Citi Ventures Champions AI Investments in Lending and Automation
- Key Insight: Citi Ventures is strategically investing in artificial intelligence firms that focus on automating lending processes, coding tasks, and a broad spectrum of manual labor.
- Expert Quote: “In the past, lending and underwriting were guided by predetermined rules. One would establish these regulations, then assess whether the client met specific cash flow criteria or possessed a requisite FICO score.
Nowadays, it will integrate some established rules with the insights generated through artificial intelligence.” — Arvind Purushotham, Head of Citi Ventures
Amidst discussions regarding a potential AI bubble, Arvind Purushotham, the leader of Citi’s venture capital division, remains undeterred.
His focus remains steadfast on long-term perspectives, fundamental principles, and strategic investments in technologies that align with Citi’s operational interests, as he articulated in a recent dialogue with American Banker.
“What is the company’s purpose? Does it deliver genuine value? Is its business model robust and sustainable?”
Purushotham reflected, emphasizing the challenges of maintaining such clarity amidst rapid changes and trends in the marketplace.
Citi Ventures has made over 200 investments since its inception in 2010, maintaining an active portfolio of 125 companies, which includes both AI and non-AI sectors.
In this edited interview, Purushotham elucidated his evaluative criteria for potential investments, identified promising AI applications, and discussed the underwhelming adoption of robotic process automation (RPA).
Recently, you shared an article on computer use agents on LinkedIn, seen by many as the evolution of robotic process automation (RPA)—which typically employs basic, non-AI “bots” to replicate human actions and streamline repetitive tasks.
RPA was once heralded as a groundbreaking innovation, but the term has nearly vanished. Do you believe computer use agents represent the next significant milestone?
ARVIND PURUSHOTHAM: This subject is currently on the minds of numerous venture capitalists and enterprise chief information officers.
When reflecting on RPA’s initial allure to enterprises, particularly within financial services, it’s primarily due to the extensive knowledge work necessitating human oversight, such as maker-checking and various manual processing tasks.
In sectors where extensive manual labor is prevalent, such as finance and insurance, this technology can genuinely enhance productivity.
What are the reasons RPA failed to fully materialize?
The consistent feedback has highlighted sustainability challenges when implementing RPA within enterprises. Operating in dynamically changing enterprise environments proved RPA to be somewhat fragile; systems require the agility to adapt to new conditions.
Without this adaptability, these agents may falter, necessitating maintenance that ultimately adds to workloads. From an industry perspective, RPA’s first iteration encountered significant hurdles.
However, as we look to 2026, the emergence of reasoning agents capable of interpreting tasks, devising plans, and executing individual steps marks a substantial evolution.
Evidence of this advancement is apparent not only in personal tasks but also within enterprise frameworks.
The pivotal inquiry remains: Can these computer use agents deliver notable productivity enhancements and operate reliably in an enterprise context?
Citi has been exploring agents extensively, and with the advent of cutting-edge models, there lies a promising opportunity to realize significant productivity enhancements across various knowledge-driven tasks.
Currently, initial productivity gains are predominantly observable in coding.
Citi Ventures has invested in AI-powered lending software providers. What draws your attention to this sector? Are lenders increasingly seeking AI-driven underwriting solutions?
Absolutely. Major banks engage in diverse forms of consumer lending—including credit cards, personal loans, secured lending like mortgages, and other institutional lending for corporate clients.
When discussions arise around AI in lending, the focus often gravitates towards underwriting, yet underwriting comprises only one element of a complex process involved in lending.
For new clients, extensive onboarding, information gathering, application compliance, underwriting, servicing, and taxation reporting all play critical roles.
This extensive workload encompasses aspects like identifying fraudulent applications and conducting KYC due diligence, which can be labor-intensive but essential in providing superior customer experiences.
AI can enhance each of these phases. It has applications ranging from fraud detection and customer due diligence to effective servicing and client support.
Citi Ventures is actively pursuing investments across these critical operational areas, recognizing the immense size of lending within banking.
Regarding underwriting specifically, consider how the landscape has evolved: Historical lending practices hinged on rigid rules. Assessments were based on parameters such as cash flow profiles or FICO scores.
In today’s context, a hybrid approach emerges—integrating established rules and leveraging artificial intelligence, including machine learning models and contemporary large language models (LLMs).
These innovations facilitate real-time decision-making, enabling a more nuanced understanding of loan applicants and potentially broadening credit access for individuals who previously might not have qualified while adhering to regulatory standards.
In consumer lending, AI can incorporate alternative data such as utility payments or rental histories into underwriting models for individuals lacking robust credit histories.
What parallels exist in corporate lending? Is it primarily cash flow metrics or corporate background information?
The focus lies in leveraging real-time company data. AI excels in discerning patterns often imperceptible to human analysis.
Still, banks must uphold responsible lending and ethical AI usage as foundational principles. The integration of AI within financial services is inherently more measured due to regulatory considerations, necessitating a meticulous implementation approach for both internal functions and client-facing applications.
While you highlighted AI in lending and coding as promising avenues, are there other AI use cases in finance currently capturing your interest?
I’ve previously mentioned coding, which represents a vital component as software-driven tasks proliferate across sectors, including finance.
The extent to which coding agents enhance productivity is profound, influencing not just efficiency but also driving innovation.
This accelerated capacity for introducing new products and enhancing customer experiences leads to higher net promoter scores, underscoring the broader impact beyond mere labor savings.
Yet, the arena of general knowledge work remains an area with untapped potential, where Citi is proactively engaged in the adoption of AI technologies, in parallel with trends across various industries.
Concerns have emerged regarding a potential bubble in AI investments that could precipitate market disruptions. What is your perspective on this issue?
Many liken it to past cycles, such as the internet boom of the late 1990s. When evaluating investment strategies regarding valuations and anticipated returns, we prioritize a long-term viewpoint, acknowledging the volatility associated with weekly shifts in market sentiment.
We emphasize the importance of fundamental assessments—what value does a company deliver? Is its business model viable?
Though these inquiries may sound simplistic, they become complex in the midst of disruptive trends, where underlying principles may be overshadowed.
Furthermore:
We consistently evaluate how investments align with Citi’s strategic objectives. As a venture investment group, it is crucial to ensure that our investments not only align with the bank’s technology or functional areas but also indicate sustainability.
When a prominent entity like Citi identifies a strategic alignment, it bodes well for the ongoing viability of a business model.
Lastly, we consider valuation potential—specifically, the prospect for returns. This aspect leans more towards artistry than science, particularly in early-stage ventures.
Uniquely, this era is characterized by remarkable revenue growth among several AI companies, many of which reside within our portfolio, as well as within the broader private sector.
While rapid revenue increases are promising, it remains essential to scrutinize gross margins and other vital factors.
Source link: Americanbanker.com.
