Financial Services Practice

Building a winning
AI neobank
Firms launching a neobank can increase their odds of success in
today’s uncertain environment by targeting specific characteristics
and capabilities in their strategy and execution plans.
by Anubhav Bhattacharjee, Violet Chung, Shwaitang Singh, and Renny Thomas

© Xuanyu Han/Getty Images

November 2022
 Licensed neobanks are disrupting the global of these banks, directing approximately $32 billion
financial services industry on the back of strong in estimated global venture capital funding to
trends, but their long-term success hinges on neobanks between 2017 and 2021.2
their ability to create an artificial-intelligence-
When it comes to financial performance, however,
powered banking model that is customer centric,
only a handful of neobanks have demonstrated
operationally efficient, and profitable at scale.
success. The rest have struggled to create value.
The last decade has seen around 400 launches Some of the world’s biggest neobanks have, in
of licensed neobanks,1 an overarching category fact, seen their valuations plummet in recent
that includes digital-only banks, virtual banks, months, responding to ongoing macroeconomic
and challenger banks (see sidebar, “What is volatility and shifting investor perceptions on the
a neobank?”). Incumbent banks, nonbank neobanks’ valuation drivers.
challengers (such as fintech players), stand-
In such a scenario, how can neobank start-ups
alone digital attackers, and large consumer and
set themselves up for success today? While we
payments platforms have all launched neobanks
believe long-term trends will keep the digital-
in recent years, making the competitive pool
banking industry relevant and appealing, even
increasingly vast and diverse.
amid an economic downturn, they will not
Neobanks have reset the paradigm for the guarantee profitability. In this article, we will
traditional banking industry in terms of customer describe a framework of winning characteristics
experience, product innovation, and pricing, and capabilities that any new entrant should
enabled by burgeoning consumer demand and incorporate in its strategy and execution to
supportive regulation of licensing frameworks. become a future-ready neobank.
Investors have been drawn to the disruptive power

What is a neobank?

The term neobank has been used at least been called challenger banks, virtual Today, licensed neobanks—entities
since the mid-2010s to describe fintechs banks, digital banks, online banks, and that have a regulatory license in their
that are challenging traditional banks by internet-only banks. respective geography to provide digital
providing an increasingly comprehensive banking services—may exist as stand-
Over the years, the definition of neobank
suite of banking services (as opposed alone fintechs, digital units launched by
has expanded and blurred. Many
to just payments, personal financial traditional banks, or as digital financial
traditional banks have launched their own
management, etc.) through innovative and services plays by other institutions.
exclusively digital plays, and fintechs have
low-cost digital channels. Depending on
partnered with traditional banks to provide
geography, these institutions have also
banking services.

1
“Simon-Kucher introduces the Global Neobanking Radar database to rank and track challenger banks and their path to profitability,” press
release, Simon-Kucher, May 27, 2022.
2
Data were aggregated from multiple sources, including S&P Capital IQ and Panorama Fintech.

