Winning the First Impression:

Special thanks to our summer research associate, Tireni Ajilore, for leading research and writing efforts for this series.

Earlier this week, we introduced how generative AI will reshape the retail and SMB banking journey, and the opportunities that it presents for both startups and incumbents at each stage. Today, we’re exploring the first two stages: the crucial moments when banks first meet, evaluate, and onboard new customers.

These initial interactions, from first click to final approval, shape everything that follows. They determine not just whether a customer joins a bank, but how much they’ll deposit, how many products they’ll use, and how long they’ll stay. Yet despite decades of digital transformation and billions invested in technology, these early touchpoints remain costly for banks, frustrating for customers, and surprisingly ineffective at building lasting relationships.

With generative AI, banks can now reduce costs and increase conversion rates while fundamentally reimagining how financial relationships begin in the digital age.

Stage 1: Customer acquisition

For banks, acquiring new customers has always been a costly and complex challenge. In 2024, customer acquisition ranks as the second most pressing issue facing banks, just behind deposit gathering—an issue closely linked to bringing in new customers. In an increasingly competitive market where fintechs and neobanks are lowering CAC and offering personalized, frictionless services, incumbent banks are feeling the pressure to evolve.

Historically, banks have relied on traditional data analytics and rule-based systems to acquire customers, drawing from static demographic data and basic segmentation to build marketing campaigns. These methods are reactive and often limited in their scope—essentially painting customers with broad strokes and missing the nuances of their behavior and preferences. Generative AI shifts the paradigm by enabling banks to analyze both structured and unstructured data, providing the ability to create more nuanced customer profiles and deliver truly personalized content at scale.

Generative AI isn’t just changing how banks market; it’s transforming when and how banks engage with potential customers. Instead of relying on periodic, pre-planned campaigns, generative AI empowers banks to make real-time adjustments based on customer behavior, allowing for micro-segmentation and hyper-personalization that delivers tailored messages precisely when they’re needed.

Opportunities for incumbents

Incumbents possess the infrastructure to support new customer onboarding but lack the technology required to meet new customers where they are. GenAI changes that.

1. Dynamic personalization engines

Generative AI is taking customer segmentation to a new level. Banks now have the ability to move beyond traditional segmentation—based solely on demographics or transaction data—and build richer, more detailed customer profiles by incorporating unstructured data, such as browsing history, customer interactions, and social media activity.

Incumbents can also leverage generative AI to dynamically adjust their marketing strategies in real-time. These AI-powered personalization engines learn continuously from customer data, improving the accuracy of predictions and enabling personalized experiences that feel uniquely tailored to each individual. For example, companies like Personetics use generative AI to analyze transaction data and customer behavior, creating evolving customer profiles that allow banks to adapt their strategies on the fly—whether to acquire new customers or strengthen relationships with existing ones.

Generative AI also enables micro-segmentation, allowing banks to craft highly personalized offers for specific customer groups or even individuals. Unlike traditional AI, which follows predefined rules, generative AI can dynamically evolve based on new inputs, enabling it to predict customer needs with increasing precision and offer personalized content and services at scale.

“As a bank, I can engage customers to tell me more about what they want to do and share insights about their financial lives. By combining this information with the first-party data I already have—much of which is unstructured—and integrating it with additional data the customer shares through open banking, I create a rich tapestry of exactly what the customer is doing today and where they want to go.” – Alex Johnson, “What Large Language Models Mean for the Future of Fintech, with Rohan Ramanath of Hyperplane,” Fintech Takes, May 22, 2024

2. Conversational marketing

In the era of digital banking, the in-person interactions that once drove customer acquisition have largely disappeared. However, conversational AI is filling this gap by allowing banks to create experiences that feel just as personalized and engaging as a face-to-face meeting.

For example, AI models can predict when a customer is likely to make a significant purchase, such as a car, by analyzing large volumes of consumer data like social interactions and browsing history. Generative AI can then personalize the customer experience based on these insights, such as creating tailored marketing messages, educational content about the car-buying process, and tools to compare financing options. Conversely, if the customer needs help improving their credit, generative AI can shift the conversation to providing tips on rebuilding credit, delaying an auto loan offer until the customer is better positioned.

Currently, only about 30% of bank customers receive personalized financial advice, despite the high demand for it. Generative AI can close this gap by delivering personalized advice at scale, whether through interactive reports, virtual advisors, and chatbots.

For example, Intuit Assist within Credit Karma uses generative AI to provide personalized financial guidance by analyzing user data and offering tailored advice on budgeting, savings, and investments. Similarly, OneClick Financial enables credit unions to offer similar AI-driven personalized services, such as tailored loan offers and financial education, to their members.

