AI Chabots: Education Counseling in Latin America and the Caribbean (LAC)
- United States
- Nonprofit
- Provide the skills that people need to thrive in both their community and a complex world, including social-emotional competencies, problem-solving, and literacy around new technologies such as AI.
- 1. No Poverty
- 4. Quality Education
- 5. Gender Equality
- 10. Reduced Inequalities
- Pilot
- Tested in Colombia in a specific type of intervention
- Research on the impact is still building but the preliminary results are promising
- Adding since 2023 the use of gen AI -- this is new and we need to pilot.
- Build capacity to stablish a sustainable plan to make the org grow organically and sustainable fashion
- Learn from peers how to establish continuity of project implementation through changes in political leaders that affect the educational leaders
- Capacity to streamline our technologies
- Legal advice to grow our org in other countries and be able to partner with local orgs that can pay part of the cost of implementation
-
- Business Model (e.g. product-market fit, strategy & development)
- Human Capital (e.g. sourcing talent, board development)
- Public Relations (e.g. branding/marketing strategy, social and global media)
A Multifaceted Approach
The education counseling landscape, particularly in LAC, faces a challenge: while counseling needs vary greatly by country and context, the technology to deliver personalized guidance has only recently become widely available. ConsiliumBots tackles this challenge with a unique and innovative approach that merges three key elements:
1. Leveraging Cutting-Edge Generative AI: The emergence of powerful generative AI tools like OpenAI's GPT-4 offers a game-changer. We've recently formed a partnership with OpenAI, granting us access to their solutions and credits specifically for developing new use cases and fostering innovation within our educational guidance chatbot. This ongoing collaboration allows us to continuously push the boundaries of what AI-powered guidance can achieve.
2. Bridging the Gap Between Technology and Education: Our extensive network of connections with education authorities across LAC sets us apart. ConsiliumBots, along with its CEO's established expertise in regional education information, positions us perfectly to deliver the solution through official channels. We partner with Ministries or Secretariats of Education, the government entities responsible for the education system. These institutions often possess student contact information, allowing for targeted outreach and information dissemination.
3. Deep Subject Knowledge and Data Expertise: Our team boasts in-depth knowledge of the higher education landscape in LAC. Years of experience working within this sphere have equipped us with subject-matter expertise. This includes familiarity with key stakeholder networks and relevant databases. We understand the decentralized nature of educational information in many countries and have taken the initiative to build comprehensive databases. We combine data sets from various sources, including employment rates and graduation statistics, often held by different ministries. Our expertise lies in seamlessly integrating these disparate data points to create a holistic knowledge base.
This multifaceted approach sets ConsiliumBots' solution apart. We don't just offer innovative technology; we bridge the gap between powerful AI and the specific needs of students in LAC. By efficiently delivering personalized guidance through established channels, we aim to create a ripple effect. Our success could inspire others in the education technology space to adopt similar innovative approaches, ultimately leading to a more impactful and accessible education landscape for students across the globe.
Finally, it is important to note that digitalization and the use of technologies still remains a problem in some more disconnected areas of larger cities. Given the above, our effort to bring these technologies to all places, without distinction, allows us to sow seeds of a digitalization process that we hope in the future can attract various projects or programs that facilitate teaching, learning, and the success of educational communities.
ConsiliumBots' chatbot solution tackles the challenge of limited access to high-quality educational guidance in LAC. The Theory of Action behind it is: If students interact with our chatbot and gain increased knowledge about higher education options, reduced misconceptions, and make informed decisions (applying for scholarships, exploring new options), then they will experience improved graduation rates, increased likelihood of finding career-aligned jobs, and reduced financial burdens (medium-term). As a result, students will have improved well-being, reduced anxiety, and greater confidence in their chosen paths (long-term).
Here's how we expect it to make a positive impact, step-by-step:
1. Increased Knowledge and Reduced Misconceptions: By interacting with our chatbot, students gain greater awareness of realistic higher education options available to them. The chatbot helps dispel common myths (both overly optimistic and overly pessimistic views) and provides accurate information about potential career paths, financial aid, and financial returns associated with different programs. Research on our previous pilots, less-conversational chatbots has already shown demonstrable evidence of this learning process.
