AI, Trust, and the Underserved - Interview with Paula Grieco, SVP at Commonwealth

Paula Grieco is Senior Vice President at Commonwealth.

Discover top fintech news and events!

Subscribe to FinTech Weekly’s newsletter

Read by executives at JP Morgan, Coinbase, Blackrock, Klarna and more

Financial AI has a long way to go — not just in terms of speed, accuracy, or even regulation, but in the way it earns trust. Especially from those who haven’t traditionally been the first in line when new tech rolls out.

At FinTech Weekly, we’ve been following the work of Commonwealth, a nonprofit focused on building financial security for low- and moderate-income (LMI) households. Their fieldwork, explored in our recent editorial, revealed a clear tension: while LMI users are open to tools like chatbots, they’re still waiting for experiences that actually serve them — not just repackaged features built for someone else.

This week, we went deeper.

We spoke with Paula Grieco, Senior Vice President at Commonwealth, to understand what’s really needed to make AI effective — and safe — for underserved communities. From design principles to earned trust, from co-pilots to chatbot fatigue, she shares why intention matters more than innovation alone.

It’s a grounded and thoughtful view of what inclusive financial technology could — and should — look like.

Read the full interview below.

2.  Commonwealth's recent collaboration with JPMorganChase provided key insights into AI’s role in improving financial security for LMI households. What were the most surprising or impactful findings from this research?

Our research illuminates the immense potential of AI, specifically chatbots, to provide personalized guidance and support to communities living on lower incomes —  if the chatbots are designed thoughtfully with the needs and perspectives of this group in mind.

Two key findings:

*   Customers largely view chatbots as positive tools for improving their financial well-being. Our data showed that 57% of respondents said using chatbots improved their financial situation. The research also showed people earning low to moderate income (LMI) want credit-building, budgeting, and debt management features.

*   Respondents valued the judgment-free space with a chatbot for asking sensitive financial questions without worrying about the shame or self-consciousness that might accompany that same conversation happening face-to-face with a human representative.

3.  How do you see conversational AI evolving in the financial services space, particularly for underserved communities?

Ideally, the next generation of chatbots fueled by generative AI will be AI financial assistants that better support the financial activities of these households, and earn trust with populations who are often wary of engagement with the financial system and of sharing data online. There is a major opportunity for financial service providers to provide more complex, nuanced, and action-oriented capabilities for their chatbots.

When customers use financial chatbots now, they are primarily seeking account information or trying to resolve a problem. Less than 20% of our national survey respondents had used chatbots for financial advice and education, product recommendations, applying for credit or loans, and opening or closing accounts. However, our research finds there is a demand for chatbots that can assist with these types of banking actions. Focusing on these types of features when developing chatbots may increase their use and usefulness among these customers.

For banks and financial institutions that are not ready to launch generative AI financial co-pilots directly to consumers, this technology can support bank employees such as customer representatives to provide better, more accurate and more timely responses to customers during interactions.

4.  What are some of the biggest challenges in ensuring AI-driven financial tools are equitable and effective for Black, Latinx, and women-led households?

With all emerging technologies, there needs to be an intentional effort made to ensure that the needs of those earning low to moderate income are included in the development process and design decisions. We’ve found that a private/philanthropic partnership with financial institutions early on helps to build momentum for these efforts. By growing an evidence base, we also help build the business case.

We’ve seen significant potential for design guidance around things like increasing earned trust that can enable conversational AI to support financial health without major cost increases.

5.  Based on your research, what are the key design principles financial service providers should consider when integrating AI to support LMI users?

Commonwealth has created a resource, the Financial AI for Good Guide, to provide actionable design guidance to financial service providers who serve LMI populations. We developed these recommendations based on comprehensive research with financial institutions, chatbot providers, and people living on LMI.

The guide is organized around four primary design goals. I’ll give you an example or two for each:

2.  Earn trust: The top concern among the majority of our research respondents when using a chatbot was security. This is where financial institutions can emphasize data security through upfront messaging about the measures the bank is taking as well as give users control over what data is stored.
4.  Drive engagement: Make the experience one in which users know what these tools can do for them and when by creating clarity around their functions. Also, pursue “intelligent proactivity.” For instance, integrate chatbots that pop up when they can be most useful, without being too pushy or aggressive, which can come off as spammy.
6.  Increase value: Anticipate the needs of your customer. Limited access to physical branches creates an opportunity for chatbots to complete small actions that customers may have had to go to the bank for. Balance automation and control by allowing users to turn the automated finance features on and off, and include “safety net” features that pause the automated movement of money if a balance falls below a certain threshold.
8.  Improve accessibility: Offer multilingual support and guidance matched to your customer base and focus on mobile-friendly features. Our research showed that more than half of all respondents preferred to access their banking through their mobile phones.

6.  Can you share any success stories or case studies where conversational AI has significantly improved financial well-being for LMI individuals?

What we know is that 57% of users in our field test study indicated that using a financial chatbot had a positive impact on their financial situation. While these early results are promising, generative AI tools are still in their infancy, and our ongoing research will continue to build an evidence base on their effectiveness in improving financial well-being for LMI individuals.

7.  What risks or unintended consequences should financial institutions be mindful of when implementing AI-driven financial tools?

What’s important is that people earning LMI are not left out of the equation. When financial institutions are developing tools, it’s important that they understand the inherent opportunities and ways to serve the LMI customer base.