“We can determine someone’s propensity to buy an auto and then provide them with information about the auto buying process, educational resources, and pre-approval offers. Conversely, if they need to improve their credit score, we won’t show them an auto loan offer. Instead, we’ll present content on how to become creditworthy enough for an auto loan, including products to help rebuild their credit.” – Chief Innovation Officer at Large Credit Union

Opportunities for startups

Data is the crux of opportunity for startups—being able to create, analyze, and serve the important unstructured data that incumbents have difficulty processing. 

1. Unstructured data infrastructure

The banking industry is swimming in data, but much of it remains untapped. According to Fintech Futures, around 80% of banking data is unstructured—including audio, video, and email files—while only 20% consists of structured data, such as names, addresses, and credit card numbers. This imbalance poses a major challenge for traditional personalization engines, which struggle to process unstructured data effectively.

This challenge is widely recognized in the industry. In a recent survey, 73% of banking leaders acknowledged that turning customer data into actionable insights is a significant hurdle, with 95% pointing to restrictive operating systems as the primary obstacle.

This is where startups come in. Fintechs like Cognaize and Senso are developing generative AI tools that clean, label, and structure unstructured data, making it easier for banks to extract valuable insights and automate processes Rather than sell generic “insights,” startups can target specific lines of business and address use cases that directly impact business outcomes—for example, improving customer acquisition metrics by leveraging unstructured data for more personalized, dynamic engagement.

“Personalization has been something we’ve been working on for years, initially branded as a personalization engine. However, it hasn’t yet lived up to the hype, primarily due to the quality of data it leverages. Our customer data isn’t fully connected in a way that allows for truly meaningful personalization. There’s definitely an opportunity here, especially with AI, but we don’t currently have anything live in production that’s groundbreaking.” – SVP Unsecured Lending, Top 20 US Bank

“The cost of transaction monitoring is sky high—and regulations are nowhere close to keeping up. Meanwhile, we have to do data sharing better to make finding financial crimes more effective.” – CEO of a Community Bank 

2. Synthetic data generation

Another major challenge banks face is the ability to access high-quality data while maintaining security and adhering to stringent privacy regulations. Traditional methods like data masking or anonymization are often slow, resource-intensive, and carry a risk of reverse engineering. This is where synthetic data steps in.

Synthetic data is revolutionizing how banks train and test their AI models. By generating artificial datasets that mimic real ones, synthetic data enables banks to innovate in a compliant, secure environment without risking customer privacy. Startups like Hazy and Gretel are building platforms that generate synthetic datasets with the statistical properties of real data, while ensuring complete anonymity.

“Synthetic data generation allows us to think about the full lifecycle of a customer’s journey that opens an account and asks for a loan. We’re not simply examining the data to see what people do, but we’re also able to analyze their interaction with the firm and essentially simulate the entire process.” – Manuela Veloso, Head of AI Research, JP Morgan

“Achieving our vision for perfect personalization, especially for a client-facing portal, requires flawless data and perfect data matching. We need to accurately identify users, whether they have accounts with us or not, and combine third-party data with our existing and newly collected data. The challenge of accurately understanding users to personalize their experience is an ongoing problem, even with advances in AI.” – AI Lead at Mortgage Lender

3. Marketing compliance

In a market where financial products often appear similar, standing out while staying compliant is a significant challenge. For banks, creating marketing that resonates with customers while meeting regulatory standards is both a challenge and an opportunity. In Q1 2024, regulators finalized 17 enforcement actions against consumer finance companies, with penalties totaling over $59 million for violations like misleading offers, discriminatory practices, and misrepresentation of FDIC insurance.

Startups are now helping banks stay compliant by using generative AI to analyze marketing content in real-time. Unlike traditional tools that scan for specific keywords, generative AI understands the context of a message, allowing it to detect subtle risks or potential compliance violations that may otherwise go unnoticed. Companies like Sedric offer products that can screen potential partners, monitor affiliate content, and flag risks, ensuring that marketing materials remain compliant at all times.

“Even with traditional statistical models for targeting, our robust fair lending assessments create barriers to nimbleness. While the vision of a future where we can crank out a marketing campaign in 24 hours is enticing, governance extends these timelines significantly. The same fairness requirements apply whether we’re using AI, heuristic criteria, or statistical models. In marketing, we can demonstrate fairness by analyzing treatment outcomes and target distributions. However, in underwriting, we must explain lending decisions clearly, which often means sacrificing predictive quality for explainability. This makes navigating regulatory constraints in marketing more feasible than in underwriting.” – EVP Lending, Top 20 US Bank

Stage 2: Onboarding and underwriting 

A 2020 study by Deloitte found that digital onboarding in banking has increased registrations by nearly 30x. Digital onboarding simplifies the customer’s journey, offering a seamless and user-friendly experience that aligns with modern expectations.

