No AI Was Used to Write This

Generative AI (AI) is so good at so many things that it often feels like it will replace everything built from knowledge and bits. And maybe someday it will. Like much of the world, at QuantiFacts we are working with AI to increase productivity and enhance our applications. But we also think it's important to recognize what AI still can't do well. What is the non-AI core about us that gives us the ability to compete and continue to provide unmatched value for our clients?

We all know that AI can help us with daily tasks like writing, coding, finding trends in datasets, finding and organizing information, and AI agents can automate and replace routine functions and even work in teams to accomplish complex assignments autonomously. I can't remember the last time I didn't use it to at least polish a draft, but this time I resisted the urge to run this article through the LLM machine. In truth, it was somewhat painful to deliberately skip using ChatGPT just for the gimmick of creating something without the help of AI, and the result is an article that took longer to write and is undoubtedly less organized and concise. How quickly I have become spoiled. But the point of not using AI to help with this article is to emphasize that the core value we provide does not come from AI.

Knowing that in general, firms that provide software and consulting to support large organizations are at risk from AI developed in-house replacements, why are we at QuantiFacts still confident in our future? The answer has to do with what AI cannot or should not be used for, and to explain this I need to first provide some background on our company and the problems we solve.

QuantiFacts has been helping the country's largest banks with Wealth Management pricing and fee implementations for decades. For these institutions, asset-based fees are complicated. They are based on tiered fee structures, with various accounts and sub accounts in fee combinations, layers of fees, numerous legacy and custom fee schedules, and many forms of discounts and asset level exception pricing. With all of this complexity, and tens of thousands of accounts, management needs tools to have visibility into their current fees to identify "leakage", and for precise modeling and guidance when it is time to make and implement price changes. QuantiFacts has proven applications specifically for analyzing, managing, and modeling fees at the macro and micro levels. And through decades working with the largest banks on pricing decisions and implementations, we have deep experience supporting clients on optimum fee schedule selection and effective implementation across their footprint.

With data from thousands of accounts, it would seem AI would be a useful tool to analyze current pricing based on recent fee transactions, for example, and estimate the impact of changes. But any estimate, no matter how good, will be less accurate than a precise calculation of what will really happen, which is what the QuantiFacts applications provide. Just like the bank's fee taking system, the QuantiFacts applications use the fee schedules, combinations, discounts, cusip-level exceptions, and all of the existing complexities to calculate fees and model changes to any of the underlying pieces. When the core of the model already calculates the precise changes, the only other answers are less accurate.

Another challenge for AI in modeling fee changes in the real world is that the required expertise has been developed over many years, and is not available for or translatable to machine learning. Large banks typically change prices every 4 to 8 years, so there is not a trove of price change history available for AI to "learn" (be trained) from within the bank, and none available from their competitors. Having worked with top-15 banks on price changes for years, QuantiFacts has "learned" from many different bank-wide fee changes. Additionally, when QuantiFacts works with a client on a fee project, we are exposed to all aspects, from the human experience with management, relationship officers, fee operations, to the hard data results from actual fees changed, phase-ups over time, and attrition.

Human experience is also invaluable when "art and science" is required to meet soft objectives, such as fee schedule appearance or product differentiation, at the same time as hard objectives, such as revenue goals. When we meet with management at the beginning of a fee project, the first thing we establish are management's goals. Is there a revenue target, or a need to match fees with costs? Is there a goal to simplify fees? Is there a desire to unbundle fees and charge different rates on investments that have higher internal costs? Does the pricing structure need to change to reflect a sales goal of selling larger relationship sizes? Does the headline rate need to be lower than a competitor's? Are there compliance issues combining IRAs with other account types? There are usually a number of goals in any price change, and some of those goals conflict and need compromises. And fee schedule appearance matters; the schedule should be marketable and not look like the rates and steps were determined by a computer.

Another challenge with using AI in a fee project, for example, is the need to overlay real world behavior onto a model of precise calculations. This is similar to the competition we face from an in-house team who try to provide the analysis and support for a fee project using something they built to calculate the new fees. An inexperienced analyst with a fee calculator is going to miss key issues, as would an AI not trained in real world fee implementations. A mathematical calculation of proposed fee schedules is useless without a realistic model of what will happen in the implementation after the give and take between management, relationship managers, and customers.

There is also an outsized risk in using AI for something that is both infrequent and has a huge impact to your core business, such as a fee schedule change initiative for a large Wealth Management institution. AI today really needs to have an environment to learn and make mistakes so that it can be tuned and verified. The challenge to using AI for in-house fee modeling and implementations is that a mistake is so costly that the risk is too great to undertake the project without experienced oversight and proven tools. If the fee schedule design was built on a flawed model or one not tuned for real-world use, will you have promised management millions of dollars in lift that won't happen? Will you have created an implementation nightmare with a schedule that doesn't "fit" your account base? If a price change impacts 50,000 accounts, do you want to risk a botched roll-out that could either cause embarrassment, unnecessary attrition, fee decreases, or a discounting nightmare that causes you to lose all control of standardized pricing? As a manager in charge of the fee project, are you willing to take those risks?

At QuantiFacts, we know AI today can't beat us in value, but we do face risk from the perception that it can. Getting prospects to understand that, for years, they have been flying blind without the right fee analysis and modeling tools has always been a challenge. We have seen poor pricing decisions and ineffective fee roll-outs when management opted to do it alone or with a consultant without specialized experience and tools. Competing against the "AI can do it" perception adds another challenge. These days there is a strong bias, and for good reason, for new technology onboarding to be all AI related. And we know we will be competing against others who promise AI in their solution even if it doesn't add any value.

The solution to this problem for QuantiFacts is likely to be AI itself. Having been in business for over 40 years, we have reinvented our technology many times already. Now, with AI, we have a tool to enhance our applications, such as through better visualizations, smarter user querying, deeper output analysis, and better integration with other systems, while still protecting the proven non-AI core. We think the way forward for us is to use AI to make something that already works, work even better.

About the Author

Chris Nichols is President and CEO of QuantiFacts. Chris has been with QuantiFacts since 1995 and has worked as a consultant to some of the largest wealth managers in the country in new fee schedule strategy, fee modeling, selection, and implementation. He led the team to develop the Q4 Wealth Management System and Deal Manager applications, and through the years has worked with data and designed uploads from all the major trust accounting and fee taking systems. Chris understands that providing our clients value starting at the highest level of the organization relies on our precision with the lowest level of detail.