities—and how AI and machine learning can bring that insight into focus.

In Part 2 below, we discuss how to maximize ROI and increase AUM by focusing on the opportunities with the most potential, saving time, resources, and money.

In Part 3, we discuss how AI and machine learning are used by asset managment firms to make the cumbersome task of trip planning easier and more fruitful.


Part 2: Using AI and Machine Learning to Manage Institutional Pipelines and RFPs

A part of any asset manager’s day is evaluating opportunities. With numerous RFPs, RFIs, and general requests coming in—requiring different levels of information and varying degrees of effort—you cannot respond to every single one. How do you decide which to focus on? Most firms have the following issues when it come to managing institutional pipelines and RFPs:

  • Not knowing which opportunities to pursue
  • Inability to effectively manage sales and Operations teams (the 80/20 rule)
  • No comprehensive way to track RFPs
  • Difficulties managing complex RFP scenarios involving both sales and operations

In addition, in the institutional line of business, asset managers need to work with few powerful third-party consultants who act as gatekeepers…gatekeepers that must be worked with to win larger institutional client deals. These third-party consultants add complexity to the sales process. You not only need to understand more about the consultant—e.g., track record, ratings, etc.—but you also need to understand much more about the request to determine if it’s worthy of your firm’s time and energy.

Typically, this is more of a gut decision than anything. A tenured salesperson would review the RFP and advise whether you have a good shot at winning or not. There are some decision rules, of course…is the range of assets under management acceptable? Do your strategies align? Can your firm deliver what the RFP is asking for? The salesperson also might choose to pursue an opportunity because he or she has worked with that client before or has a good relationship with the consultant. In other words, human factors come into play. Beyond very general business criteria and an unknown list of human criteria, it comes back to at best, an educated guess based on qualitative information…and that’s no way to make a decision.  

But let’s say you do decide to pursue an opportunity and begin work on responding to the RFP. RFPs can be incredibly lengthy and complex, often requiring hundreds of man hours of work to research, verify, and complete.

Account managers and salespeople can manage RFPs and pipelines, but it’s very time-consuming and not the best use of their time and resources. And, as discussed, they—as humans—cannot process every piece of data that contributes to making these decisions, so they might not be choosing the right opportunities to pursue. Regardless of the form in which they’re presented, they all require some level of assessment, research, due diligence, and time to complete, so you want to focus your time and resources on those with the most potential for return. This is where AI and machine learning come into play.

Data is the key to success with AI and machine learning

With AI and machine learning, data is key. The first thing to do when you are presented with an opportunity—an RFP, for example—is to collect data about the following:

  • The attributes of the opportunity or the RFP
  • The attributes of the client, like AUM
  • Internal attributes – What are your strategies? Strengths? Similar client experience?
  • The best resources for that opportunity – Who is the salesperson best suited to work on that opportunity?

With AI, you get some very good news:

Good news #1: Machine learning starts with collecting data, through integration or through people entering information, and AI works with structured (data in fields, for example) and unstructured data. That means you’re not forced to do data entry. You can scan a document–an RFP, for example–and the machine will read it, look for patterns, and use those patterns as part of its decision-making algorithm. Right off the bat, you’re saving a great deal of time.

So, you get the data in various ways into your AI engine to determine whether it is a viable pursuit. In addition, you provide the AI engine with what you believe, based on experience, the overall effort to be for pursuing and winning that opportunity.

Good news #2: Based on that quantitative data, machine learning can tell you which opportunities should have the best close rate…and which to pursue based on potential return on investment. In short, machine learning is optimizing for win percentage and AUM. So, the second piece of good news is that you do not have to turn on the sales engine, which is slow and costly. The machine will review the RFP for you based on predictability, the way the RFP is phrased based on past RFPs the firm has won, and so on, and it will use all this data to make a recommendation.

Over time, as you gather more information and feed it into the AI engine, the richer the data set becomes, and the better the algorithm will become at tracking your wins or losses over time. This is where the “learning” part comes in. The AI engine “learns” as data comes in, getting smarter and smarter. It looks behind the scenes to see what’s driving a number and helps you make smarter decisions. It also helps you refine your model.