2 Building a winning AI neobank

Differentiating characteristics 1. Launching products at high velocity
of winning neobanks Most banks aim to be customer focused, but
The banking ecosystem has undergone seismic winning neobanks go a step further in ensuring
shifts in recent years. In the case of retail customer delight by rapidly launching and reworking
banking, profit pools are large and attractive, new products and propositions. They adapt to
but banks have historically had to cope with high evolving customer needs, as well as attract, engage,
barriers to entry (for example, high fixed costs and monetize various customer segments.
and a need for trusted consumer brands) and long
While a typical bank could take years to launch a
paths to reach breakeven points. Neobanks have
new product, neobanks with an AI-first mindset
challenged these barriers by operating with lower
do it in months, sometimes even weeks. To do this,
cost structures—typically a fraction of traditional
they make substantial investments in gathering and
banks’ costs—and disruptive pricing practices;
analyzing extensive customer data. They organize
compared with traditional banks, they offer lower
full-stack teams that include people from each of
prices and more transparent pricing terms with
the relevant functions—product owners, designers,
fewer hidden fees.
and data scientists, as well as members of the legal,
That said, the real challenge for neobanks today risk, operations, and marketing teams. They also
is how to strengthen that competitive advantage maintain flexible and configurable tech platforms
by capturing a higher share of consumers’ wallet that allow relatively easy creation of new products.
and generate material profits. The answer, to
A case in point is Revolut, which had only two
a large degree, lies in embedding data and AI
products in 2015: its flagship multi-currency travel
capabilities extensively across all aspects of
card and a mobile app. By 2020, it had branched
neobanks’ operations.
into 20 products and services,3 or on average, a
AI can help neobanks deepen customer new product launched every quarter, driven by
relationships (for example, designing intelligent Revolut’s strategy of experimenting with launching
value propositions that solve unmet needs, new products across the stack to address evolving
harnessing big data to deliver hyper-personalized market trends and then doubling down on the
services and enhanced cross-selling), and products getting the most traction. In 2018, for
improve financial performance (for example, example, it launched cryptocurrency features within
maximizing customer lifetime value, dramatically its app, becoming one of the few licensed neobanks
lowering the cost to serve through automation, to do so at the time.
and adopting superior data-driven risk
management practices). In addition to having an 2. Focusing on customer engagement
AI-first mindset, neobanks also must ensure that Successful neobanks prioritize and excel at
these AI capabilities work to their advantage and customer engagement. Like their peers in the
deliver actual business value. consumer tech sector, they think of digital
engagement as a precursor to monetization—
Many of the world’s leading neobanks that have operating on the idea that share of customers’ time
achieved scale or profitability have leveraged AI precedes their share of wallet.
effectively. According to our analyses of these
successful players, they all demonstrate the As a result, their strategy is not limited to providing
following seven characteristics. the best transaction experience or pricing, or
bundling their core products. They also tend to

3
“Revolut: Neo-bank strategy deep dive,” WhiteSight, November 11, 2020.

Building a winning AI neobank 3

 go beyond pure-play financial services to deliver customer touchpoints. The more successful ones
complementary services that provide not just new are also personalizing user experience by matching
forms of utility (for example, “save and earn more,” customers’ present context and historical behavior,
“get better deals”), but also inform and entertain direction of movement (for example, predicting
(“play games,” “see what’s trending,” “learn about their current state of satisfaction and next-best
new topics”). action), and aspirations. The industry standard
for digital banks is rapidly shifting toward the n =
Offerings boosting engagement range from digital
1 level of personalization, in which each customer
and user-generated content, commerce, social
can receive unique propositions and experiences
features, gamification, and personal financial
generated by a combination of multiple machine-
management. Many neobanks provide deals
learning models.
and discounts, helping customers discover new
products, gamifying financial services (for example, Some leading neobanks are also creating
managing a “demo” investment portfolio that propositions that can automate mundane high-
tracks the market in real time), offering insightful friction tasks, based on customers’ preferences.
and entertaining content in their area of interest The UK-based virtual bank Monzo, for example,
(vacation planning, entertainment booking), and introduced automatic saving features in its app.
offering social features (for example, “send a gift to The features can round up transactions every time
your friends and family”). a customer spends and deposit that money in a
separate account and can automatically transfer a
These neobanks are effectively curating and
preset percentage of their monthly paycheck to a
designing products and services based on several
savings account on payday. While these features
factors, including the ability to engage target users
are arguably common in the online-banking
based on their interest patterns and alignment with
industry, their effectiveness depends on how
the bank’s overarching brand and purpose. For
well these nudges are dynamically personalized.
example, besides allowing customers to access
A successful bank can build models on a user’s
seamless payments and credit products at point
transaction data and demographic information,
of sale, Sweden’s Klarna uses a machine learning-
among other data, to identify actionable budget
based recommendation system to determine
and savings goals automatically.
consumer purchase patterns, and offer appropriate
shopping recommendations and financing offers.
4. Adopt conversational design
Additionally, leading AI-powered neobanks Chatbots play an essential role in customer
rigorously track their engagement performance. communication, with most banks using them to
They have dedicated teams to analyze several address customer queries and service requests.
engagement metrics, such as the average time AI-focused neobanks are going a step further
spent by a customer on the app, the number of daily by replacing forms and questionnaires with
active users (DAU), number of monthly active users conversational design in the form of chatbots, voice
(MAU), and the ratio of daily to monthly active users. assistants, and live video consultations.
Some of the best performers have achieved a DAU-
By doing this, they can combine the customer’s
to-MAU ratio of 25 to 35 percent for their overall
context and intent, using natural-language
user base.
processing techniques to deliver superior
3. Hyper-personalizing experiences engagement. China’s WeBank extensively deploys
and propositions AI in user interactions and claims that almost 98
Neobanks have demonstrated ubiquitous use of percent of customer queries are addressed by
machine-learning-enabled personalization across AI chatbots.4 This allows the bank to scale with