Yet as digital onboarding becomes the norm, banks face a new challenge—scaling these processes to handle increasing volumes of data, prevent fraud, and deliver the level of personalization that customers now expect. For example, onboarding new business customers can still take anywhere from 90 to 120 days, largely due to compliance requirements like KYC, KYB, and AML. These procedures are typically manual, requiring redundant data collection and repeated back-and-forth communication, which not only slow down the process but also increase the likelihood of errors.

​​The impact of these slowdowns is staggering: 68% of customers abandon financial services applications due to cumbersome onboarding processes. For small business owners, the frustration is particularly acute, with up to 75% abandoning their applications due to repetitive paperwork and long waits for status updates.

Opportunities for incumbents

Banks have made incremental progress in automating onboarding and underwriting, but meaningful inefficiencies remain. Integrating generative AI into these workflows can help banks overcome them.

1. AI-Assisted Workflows

According to Capgemini’s 2024 World Retail Banking Report, customer onboarding teams in banks spend 55% of their time on routine tasks like paperwork and document verification, and another 36% on compliance and risk checks. This not only wastes valuable time but also creates friction points that can lead to missed opportunities and lost revenue.

Generative AI can solve these problems by automating key processes like document verification, customer onboarding, and loan origination. With AI-powered systems, customers can upload documents online or via mobile apps, and the system instantly verifies them for accuracy and completeness. AI can cross-reference these documents with regulatory databases, identifying potential issues and ensuring compliance without manual intervention. 

When it comes to originating loans, generative AI can gather financial data, check credit histories, and look for red flags in public records—all automatically. Together, these advances enable banks to save time, reduce errors, and focus more on opportunities that drive revenue.

Opportunities for startups

Startups have the flexibility to design AI-driven solutions from scratch, allowing them to create more agile, customer-centric approaches to onboarding and underwriting. 

1. Virtual onboarding assistants

Startups should aim to move beyond merely embedding AI into existing workflows and instead focus on fully automating specific verticals of lending using AI-powered solutions. Traditional methods often frustrate customers with confusing steps and long wait times, leading to low completion rates. In contrast, a fully automated AI-driven system can dynamically adjust onboarding steps in real time based on user behavior, offering tailored guidance that reduces friction and keeps customers engaged.

A great example is Foundation Capital portfolio company Lama AI, which is using LLMs to disrupt loan origination for banks of all sizes. Lama helps financial institutions, fintech companies, and other SMB-focused organizations launch and scale business lending products through its AI-powered, API-first technology. More specifically, they’re leveraging generative AI to address a key challenge in small business lending: processing vast amounts of unstructured data such as financial statements, business plans, and cash flow projections.

Traditionally, lending institutions have had to invest significant time, resources, and labor into cleaning and categorizing this data before feeding it into their underwriting models. LLMs, with their ability to directly process and act on unstructured data, allow Lama AI to bypass this costly underwriting process, and make even the smaller ticket loans profitable. Lama AI is also able to continually train its LLMs with additional, domain-specific data to ensure data accuracy and better meet customer needs.

This allows banks to automate nearly the entire origination continuum—from efficiently onboarding new borrowers and enhancing the application process with AI-generated suggestions and next steps for relationship managers and loan officers, to fully automating email and text communications. By reducing underwriting costs by 10x, Lama AI enables banks to profitably expand their reach, tapping into online marketplaces and embedded lending ecosystems that were once dominated by alternative lenders.

Another example is Casca which provides an AI assistant for SMB loan origination that claims to have meaningfully boosted completion rates. Their AI assistant prompts users to upload necessary documents and sends personalized follow-up messages, allowing small business owners to complete their applications outside regular business hours. This kind of end-to-end automation in specific lending verticals offers a clear path for startups to differentiate themselves from horizontal solutions.

2. Alternative data

As AI underwriting models become increasingly commoditized, simply offering “better” credit scores or risk assessments will no longer provide a sustainable competitive advantage. AI tools that generate accurate credit scores are becoming more accessible to financial institutions, diminishing the unique value they once provided.

To stand out, startups need to go beyond conventional credit assessments and offer specialized insights tailored to specific verticals. For example, in home lending, a startup could develop a platform that aggregates data on property value trends, neighborhood development, and local economic indicators. By analyzing this data alongside borrower creditworthiness, lenders gain a more nuanced risk profile—allowing them to consider not only the borrower’s ability to repay but also the inherent risks of the property itself.

By building these kinds of deep, sector-specific insights, startups can provide financial institutions with tools that offer a more holistic view of risk, helping lenders serve their customers more effectively and profitably. This strategy positions these startups as essential partners for lenders and helps them maintain a competitive edge in an industry where basic AI underwriting is rapidly becoming table stakes.

For an overview of the consumer banking lifecycle and how generative AI fits in at each stage, check out our recent post. In the coming weeks, we’ll take a closer look at the remaining stages, concluding with a market map of key startups in the space.