Good news #3: Machine learning empowers you to change the chance of winning a deal by showing you how you can adjust the criteria. The machine tells you that, based on your mandates, certain criteria need to be changed to increase the odds of winning. With that recommendation–backed by quantitative data–in hand, you can go back to the consultant and ask the fund to change those mandates.  You win more deals, and your relationships with your consultants get stronger because you win more of the deals sent your way and because the consultant is willing to work with you to make those deals happen. Everybody wins.

This sounds simple, and quite frankly, it is. What is not simple is how the AI engine is able to take tremendous amounts of data, learn from it, and make recommendations so quickly. This is something no human can do—and definitely not with the same accuracy and not without human bias.

Can you trust AI and machine learning?

In essence, you are being asked to trust business-critical decisions to a machine, which can be pretty scary, especially since you can’t really see what’s going on behind the curtain. So, why should you trust the “black box?” Here are a couple of objections you will often hear:

Experience and intuition work just fine, thank you. A senior salesperson will say something like, “I’ve been doing this for thirty years! I know my business. I know how to pick the right opportunities.” That might very well be the case, but can that one salesperson do that for everyone else in the company? Not without becoming a bottleneck, which is not good. And regardless of the level of experience or success rate, that person is still human and WILL make mistakes. Machine learning uses the same information objectively and consistently and is available to everyone.

We can build our own system. What if we build our own algorithm/calculation? That is certainly an option—lead scoring—is a common practice. So, you come up with a way to assign weight to different criteria and then adjust that over time. However, that model doesn’t improve upon itself, and how do you know there isn’t a better model out there or that new criteria or changes in the landscape are not falling through the cracks? Unfortunately, you do not. Machine learning will improve itself over time and will use multiple models and look at the data in different ways that no human could ever do.

But let’s say you were able to build a (nearly) air-tight algorithm. How do you keep it relevant and up to date? If you don’t, it would quickly become irrelevant. You would need a resource dedicated to the maintenance of it. That is not practical, either.

So, going back to the question of trust. If you really believe your current model works, keep it, but do an experiment—test your model against an AI model to see which is right more often. Make sure you do this over a period of time, because AI gets better and better with time, while home-grown models stay the same. As an AI model learns and self-improves with more data, the human brain will always have limited processing power—and it takes years and years of experience to master what machine learning can do in a very short amount of time.  

Let’s take a moment to talk about intuition, experience, and gut instincts. They are not useless with an AI model; in fact, they are still very relevant. An AI model can provide data to back up that intuition, to confirm your instincts—and that in and of itself is invaluable.

One minor—but manageable—“gotcha”

It’s not really fair to call this a “gotcha.” But it is imperative that you understand this: For an AI model to work, you must feed it data. Your firm likely already has good data (it doesn’t have to be a lot of data) sitting there—but someone has to input it into the system and set up that system. An AI system is smart, but it’s not smart enough to do this on its own. So, if you decide to embrace AI and machine learning, commit to doing the work on the front end. The benefits you’ll realize on the back end will make it well worth the effort.

Ready to see machine learning in action?

The purpose of technology has always been to make things easier and better. AI and machine learning are no different. You have smart people who make smart decisions, but AI can sharpen those decisions even more—and in ways a human cannot.

Machine learning isn’t science fiction; it’s real, applicable, and has the potential to take your business to the next level. Read Part 3 to learn how AI and machine learning are used by asset managment firms for trip planning

Want to see machine learning in action?  Contact AKA’s Financial Services experts.

By | 2018-10-19T16:08:55+00:00 September 28th, 2018|Machine Learning/AI, Sales & Service (CRM)|0 Comments
Alternative Text

Contributor: AKA Enterprise Solutions

AKA is comprised of professionals with deep experience in business, technology, and their respective industries. Our team members regularly share their knowledge and expertise through blog articles. We hope you find them helpful, and we welcome your comments.

Leave A Comment