There are many bodies focused specifically on the inherent risks and consequences of AI- driven tools, and the bias and accuracy of large language models. Beyond that, we want to ensure that a primary concern is addressed: the relevance of financial recommendations for users’ individual financial situations. Financial institutions can increase customer engagement and earn their customers’ trust by ensuring that the information they’re providing is accurate and that there is real transparency.

AI presents an unprecedented opportunity for people earning LMI to access advice and tools that have traditionally not been available to them, whether it be investing tools or personal finance management. These tools can be personalized and customized for people earning LMI and their unique situations. This is a tremendous opportunity for financial providers to grow their customer base.

8.  How can financial institutions measure the real impact of AI-driven tools on users’ financial security and well-being?

The financial wellness fundamentals:  Is there an increase in savings, a reduction in debt, an improvement in credit scores when using these tools?

We can also survey the experience around interacting with the chatbot — has trust increased? Is there an increased interest in products that would be helpful for improving financial well-being? When it comes to advising, were there actions taken after receiving advice?

Banks can also do A/B testing among different groups of consumers that are interacting with chatbots vs. those who aren’t to see if there is a measurable difference between them.

9.  What role does human oversight play in the deployment of AI tools for financial services, and how can providers strike the right balance between automation and human support?

One of the ways to increase earned trust around AI is to ensure that there is a human accessible at the right times around the interaction.  This is where the use of co-pilots by customer-facing bank employees can be beneficial. Access to a live human when needed increases the trust and experience with the AI tool.

Using conversational AI will allow customer service representatives to better and more rapidly serve the complex needs of their customers and members while providing the human touch at key points in the interaction when a live agent is desirable.

Transparency is also critical to build trust in any interaction. You should know, for instance, whether you are speaking with a chatbot or with a real person.

10.  Looking ahead, what are the most exciting opportunities for AI in financial inclusion over the next five years?

Generative AI represents the next evolution in conversational AI support offering personalized and context-sensitive engagement at a level that much more closely approximates human support than the decision-tree structure of most financial chatbots today. Initial applications of generative AI in finance have focused primarily on back-office applications, where there is opportunity to support customer service agents. Identifying how generative AI can provide personalized support at scale in a financial context is a key opportunity to drive development in this sector.

Earned trust-building will be particularly critical for broader adoption of generative AI, which participants in our field tests and focus groups remain more skeptical of than traditional chatbots. Still, the potential benefits of providing a more advanced level of support across financial service applications make generative AI the most exciting technology to watch in the financial sector. Those who can develop trusted and reliable generative AI support will be at the frontier of this new era of customer relationship building at scale.

Some other specific opportunities we see are co-pilots and personal assistants that can provide comprehensive financial guidance tailored to individual needs, a personal financial coach if you will. We also expect advances in conversational AI to play a valuable role in promoting workers’ financial health by providing information and guidance to navigate complex employee benefit systems.

11.  How do you see the role of nonprofits like Commonwealth evolving in shaping the responsible use of AI in financial services?

Historically, the design of new technologies has focused on adoption by upper-income consumers while overlooking the needs of LMI households. Through our Emerging Tech for All (ETA) initiative, we are focused on making sure that the needs of financially vulnerable people are understood, visible, introduced into relevant conversations, and integrated into solutions. We’re at a critical inflection point in scaling AI, and believe it is urgent to continue to research and identify the ways in which AI can positively impact this population.

Relatively little research and adoption in the field exists on this topic today, and some providers we interviewed cited the need for larger-scale studies to build the kind of evidence they could use to make the case for this kind of design internally. We are rising to this challenge by producing impactful research and on-the-ground field tests that demonstrate how generative AI can support the financial wellness of households living on LMI and makes the business case for more actively designing for this underserved consumer segment.

Looking to the future, the systemic impact of inclusive tech design will depend on scale applications of these insights by major actors in financial services. For us, moving inclusive design to scale will depend on leveraging our research to partner with larger organizations seeking to capitalize on advances in AI to support the financial health of their customers and workers.

12.  What advice would you give to financial institutions looking to leverage AI while maintaining trust and transparency with their customers?

LMI households are more interested in banking directly with a person yet have the least access to in-person branches. This gap highlights a key opportunity for AI to provide the kind of personalized support that households living on LMI seek without needing to increase the number of branches or customer support staff.

However, to drive wider adoption, financial institutions must earn and build more trust in chatbots from people earning LMI — some of this is specific to the chatbot experience, while some is industry-wide as AI technology gains more acceptance and improves in overall security and quality.

The top concerns for people engaging with chatbots are security and privacy.  In general, people have expressed a lack of trust in conversational AI to be helpful, to protect their data, or to act in their best interests. While many in the business world are excited about the potential of AI, people living on LMI likely view it with more skepticism as a new technology that has yet to demonstrate its direct value to them.

Transparent data policies, reassuring branding and messaging, and maintaining the connection to a human agent as a backup option will all help toward building and earning trust. Developing useful and personalized interactions through generative AI that move beyond providing the basic information offered by chatbots today, such as account balances and recent transactions, will also help demonstrate the technology’s value.

It’s also important to emphasize the concept of earned trust. The goal is not merely to convince people to trust chatbots, but to actually design chatbots in such a way that this trust is warranted.