4
“WeBank: The world’s leading digital bank decoded,” press release, WeBank, October 31, 2019.

4 Building a winning AI neobank

higher operating leverage while serving millions of scale in a cost-effective manner by going beyond
customers every day. traditional customer acquisition routes, and
efficiently leverage partner data to provide
This characteristic also drives cost-efficiency. One
contextually relevant offers to customers in real
report estimated that operational cost savings from
time, without sacrificing their privacy. As a result,
effectively using chatbots in banking could reach
leading players can increase new-product sales
$7.3 billion globally by 2023, equivalent to roughly
within the partner ecosystem by around 50 to
862 million hours of time saved.5
70 percent, according to McKinsey estimates.
5. Integrating open-banking features seamlessly Given the strategic importance of partnerships,
While several banks are adopting open-banking many neobanks are now set up as joint ventures or
enablers, successful neobanks gain competitive equity partnerships between traditional banking
advantage by building an open platform from the players and large consumer platforms. Others are
outset. With a default open-first approach, they creating their own mega ecosystems. One example
can invest extensively in building the right API6- is South Korea’s KakaoBank, which was formed
first architecture and experiences at the time of by the parent company of the country’s dominant
setting up the bank. This allows them to seamlessly messaging app, KakaoTalk, in 2016. KakaoBank
tap into the broader open-banking ecosystem and has successfully leveraged the scale and popularity
offer customers superior products and services. of Kakao’s various ecosystems. As of year-end
Chime, a San Francisco–headquartered fintech, 2021, the bank had 18 million customers,8 and its
uses open-banking features (powered by its partner Kakao partner ecosystems now include a range
Plaid) to let customers link all their bank accounts of services, including Kakao Commerce, Kakao
within the Chime app. Customers can see all their Mobility, and Kakao Games.
accounts, balances, and transactions in one place,
7. Using customer-lifetime-value
which increases their engagement and gives them
(LTV) to guide actions
a full picture of their finances. In fact, Chime reports
Successful neobanks have a different mindset
that users who link a non-Chime account typically
when it comes to measuring performance and
spend five times more money through various
creating value. In addition to tracking traditional
purchases with their Chime debit card in their first
balance sheet metrics like the return on average
12 months on the platform.7
assets and net interest margins, which summarize
6. Leveraging partner ecosystems to scale the aggregate performance of the business,
Most banks have established partnerships with they also analyze customer-centric metrics like
social-media platforms and digital commerce, customer lifetime value (LTV), customer acquisition
healthcare, and lifestyle brands. Winning costs (CAC), and return on investment (LTV/CAC).
neobanks seamlessly embed their services within
Doing so enables neobanks’ planning and operating
these ecosystems, using technology to provide
teams to take more granular and tailored actions
customers easy, low-friction access to their
quickly, which eventually boosts customer value.
banking services every time they engage with
These actions could be strategic, such as a new-
the ecosystem partner. By doing this, neobanks
product launch that aims for a material increase
increase their discoverability, accelerate their

5
“Bank cost savings via chatbots to reach $7.3 billion by 2023,” press release, Juniper Research, February 20, 2019.
6
APIs enable data sharing between banks and third parties.
7
See, for example, Catherine Bakewell, “A better banking platform: Chime aims to help its members get ahead by making money management
easy,” Plaid.
8
Chris Wright, “KakaoBank turns to platform growth,” Euromoney, March 31, 2022.