2. Informed Decision-Making: Equipped with this knowledge, students are empowered to make more informed decisions about their higher education applications. This may involve applying for scholarships and benefits on time (as documented in our 2021 pilot in Colombia), exploring previously unconsidered options, and developing realistic salary expectations. Evidence of these effects are recorded by higher education recommendations in Chile. During the 2021 and 2022 centralized application to higher education in Chile, with ConsiliumBots’ support, Fabre et al. (2023) designed and implemented a large-scale outreach policy to reduce application mistakes. The authors find that showing personalized admission probabilities and non-assignment risk has a causal effect on improving students’ outcomes, significantly reducing the risk of not being assigned to the centralized system and the incidence of over-confidence mistakes. In this sense, policy interventions that reduce frictions of the process can significantly promote access to higher education.
3. Improved Graduation Rates and Job Placement: In the medium term, better-informed decisions are expected to positively impact graduation rates. Students are more likely to finish their chosen programs on time, reducing the financial burden of higher education. In the long run, these students are more likely to find careers aligned with their chosen programs, leading to greater career satisfaction and financial stability. For instance, Bettinger et al., 2012 finds that assisting students during the application process can substantially increase completion rates for financial aid, which leads to an improved likelihood of college attendance and persistence.
4. Beyond Financial Benefits: Our solution offers intangible benefits like well-being and confidence. The chatbot acts as a virtual counselor, alleviating anxiety and confusion often experienced during this critical life stage. This translates to improved well-being in the short term. Additionally, a well-considered decision can lead students to begin their higher education journey with greater peace of mind and a clear sense of direction, positively impacting their mental health alongside financial gains.
Our impact goals are the following:
To increase our participation rate (or take-up rate), i.e. the percentage of students interacting with the tool, to 50% in Colombia and at least 30% in the Dominican Republic. Indicator: take-up rate
Although participation is important, the number of interactions is extremely valuable in knowing the extent to which students received information that could be valuable for better decision making. Therefore, in each intervention, we try to evaluate which aspects or elements are crucial to improve this indicator. For example, in Colombia we aim to increase the number of interactions by 15% and in the Dominican Republic we aim to reach an average of 12 interactions when this tool is introduced. Indicator: Average number of messages sent by a student in a conversation with the chatbot.
Student evaluation of this tool is crucial. If we fail to build a tool that is friendly, easy to use, with good content, and that allows the student to have an excellent experience, then the participation rate and the average number of interactions will be lower than expected. With this in mind, we aim for our tool to be rated at least a 4 on a scale of 1 to 5. Indicator: level of satisfaction with the tool.
Increase application to higher education benefits and financing options by 10%. Indicator: rate of students that apply for benefits or loans. Measured by ICETEX data in Colombia. In the Dominican Republic most students have financial aid or heavily subsidized tuition, therefore this indicator is less relevant.
Increase inmediate access to higher education by 5%. Indicator: rate of students who leave school and immediately enter higher education (university, technical, or other).
Increase retention after the first year of college enrollment by 5%. Indicator: Number of students who remain enrolled in the same program for the second year.
The technology we use to power our guidance chatbot includes the following:
User Interface: WhatsApp
We leverage WhatsApp, the most widely used messaging service in LAC. Accessible and often free of data charges in many countries, WhatsApp allows students to interact with the chatbot seamlessly within their daily lives. Additionally, utilizing this established platform eliminates the need for us to develop and maintain a separate user interface.
In-House Chatbot Studio
Our team created a custom chatbot development platform. This studio functions as a workspace for designing the chatbot's conversation flow. Here, we define conversation nodes (decision points), rules connecting user inputs to chatbot outputs, and the overall chatbot structure. This versatile platform supports the creation of multi-channel chatbots, allowing connections to various APIs and libraries. Within this studio, we design, develop, test, and finally launch the chatbot.
