ChatGPT is another "shiny" distraction

KeyBank Going "All In" on Cloud, No-Code Fantasy No More

Welcome to the inaugural issue of FSI Digital Transformation Weekly! If you end up finding this content worth your time, please share it with your network.

ChatGPT is another "shiny" distraction

In a quest to transform everyone & everything except themselves, FSI executives naturally gravitate toward groundbreaking announcements that promise easy digital wins. Who wouldn’t when “Cloud” “no-code” and “AI“ are enticing you with the promise of accelerating business value without the hard work, all thanks to the genius of technology itself.

The latest excitement comes from AI’s spin-off, ChatGPT, which created so much buzz that even the world’s largest hedge fund hopped the AI train. It’s now conducting layoffs to free up every penny for developing AI:

If ChatGPT is so ground-breaking, maybe it could even advise FSIs on how to leverage it (sorry for the large image below - ChatGPT had a lot of specific suggestions for you):

It looks like a standard list for leveraging AI in financial services circa 2016. Underwriting is missing explicitly, but it might be implied under “Risk management.” An increasing lack of novelty and related marginal impact is common for innovation. In the case of Customer Support, 20th-century interactive voice response (IVR) and web FAQs are mostly remembered for annoying customers, but they did enable significant efficiency gains. Scaling natural language processing (NLP) in the last 10 years made automated support less annoying but also had a smaller relative impact. There might still be some remaining efficiency gains left, but a typical online UX by FSIs in 2023 doesn’t leave much of FAQs left to fix.

For Fraud Detection, here is the current state of trade-offs in leveraging AI from the interview with a cybercrime expert in PYMNT:

“AI can be easily fooled. I’ve seen it write more programming mistakes and introduce more cybersecurity vulnerabilities than a normal person would do… I don’t think that a definitive AI solution against hackers is coming soon, and even in the short-term it may introduce more problems than will fix problems…”

What about using ChatGPT for Personalized Recommendations? Here is how my primary bank described its AI focus, by its Chief AI Officer (!) in 2022:

"AI is helping to transform the financial sector, and at [bank name], we're adopting AI to help build meaningful relationships with our customers and to help improve internal processes so that our colleagues' experiences are enhanced"

And how is that AI adoption going so far? After a decade of having me as a customer, the bank’s two best offers for me are 1) a credit card with free balance transfers, 2) free A2A transfers:

I never carried a card balance, and I have been doing free A2A transfers with this bank for as long as I remember. Does an FSI really need a more advanced AI engine to tell them to stop sending customers such “meaningful” offers?

And how much difference would it make if your whole FSI was “AI-first”? You don’t have to guess, insurtech Lemonade was launched under that premise in 2015 and keeps claiming AI as its key differentiation today:

And yet, Lemonade’s market cap remains 80 times smaller than that of an 80+-year-old Progressive Insurance with no signs of imminent turnaround:

What went wrong? Lemonade violated the #1 rule of scaling, the importance of focus: add an additional segment, product, channel, or region only after winning market share or shutting down the previous ones. Lemonade launched 5 products and opened in 5 countries before succeeding anywhere. Maybe one day AI becomes capable enough to prevent humans from making bad strategic decisions, but there might be some unintended consequences of such judgments.

So why do you, FSI executive, really care about ChatGPT since it’s clearly not going to make your company more money this year? In part, it’s due to FOMO as being honestly acknowledged by Chime’s CTO in the Insider interview:

Barrese acknowledged that, beyond the initial security and ownership concerns, the wide use of AI will bring out a lot of philosophical and ethical questions around how the tech is used and abused "that used to be theoretical and now are real."

But for now, the fear of missing out is enough to get Chime in the game. 

"I've seen this play out time after time, right? Where there's a new innovation, and then everyone in the industry widely acknowledges that this disruption is coming, but then they turn around and refuse to engage in it," said Barrese, who was the former CTO of PayPal and VP of technology at eBay. 

But, more importantly, you are spending invaluable hours on another “shiny” topic as an easy excuse for not personally addressing the persistent and well-known gaps in your company’s business, operating, and technology models. Sorry, but AI can’t solve those for you.

KeyBank Going “All In” on Cloud

It’s one thing when a tech/SI/consulting/research firm tells an FSI that they must be “all in” on X technology (cloud, API, no-code, you name it), or something really bad would happen (usually disruption by Big Tech or fintechs). Fear sells.

It’s less obvious why, in 2023, so many FSI executives are still signaling their unequivocal support for any technology, like KeyBank’s CTO view on Cloud in the Wall Street Journal article:

“We have been on a multicloud journey since 2017,” says Keith Silvestri, KeyBank’s CTO. “After analyzing the benefits, it became clear in early 2022 that it’s time to go all in on cloud.”

