How AI and Human-Centered Design Can Transform Financial Inclusion in Colombia

María lives in a small rural community in Colombia. A few years ago, she opened a savings account through a financial inclusion program in her region. However, like most Colombians in rural areas, she does not actively use her account. As of June 2023, only 56% of adults in dispersed rural municipalities had access to at least one financial product, and the usage rate was even lower at 44%. 

In today’s fast-paced digital financial landscape, technology has the potential to drive positive financial behaviors and improve economic stability. However, the rapid digitalization of financial services risks deepening financial exclusion, particularly among low-income populations who may lack the digital literacy or confidence to navigate these platforms. Without personalized guidance, many face barriers to engaging with the formal financial system, limiting their ability to build financial security and long-term economic resilience.

Recognizing this challenge, DataKind and Fundación Capital, with support from the GitLab Foundation’s AI for Economic Opportunity Fund, launched an innovative initiative to strengthen savings habits among Colombia's most underserved communities. The pilot, in partnership with Fundación Capital’s Tierra de Oportunidades program, primarily served women, 70% of whom were Venezuelan refugees, migrants, or Colombian returnees.

By integrating AI-driven personalization with human-centered design, the project demonstrated how combining digital tools (“tech”) with personalized support (“touch”) can drive financial inclusion, improve user engagement and foster lasting behavior change.  

Using machine learning to generate personalized financial recommendations 

In 2024, the two organizations collaborated to integrate and optimize the user journey of two distinct flagship products:

• LISTA, a modular digital financial education platform for smartphones and tablets.
• Con-Héctor, a virtual assistant that provides information on products, services, and other relevant content related to financial inclusion. 

By combining these solutions, we aimed to create a seamless, data-driven ecosystem that enhanced participants’ financial knowledge and fostered long-term savings behavior. 

Machine learning helped us generate personalized financial recommendations based on inputs into LISTA, where users had identified their personal financial goals. The AI model first assessed goal feasibility based on market value, goal type and savings timeline. It then delivered tailored recommendations through the Con-Héctor chatbot. These AI-driven insights aimed to reinforce savings habits and provide actionable guidance to improve financial decision-making. 

With these recommendations in hand, users could then refine their financial and savings strategies. At this stage, we also provided information about some available financial products so that users could explore banking options and decide whether to connect with partner institutions to open savings accounts and digital wallets.

Overcoming barriers to adoption 

The project was not without its challenges. Using a human-centered design process of iterating and testing adjustments to the products and implementation strategies was key to helping us meet the needs of the population. 

For example, we found that some users were hesitant to participate at first, and that low digital literacy was a barrier. So we engaged field coordinators and local leaders to support onboarding and address concerns about digital security. Field staff played a critical role in building trust and providing hands-on training, which significantly improved adoption. This hands-on approach enabled the project to be successfully integrated into community activities. 

Another issue was connectivity constraints. By integrating LISTA and Con-Héctor through WhatsApp, we were able to facilitate access in low-connectivity areas, although offline resources and scheduled follow-ups were sometimes needed to maintain engagement. 

A key element of the pilot involved a feedback mechanism within Con-Héctor to gather user insights and refine future iterations. By addressing these challenges through a human-centered design process, the team developed a model for future scaling. 

The impact of personalized messaging

In addition to the quantitative findings displayed above, qualitative feedback further highlighted the impact of personalized messaging. Users reported greater awareness of financial goals, better budgeting techniques, increased confidence in their ability to save, and a more consistent savings approach. 

"After using LISTA, I did a quick self-analysis and recognized some mistakes in my savings habits. LISTA helped me learn a better way to save, and now I’m putting into practice what I learned." 

"The messages in CH showed me what I was doing well and where I needed to improve my savings. I would like to receive more information or advice because it will help me improve, clarify doubts, and give me options to make better financial decisions." 

These testimonials underscore the effectiveness of integrating AI-driven insights with financial education to encourage meaningful behavior change.

Looking ahead: Scaling AI-driven financial inclusion 

The seamless integration of LISTA and Con-Héctor showed how digital tools can work together to deliver personalized and practical financial education.  However, tech alone isn’t enough—human-centered approaches remain essential for ensuring digital financial inclusion leaves no one behind. The collaboration between DataKind and Fundación Capital highlighted the potential impact of AI-driven financial education in promoting sustainable savings habits. The pilot phase has been successfully concluded, and we are actively seeking new opportunities for collaboration to expand upon the insights gained.