Building a winning AI neobank 5

 in LTV or discounts intended to reduce churn measure and report the metrics to a range of
for specific customer cohorts. Other actions are teams. The decision-making processes (machine
operational—for example, determining how much driven or human) need to be aligned on optimizing
CAC can be incurred to acquire a certain customer these metrics. Decision making also must be
(Exhibit 1). embedded in the organization’s performance
rubric and operating rhythms.
For example, San Francisco–headquartered Block
Finance’s mobile payment service Cash App tracks
and reports customer LTV curves with the returns Building the capabilities
(as a multiple of the CAC) from its customer cohorts, Players interested in launching a neobank today
plotted against the time since the customers should adopt an AI-first mindset and build a holistic
started using Cash App. According to Block,9 Cash set of capabilities in four interconnected layers
App reaches CAC breakeven within six months of (Exhibit 2). This approach will equip banks to acquire
acquiring a customer and earns six times CAC within many of the differentiating characteristics.
18 months. Block also has reported that Cash App
Engagement layer
has also consistently improved its monetization
To succeed in the current competitive
performance for each customer cohort from 2017
environment, neobanks need to create
to 2020.
experiences and build accompanying capabilities
Tracking such metrics is not easy. Several front- that fulfill a range of customer expectations.
end and back-end systems must be adapted to First, the engagement layer needs to deliver

Web <2022>
<Neobanks>
Exhibit <1>
Exhibit 1 of <2>
Cohort analysiscan
Cohort analysis canshow
showimprovement
improvementinincustomer
customer acquisition
acquisition andand
monetization strategies
monetization strategies forfor neobanks.
neobanks.
Cohort return on customer 10× Year 1
acquisition cost, illustrative, Year 2
multiple Year 3
8×
Year 4

6×

4×

2×

0×

0 12 24
Months after joining

9
Amrita Ahuja, “Block Investor Day—business model,” Block Finance, May 18, 2022, available on YouTube.

6 Building a winning AI neobank

Web <2022>
<Neobanks>
Exhibit <2> of <2>
Exhibit 2
The AI–digital bank of the future has four layers and many elements.
The AI–digital bank of the future has four layers and many elements.
Structure of the AI–digital bank of the future

Engagement layer
Intelligent Beyond-bank Low-friction Embeddable AI-enabled Digital-first
products and engagement omnichannel (into platforms) conversational brand and
services modules journeys user interfaces interfaces identity

AI- and analytics-led decisioning layer

Digital marketing and personalization stack

AI/machine learning–driven life cycle decisioning

Customer Credit decision Monitoring and Retention and Servicing and

acquisition making collections cross-sell/upsell engagement

Edge technologies
Natural- Virtual Facial Robotics Voice Computer Blockchain Behavioral
language agents/ recognition script vision analytics
processing bots analysis

Core technology and data layer

Data Modern Intelligent, Flexible and Cybersecurity Banking-as-a-
management API scalable configurable and controls tiers service modules
for AI world architecture technology product core
infrastructure

Operating-model layer
Autonomous Agile, learn-fast Modern, Digital-native Hybrid teams and
full-stack business way of working nontraditional culture and operating model
and technology talent capabilities
teams

experiences that are not only easy, intuitive, fast, Third, the experiences need to be tailored to each
and responsive but also delightful. To do this, they customer’s needs and context and be hyper-
could create low-friction, low-latency, and highly personalized. For this, neobanks need to set
customizable mobile-first journeys. up a product innovation engine that allows for
the discovery of unarticulated customer needs
Second, these experiences need to be available
and the flexibility to introduce innovative new
where customers need them the most. One way for
products. Equally important is to ensure the
banks to do this would be to build their products
flexibility of the entire interface (for example,
and interfaces in a way that they can be seamlessly
enabling it to change the products and journeys in
integrated with various ecosystem partners,
varied ways for each customer). The engagement
allowing customers to discover and consume
layer should also be suitably linked to the models
propositions beyond the bank’s core platforms.