What were those clear benefits to warrant “all in” Cloud transformation?

• economies of scale

• faster iteration

• environmental impact

It’s a standard justification package for new technology: cost savings, faster time to market and some feel-good add-ons about community/environment. None of those reasons are necessarily true. While on-prem costs have been stable, Cloud spending is growing exponentially, impacting business models of the at-scale enterprises (remember this seminal 2021 analysis from a16z)…

and causing a growing Cloud repatriation trend:

Another benefit of Cloud, faster iteration, could be often counterproductive for FSIs since most of your companies are stuck in the IT Product phase of operating model evolution. Yes, it’s fun to deploy more features in each sprint, but without hands-on business ownership, many of those features are at best a waste and at worse would annoy your customers. Cloud migration in order to decrease environmental impact could work - you just need to agree with your cloud provider on using its service only during sunny & windy hours.

Notice what is missing in 90+% of similar FSI case studies: hard/verifiable numbers around the magnitude of benefits and related ROI. Getting those estimates directionally correct requires straddling business and technology knowledge which very few FSI executives are capable of (yes, including you). Doing this exercise routinely would enable KeyBank to not just better predict Cloud’s TCO which is very high as the bank learned in the last couple of years…

… but also learn that the benefits of cloud migration tend to be much smaller due to more systemic impediments in the operating model. KeyBank’s CTO does want to avoid a “big-bang” approach:

“The cloud journey is like when you move into a new house—you can just grab all your stuff and do it in a hurry, or you can take the time to be deliberate, go through everything, and do it right. ”

However, such a transformative decision is not a supply-side question about the pace of setting up a new house. A more realistic analogy is of a city with different districts: business, residential, entertainment, and government. While using some common infrastructure, each of KeyBank’s numerous product lines within main LOBs has its own technology needs and effective pace of evolution. The difference in technology maturity could be up to 20 years across those products, and there is usually no ROI for narrowing that spectrum.

Only a leader of each district with P&L autonomy is in a position to determine whether they could commit to capturing massive P&L improvement in order to justify the high costs of cloud transformation. They shall be the ones to prove the realism of this strategy for their product line in front of the Executive team, not the CTO. And under those conditions, no business executive would be “all in,” and some might be even ok to wait another decade.

No-Code Fantasy No More

With Unqork’s third round of layoffs and disappearance from organic news, No-Code platforms joined the long list of FSI technology innovations that were initially sold as a game-changer but ended up as another useful tool with clear trade-offs. Unqork’s FSI client roster is impressive, and the recent adjustment is likely to be more of a pivot than a fundamental change in strategy:

I first heard about Unqork from FSIs in 2018, and the value proposition seemed incredible: FSI business executives could finally fire their obtuse IT counterparts, and, instead, get needed functionality directly from a vendor via an Uber-like experience while paying only after a proven business impact. And the business impact was claimed to be extraordinary:

“Unqork's market position in the insurance and financial services industries as the only proven platform that solves the industry's risk assessment and client acquisition digitization problem, enabling clients to focus on their core value proposition of providing great products. Simply put, Unqork has shown to reduce costs up to 40 percent and increase policy sales by 20 percent – all while providing seamless, instant customer experience.”

I was very intrigued and offered a free independent validation of those claims, but why bother with more analysis when FSI executives seemed to love Unqork? By early 2021, Unqork was valued at $2B, received $380M in funding, and was growing 277%.

Around the same time, the feedback loop became mixed. Some FSI executives among initial clients were disappointed to discover that a No-Code innovation couldn’t fix or replace slow internal processes. Instead of full risk sharing, the configuration costs were high and had to be paid regardless of the business outcomes. The configuration process was cumbersome enough, so rather than circumventing IT, Unqork mostly pivoted to a traditional engagement model with developers as core users.

By 2022, some Unqork clients started to pull partially back in order to cut costs and improve flexibility. Here is how a Liberty Mutual developer describes the benefits of migration from Unqork:

“… team had to build a similar application and help with the migration from Unqork to the in-house application. Doing this saved the company over $10 million, lets us have ownership of all our data, and more direct control and flexibility over our application.”

Today, a No-Code solution is a useful tool under specific constraints:

• Application rather than a platform - can’t build a different app on top of it

• Pre-built components - some configurations might not be allowed or could be costly, more complex requirement means more configuration costs

• Value proposition - saving development effort via pre-built components

• Risk - never-ending engagement with expensive configuration partners, high usage-driven costs

Unqork did push some business executives to finally get involved with technology initiatives, and if your FSI is lucky to have such folks more reason to consider a no-code solution in those specific LOBs. Otherwise, a positive ROI is still possible but mostly driven by targeted reductions in the development cost, and without FSI executives playing in the “citizen developer” fantasy.