Using Artificial Intelligence to Sell Mutual Funds: Three Phases

Finance is in the midst of a revolution. Processes are being digitized. Decisions are becoming increasingly data-driven and approached from the bottom up. Artificial intelligence (AI) is taking care of business while we work from home. The revolution has affected every market, firm, and department — except for product distribution.

How financial products are sold to institutional investors has remained curiously static over the past two to three decades. It’s all about relationships.

My firm, Genpact, helps capital markets participants transform their businesses with AI. We have worked with several large asset managers to reinvent how they distribute their products.

That process requires overcoming several key challenges:

• Inefficient prospect qualification due to manually compiled prospect dossiers composed of data from disparate vendors and internal sources.
• Inconsistent sales processes that make it hard to evaluate and execute channel-wide strategies.
• Forecasting challenges at the account, territory, and channel level because of difficulty matching outbound activities and their costs to fee income.

Here we will consider the first of these hurdles: target prioritization.

The Data Deluge

Today’s mutual fund and exchange-traded fund (ETF) sales teams are drowning in data. Whether it pertains to products positions; performance, flow; environmental, social, and governance (ESG) criteria; meetings and calls; macro and micro; history; estimates; and predictions — there is an information overload. Ostensibly, this data exists to help sales team identify and qualify prospects, but that’s easier said than done.

One client, a publicly traded mutual fund manager, told us that qualifying a single lead used to take staff members more than 40 minutes. Here’s what the step-by-step process looked like for the sales team:

1. Log in to the Broadridge financial solution for a list of registered investment advisor (RIA) prospects in a particular territory.
2. Collect asset flows for these prospects from MarketMetrics market intelligence software.
3. Collect additional intelligence about the prospects from the client’s RIA database financial data and marketing solution.
4. Cross-reference the data against a customer relationship management (CRM) system for information on and results from past meetings.
5. Assemble all the data in Microsoft Excel.
6. Rank the opportunities based on formulas and judgment.
7. Review the top priority prospect’s investment philosophy through its website to determine the optimal pitch.
8. Schedule a call or set up a meeting.

With our help, the client has transformed this process in three phases:

Phase 1: Self-Service

The client set up data feeds with all of its vendors, aggregated its internal and third-party data into a data lake, and packaged them for use through user interfaces. These included a self-service interface for salespeople and a more advanced one for the business intelligence (BI) team.

---

---

Before the transformation, the sales team had to drill down one client at a time. Now the self-service interface helps the team conduct analysis across multiple clients. This has opened the door to a number of fruitful data-driven conversations. For example, the products the firm had previously prioritized for the sales team turned out to be neither the best-performing nor the most sellable.

Phase 2: Report Library

Once the sales and marketing data was centralized and integrated, the client developed a library of reports to drill down into the data. The goal was not to reproduce existing reports, summarize the pipeline, or describe “how things are going.” Rather the purpose was to drive decisions about who to call and what to pitch.

The combination of domain knowledge of the industry and business with digital technologies proved critical here. The client’s best and most senior salespeople had the experience, skill, and intuition to identify patterns of likely buyers and likely pitches. The report library codified this knowledge and made it available to the entire team.

Here are two pattern examples:

• RIA 1 bought mid-cap mutual funds in 2012 and 2013, after mid-cap funds outperformed. If the client’s mid-caps begin outperforming again, it’s a good time to call RIA 1.
• RIA 2 consistently turns over their portfolio in January. So, the client knows to call them in October, not in February.

Based on patterns like these, the client identified high-probability targets and reported on them weekly by email and through the CRM. The reports were succinct and action-oriented, as in “Call firm X and pitch product Y.” Smart routing ensured the right prospect ended up with the right salesperson and team leader.

Phase 3: Machine learning

While the report library packaged and automated human pattern identification, the next step was to add machine-based pattern identification. To train the model, we had to clearly define what the sales team would look for in terms of positive outcomes from the available data. There were some nuances. The right definition depended on the team and sales process.

Here are two examples from the same client organization:

Sales Team	Sales Process	Positive Outcome
Selling to RIAs	One-on-one meetings	Inflow of $1 million or more within 60 days of pitch meeting
Selling to Retirement Plan Administrators	Request for proposal (RFP)	Making it to the shortlist stage

Once the outcome was well defined, we used historical data to train a model to predict which pitches would most likely succeed. Now, the machine learning model acts like a senior salesperson, identifying demand patterns and ranking them from best to worst.

Introducing machine learning also created bonus capabilities, including the ability to:

• Assess which patterns, both human and machine, are the most effective at generating positive outcomes.
• Rank targets across patterns, by expected value (positive outcome x inflow), creating a single call list for the sales team to act on.
• Identify new patterns on the fly, as the model is retrained on fresh data. This captures structural changes in buying behavior, for example, due to COVID-19.

Conclusion

By segmenting the transformation into three phases, we helped our client manage the change and improved the odds of success.

• Self Service helped build trust in the data among business stakeholders by getting everyone on the same page. It also delivered some quick profit-and-loss wins.
• Report Library allowed sales leadership to standardize the sales process, moving from efficiency to effectiveness. It also introduced the capabilities to the broader team in an easy-to-digest format.
• Machine Learning capabilities were introduced transparently, without changing the format of the reports the sales team consumed. This builds further credibility and enables an augmented intelligence operating model: AI supporting human judgment.

Our client has reduced the time it takes to qualify a prospect from 40 minutes to near instant. Looking ahead, outbound activity is expected to grow by 30% and inflows by 10% to 15% within a year.

To be sure, this is just one example, but it demonstrates how AI, when properly harnessed and guided by human judgment, can create more efficient and effective processes. It also is a case study in how the firms and professionals that successfully embrace and adapt to today’s data and technology revolution can lead the finance sector in the years to come.

For more on alternative assets, register for the complimentary webinar “The Global Investment Performance Standards (GIPS®): Application to Alternative Strategies,” hosted by the Standards Board for Alternative Investments (SBAI) on 15 July 2020.

If you liked this post, don’t forget to subscribe to the Enterprising Investor.

---

All posts are the opinion of the author. As such, they should not be construed as investment advice, nor do the opinions expressed necessarily reflect the views of CFA Institute or the author’s employer.

Image credit: ©Getty Images / MF3d