Building a winning AI neobank 7

 in the decisioning layer, to achieve the third goal, banks invest in digital capabilities, their money is
sending customers real-time personalized offers, mostly spent in the maintenance of their existing
nudges, and recommendations. core technology. Far less investment goes to
building new cloud, data, or API capabilities.
AI-and-analytics-led decisioning layer
Neobanks unencumbered by legacy decisions can
Hyper-personalization requires granular AI-driven
leapfrog incumbents by building their core tech.
decisions across the entire customer life cycle,
Their goal should be to develop a cloud-based,
spanning acquisition, onboarding, servicing,
flexible, and highly configurable banking core that
retention, and cross-sell. This requires many
will help engineering teams launch new product
machine-learning (ML) models across domains
variants rapidly through easy configurations. To
including credit analytics, product and channel
do this, neobanks would need to make judicious
propensity, fatigue (how often neobanks are
decisions on which components to build on their
reaching out to customers), and risk (including
own and which to outsource. For example, most
attrition, collections) to identify the next-best action
neobanks use externally available cloud-based
for customers.
infrastructure and storage services.
Some essential capabilities are required to build
As a starting point, neobanks should design a
and effectively deploy these models. The first
modular, micro-services-led, API-first architecture.
relates to data gathering. Since ML models are built
This should be able to scale efficiently, allow high
on data, neobanks need to create data ingestion
reusability to reduce the time to market of new
pipelines. Data collated from these pipelines
features and the overall cost to maintain code, and
should include traditional banking customer
ensure ease of integration with various partners
information, transaction data, and less-traditional
and the open-banking ecosystem. Additionally,
data streams such as clickstream data from a
neobanks should ensure that data management
bank’s platforms (app or website) or data from
enables data liquidity, including the ability to
third-party partnerships with various platforms and
access, ingest, and manipulate data across systems
aggregators. All the data should then be stored in
and models. This crucial element serves as the
a unified data platform that aggregates, enriches,
foundation for all decisions in the decisioning layer.
and maintains a 360-degree view of existing and
potential users, providing complete visibility on Operating-model layer
all customer interactions with the bank across The final layer—and a crucial determinant of a
products, transactions, and service requests. This neobank’s speed and agility—pertains to how it
will help neobanks develop new decisioning models. attracts and organizes talent across its teams in
a way that breaks traditional organizational silos.
Once the data have been collated and the model
Three capabilities are key here. First, a successful
has been created, neobanks also need to be able
neobank should be able to set and embody an
to rapidly deploy, test, and iterate on their models
aspirational vision and culture for the organization.
in live production environments. This requires
This will help it attract leading talent in banking and
technical capabilities such as machine-learning
critical nonbanking domains, including designers,
operations (MLOps) and close coordination among
architects, engineers, and data scientists. Second,
data science, product, technology, and business
these neobanks should have full-stack platform-
teams to ensure the models are built for live and
based teams with all the skill sets required for
efficient production deployment.
their end objective—for example, driving revenue,
Core technology and data layer enhancing engagement, or ensuring service
Developing experiences in the engagement layer quality and increasing operational efficiencies.
and personalizing the decisioning layer require a These teams should also have autonomy to
modern technology and data core. When incumbent define and execute actions aligned to their goals

8 Building a winning AI neobank

 while being loosely coupled with the rest of the Despite current headwinds, the long-term growth
Find more content like this on the
organization. Finally, the neobanks should calibrate trajectory of neobanks is expected to remain strong.
McKinsey Insights App
the organizational rhythms and processes that Those starting their digital-banking journey today
promote innovation and rapid experimentation while should be focused on preparing for disruptions of
adhering to regulation and compliance norms. all kinds. The key to success lies in making AI the
centerpiece of the neobank’s design, architecture,
operations, and the whole gamut of the customer
experience strategy.

Scan • Download • Personalize

Anubhav Bhattacharjee is a consultant in McKinsey’s Mumbai office, where Shwaitang Singh is a partner and Renny Thomas
is a senior partner. Violet Chung is a partner in the Hong Kong office.

Designed by McKinsey Global Publishing

Copyright © 2022 McKinsey & Company. All rights reserved.

Building a winning AI neobank 9