Key Capabilities of the Chatbot Studio:
Flow Design: Enables development of structured, semi-structured, and unstructured conversation flows using nodes and edges.
Targeted Interventions: Administrators can design interventions, sending the chatbot to specific user groups based on defined characteristics.
Usage Monitoring: Administrators can monitor chatbot usage and track conversation flow.
Rich Content Integration: Administrators can upload resources like PDFs and data tables for association with different conversation nodes.
Multi-Channel Support: Users can interact through various channels, including platforms embedded in web applications and WhatsApp.
Comprehensive Testing: The platform allows testing the chatbot from any specific node onwards.
OpenAI Integration: All chatbots leverage OpenAI capabilities such as fuzzy matching and relevancy evaluation, with the potential to integrate further functionalities.
Personalized Interactions: User-specific variables can be used to personalize the conversation flow.
Twilio: WhatsApp Connection
We connect to WhatsApp through Twilio, allowing us to monitor our WhatsApp numbers on their dashboards. Additionally, we have secured WhatsApp numbers with high daily message sending capacity.
Data Science: Powering Personalization and Knowledge
Our data science expertise facilitates the processing and organization of large educational data sets. We build databases that structure decentralized information about higher education systems. This information is then made readily available for chatbot consumption. Data science also allows us to incorporate student-provided information, both pre-populated and user-generated, to personalize recommendations and enhance the user experience.
OpenAI: Core Language Processing
OpenAI's GPT-4 language model forms the foundation of our solution, driving its Natural Language Processing (NLP) and text generation capabilities. NLP techniques such as named entity recognition and sentiment analysis enable the chatbot to understand the nuances of student inquiries. Additionally, embedding models capture semantic relationships between concepts. This allows the chatbot to identify relevant information within its vast knowledge base, which is constantly enriched by user interactions and real-time data analysis for continued accuracy. We've currently implemented 12 OpenAI API use cases, with 5 additional potential applications identified.
In conclusion, our WhatsApp-integrated, AI-powered chatbot leverages OpenAI's GPT-4 to offer personalized, accessible educational guidance. This innovative tool transforms how students in LAC navigate the complexities of higher education, empowering them to make informed decisions with confidence.
- A new application of an existing technology
- Artificial Intelligence / Machine Learning
- Crowd Sourced Service / Social Networks
- Software and Mobile Applications
- Colombia
- Chile
- Dominican Republic
Part-time staff: CEO and COO that devote a fraction of their time to the project development and vision. Designer time to generate branding and communication materials.
Full-time staff: 6 people; 1 Head Project Manager, 1 Project Manager in Colombia, 1 Project Manager in the Dominican Republic, 2 Data Analysts and 1 Senior Full-stack Software Engineer.
Contractors: As needed, to conduct QA, user testing, and focus groups.
Research-related staff: research assistants and Principal Investigators will work on the project to design any randomized evaluation within the implementation and evaluate the outcomes.
ConsiliumBots has worked on this solution for 6 years, since 2018, making progressive improvements.
Through the years we have developed not only IcfesBot - the higher education chatbot in Colombia- but also other guidance solutions that have taught us important lessons on giving feedback on key decisive moments in education. For instance, school admission feedback in Perú and Chile; high school selection in Spirito Santo, Brazil; teacher assignment in Ecuador; and higher education enrollment in Chile.
In alignment with MIT Solve's commitment to diversity, equity, and inclusion (DEI), our organization prioritizes creating an environment where all team members feel valued, supported, and included. We recognize that diversity encompasses the social, cultural, and identity-based attributes of individuals, while equity involves providing access to opportunities for all, and inclusion ensures that every team member is welcomed and respected.
Organizational Structure and Policies: Our organization operates on a horizontal structure, fostering an open and collaborative environment where every team member, regardless of their role, is approachable. We maintain an open-door policy that encourages transparency and open communication across all levels of the organization. This structure not only promotes inclusivity but also ensures that all voices are heard during decision-making processes, thereby enriching our solutions with diverse perspectives. We are continuously working towards creating a workplace where everyone has the opportunity to thrive. We celebrate team members' achievements and support their ongoing professional development. Many of our collaborators have leveraged this support to pursue graduate studies at prestigious universities, furthering their ability to contribute to our mission.
