- Key insight: The career path of starting as a teller and rising through the ranks is changing as AI takes on more entry-level work.
- What's at stake: Not doing grunt work means workers miss the opportunity to learn the basics and recover from mistakes.
- Expert quote: "It would have been very, very difficult for me to become an executive in loan servicing had I not processed loan extensions or payments in our operating systems, and 20 years later I can still tell you the code names and the screen transaction codes from our old green screens." —Carissa Robb, managing partner, SolomonEdwards
Banks have traditionally been a decent place for people to get their start and work their way up the ladder. There are people who started as tellers and eventually became CEOs, like Bob Rivers at Eastern Bank and former U.S. Bank CEO Richard Davis. As AI is used to automate more entry level work, some people wonder how young people right out of college will learn the fundamentals of the business.
Carissa Robb, managing partner at consulting firm SolomonEdwards and former senior vice president and head of U.S. loan servicing at TD Bank, shared her thoughts with American Banker in an interview.
A McKinsey analysis found that banks are cutting junior analyst classes by about two-thirds while sourcing 62% of their AI talent from the junior analyst candidate pool, indicating a shift in the kinds of work entry-level employees do. How do you see AI affecting entry-level jobs in financial services?
CARISSA ROBB: I'll start by saying I was a beneficiary of joining the banking world in an entry level position. I started in collections, I moved through loan servicing, and had the opportunity to understand the infrastructure of a bank and how data influences critical decision points. As we start to look at AI and the junior level positions, there is an absolute place and benefit for it, where it can accelerate the learning curve, but it can't replace the exposure that we need to accumulate to create that next generation of bankers and decision makers, as we move through a more complicated banking environment.
So there's a need to preserve some of those entry-level jobs. Do you think collections jobs will be replaced by AI?
CARISSA ROBB: I think they're being strengthened. The more access you have to information, the more you can refine the art of collections. The science is helping you be more effective, but the two need to go together, because it's a very human experience to call and collect a payment on a person's mortgage that is 30, 60 or 90 days behind, or you have to have that conversation of rising credit card debt, and maybe you also missed an auto loan payment. That is highly personal, and to recover and put them into a solution requires both insights that you can get from data at a much faster pace [through the use of AI], but it also takes that human element. People want to pay the person that they can connect with more than the automated telephone system that is harassing them for payment.
Banking has long been an apprenticeship business. New employees learn from mentors for a number of years, and then they get their chance to move up. Do you think that's going away or changing?
CARISSA ROBB: I think it's changing, and I think it's evolving, so I hope that it isn't going away. What I worry about is the training from failing and how they will get the exposure to learn to be wrong, how to recover from misinformed decisions, or perhaps a failure of a model output. There's a value in that discomfort of getting it wrong and being held accountable and being able to explain and course-correct under supervision, that I think will have to find a balance with apprenticeships that are now accelerated by the use of AI and model validation.
Will entry-level people become model validators, and if so, how do they do that without having that work experience?
CARISSA ROBB: That's the gap. The subject matter experts we have now, who have the benefit of experience from manually underwriting, for example, or manually processing a certain transaction, are the ones that should be included in model validation right now. So instead of making an op-ex move where you strike a high percentage of your workforce, those mid-managers, you need to retrain them so that they become your testers and your validators, and that in time will create this feedback loop with junior folks that are exposed to this, so you're giving them a place to interact with the AI output critically, let them interrogate that recommendation, look at the assumptions, break the assumptions, almost argue with it rather than simply accept and execute.
So today's subject matter experts become model testers and validators, and then the more junior people query that work to understand how that was done. Do all subject matter experts want to do this kind of work, or are they prepared to do this kind of work?
CARISSA ROBB: That's a cultural oversight in many institutions. Certain segments of the population are quite intimidated by the use of AI, and the adoption rates are impacted by that, so the banking cultures that are taking time to say, "you have a place here, we're including you in the proper implementation and creation of our AI tools as a teaching tool," the ones that are capturing the knowledge in the development of the tools, those are the ones that are most successful at creating space for SMEs that we still need.
Not everyone will feel comfortable or interested, and this is where we start to find the gray area between those that are self-selecting out — where do they go, are there jobs available — and those that are nervous, but still leaning into their own development. And then the ones that are able to fully transition and embrace this new role of learning, that's where we'll see the most success for the entry to mid-level folks.
How will young people right out of college develop those critical thinking skills?
CARISSA ROBB: The training programs are pretty critical at the financial institutions. If you remember way back, GE used to have cross-functional apprenticeships or internships, where you would learn different segments of the business. Those kinds of rotational programs where they can learn and interact with different areas will be really important. They need to have as much opportunity to operate in the first line as possible through those rotational programs, so that they can get as close to the operating systems, as close to the customer profile, and the functional point of what that task is doing for revenue or for customer growth as possible, and as frequently as they can.
When you attract someone out of college, if you are only posting model developer positions, validation positions, testing roles or automation roles, and you're foregoing those introductory rotational positions or internships, you're not really creating the space for them to learn at the ground level that calibrated judgment of learning how to be wrong in recoverable ways under supervision, learning anomaly detections. For certain sized institutions, that will be difficult to pull off because of pressure on operating expenses and budgets.
Is this new wave of interns that you're describing useful, or are they just in training, and eventually they'll be useful, they'll do actual work.
CARISSA ROBB: When you're hiring in customer support roles, collection roles, entry-level positions, tellers, you're hiring someone on the premise of, come into this organization and learn it. So, there's always going to be an element of training that comes with someone who is fresh out of school, more green, doesn't have that experience and exposure, and you teach that through this rotational program. The challenge that we're going to have is, will that be exciting and attractive, or do they want to just jump ship and come right into the model validation and the AI implementation, because that's where all of the noise and the excitement and the attention is.
Switching gears just a little bit, Jamie Dimon, CEO of JPMorganChase, told investors in February that his bank has already displaced workers because of AI, and though the bank's head count stayed flat at around 318,512 people, the number of operations and support staff fell by 4% and 2% while claim-facing and revenue-generating roles grew by 4% Is that a trend you see happening across the board?
CARISSA ROBB: Yes, I think the positions that are more task-oriented are benefiting at a faster pace from automation, and so that makes sense that some of the operational or support roles, the call center support roles, a lot of the analyst positions supporting some of the underwriting positions, have been automated. So where there are consistent guardrails, processes, repetitive behavior, that certainly makes sense, that is the point. We want to introduce operational efficiencies for those value-added positions, but in a more automated and scaled way.
The piece of data that I would love to see from the industry is of those customer-facing roles, how many of them started in the operational roles that you've eliminated or optimized. That's where we need to be careful of that compounding effect, how much of that workforce was created and became successful in revenue generating or customer interacting positions, because they had the opportunity to move through the operational and back office support model.
I know personally it would have been very, very difficult for me to become an executive in loan servicing had I not processed loan extensions or payments in our operating systems, and 20 years later I can still tell you the code names and the screen transaction codes from our old green screens. That foundational knowledge has influenced my ability to lead and optimize operational centers, or productize or automate certain flows. Without that, what does that transition look like? What does the influence and impact look like on the next generation of executive leaders sitting across from the CFPB or the OCC or other federal regulators in a defensible position and saying, yes, I know my business and I know what I'm great at, and I know where I'm exposed? I don't know that we've totally cracked the code on that for the next generation of leaders.