Team Dynamics and Support: We emphasize building a cohesive team dynamic where support and mutual respect are foundational. Our team members are encouraged to support each other in reaching individual and collective goals, creating a strong sense of community within the organization. This supportive environment ensures that each team member feels secure and valued, fostering a collective drive towards innovation and excellence.
Inclusive Recruitment and Talent Acquisition: In our recruitment practices, we strive to hire the best talent suited for the challenges we face, ensuring a cultural fit that respects and enhances our diverse community. We have a strong non-discrimination policy that supports our commitment to diversity and equity. By hiring local talent, we not only support the local economy but also enrich our team with individuals who bring intrinsic understanding of the cultural nuances necessary for the effective deployment of our solutions.
Feedback and Continuous Improvement: We actively seek and incorporate feedback from our team to refine our DEI strategies.
We provide information and counseling to high school graduates during the last year of school through WhatsApp. Key beneficiaries are young people -between 16 and 18 years old- who are in the process of defining which higher education program to apply for. Therefore, they need the information to make a better-informed decision and will be interested in interacting with the chatbot. The chatbot is on WhatsApp, a communication App they have installed on their phones and use regularly. Most young beneficiaries own a smartphone and are used to searching for things on the internet and most likely, interact with technology in conversational ways.
To provide the chatbot to students, we partner with the relevant education authority in the country. In every country, the government at some level has centralized information on students graduating from high school who could potentially access higher education. By partnering with them, we leverage the contact information they usually request from students and the consent they receive to send them information about the admission process.
There are two ways to provide beneficiaries with the chatbot. First, proactively sending a nudge on WhatsApp to initiate the conversation. Second, by making the chatbot's WhatsApp information available for students on websites or places they visit (i.e. official education websites, and schools). In Colombia, thanks to having the contact information we have used the direct contact method, which is much more effective in reaching students, especially those less aware of resources for the admission process.
The business model to sustain this information provision is through governments. The government wants the students to make better-informed decisions benefit their constituencies directly, and potentially reduce costs on higher education subsidies or other that students that drop out or are unemployed generate. As well, the government benefits from other of their programs being promoted through this communication channel, and improving their performance. For instance, in Bogotá, our project recommended different career guidance tools the local government had in place. These political gains are critical as the decision makers to implement this kind of solution tend to consider short-term gains according to their administration period.
- Individual consumers or stakeholders (B2C)
To date, our funding has relied primarily on research grants and philanthropic funding, with significant support secured in recent years. In 2023, we received $200,000 from The Agency Fund and a $110,000 contract with Harvard University. In addition, we are in the final stages of securing $360,000 from another US philanthropic fund, pending final review.
We supplement these funds with general operating budgets, smaller research grants and in-kind contributions such as research assistant time. A notable partnership with OpenAI provides us with credits to cover our anticipated use of generative AI technology, demonstrating a successful model of leveraging corporate partnerships for technology access and showcasing social good applications.
Our revenue model is evolving to ensure long-term sustainability. We expect fixed costs to decrease as our product matures, allowing us to offer our chatbot solutions, such as the OsoBOT for Bogotá and the national-level IcfesBot, to new government or institutional clients with minimal customization. This scalability has proven effective, as demonstrated by our ability to quickly repurpose the core structure of our chatbots for different use cases.
Looking ahead, we aim to establish service contracts with governments that not only cover the direct costs associated with WhatsApp messaging, but also contribute to a steady revenue stream. This approach should move us towards financial sustainability by reducing our reliance on fluctuating grant cycles.
Overall, our financial strategy combines initial grant funding with a transition to recurring revenue through service contracts and strategic partnerships. This mix allows us to sustain and expand our initiatives while exploring new opportunities for growth and impact.

Chief Operating Officer