Doowii - your AI-first education platform 📖

Plus: Doowii CEO Ben Dodson on revolutionizing the way schools and universities use data...

January 14, 2025

CV Deep Dive

Today, we’re talking with Benjamin Dodson, Co-Founder and CEO of Doowii.

Doowii is an AI-first data platform focused on simplifying data analytics for educators and administrators. It enables non-technical users at academic institutions to retrieve, analyze, and visualize data using natural language, providing tools to support administrators and students with better decision-making and resource allocation.

Today, Doowii serves both K-12 and higher education, with early adoption primarily among school administrators who benefit from being able to streamline data-related tasks. It packs features like text-to-SQL and dynamic context management to help make these processes more efficient and provide deeper insights into academic performance, helping users address needs like student success and retention across schools and colleges.

In this conversation, Ben discusses the origins of Doowii, the opportunity in building an AI-driven education platform, and how it’s revolutionizing the way schools and universities use data.

Let’s dive in ⚡️

Read time: 8 mins

Our Chat with Ben 💬

Ben - welcome to Cerebral Valley! First off, give us a bit about your background and what led you to co-found Doowii? 

Hey! My name is Ben and I’m the CEO and co-founder of Doowii. My background is primarily as an AI/ML engineer for about the past decade. I’ve worked at Bay Area tech companies like Google, Snapchat, and Mux—big companies, medium-sized companies, and startups—always doing something related to AI/ML. A lot of the work I was doing was technically challenging and interesting, but I felt like it didn’t have the kind of impact I was looking for. I wanted to do something more value-driven. Not just impact in terms of affecting lots of people, because some of the things I worked on certainly did that, but something more tied to mission and values.

I care a lot about education, so I started thinking about whether I could bring some of the progress I was seeing in AI—not just in the past few years with generative AI, but over the last decade in data science—to K-12 and higher education. About two years ago, I decided to start an ed-tech company. The idea is to create an AI data scientist for educators. Schools cannot  invest the millions of dollars that big tech companies do to get value from their data, so the goal is to build a data platform that gives them those benefits at a fraction of the cost.  Having seen the sophisticated tools available in tech, I believe we have an obligation to bring similar capabilities to education – helping every student reach their full potential.

How would you describe Doowii to an AI engineer or individual who’s slightly less familiar with what you do? 

The idea is to think about what a data scientist does at a big tech company—they’re building data retrieval pipelines, doing data engineering, performing analysis, creating dashboards, and working on machine learning. We're bringing all of these functions together in a way that’s accessible to anyone through a natural language interface.  A layperson with no technical expertise can get the data they need, run analyses, and have those analyses explained to them in the same way a data scientist would.

The key difference is that we’re vertically focused. There are plenty of platforms integrating natural language into data analysis, but being able to do it accurately while understanding the context (e.g. human-level definitions) of the data is part of the unique value a data scientist provides. Doowii is building that same level of context and data expertise into a platform, allowing non-technical users to effectively leverage powerful data science tools at scale.

Talk to us about your users today - who’s finding the most value in what you’re building with Doowii? Any use-cases that are the most prominent to date? 

Our vision is to make it easier for anyone in education to leverage data, unlocking capabilities that lead to better decision-making, resource allocation, and ultimately improved student outcomes. Right now, we’re primarily working with administrators because they typically have the most access to data and already interact with it as part of their daily work.

Administrators in both K-12 and higher education are our first users and are seeing the most immediate value. For example, they’re going from having a question and waiting a week for someone to run a report to now getting that answer in 30 seconds. That’s where the fastest impact has been so far. We’re working to expand access—moving to faculty, professors, and teachers, and eventually to students and parents. But starting with administrators has made the most sense as an entry point.

Are there any specific use-cases that you’d like to highlight? How are you measuring the impact you're having on your key users so far? 

There are two broad buckets where the platform adds value. One is existing reports that take a long time to run. For example, administrators who used to spend an hour updating a daily report from multiple sources can now have that process take place in just a few minutes. These could include compliance reports or regular reporting requests they need to handle.

The second bucket is data and analytics they didn’t previously have access to or couldn’t easily obtain. The most exciting part here is being able to impact student success. The way this plays out differs significantly between K-12 and higher education, due to the distinct needs and structures of each system.

In higher education, for instance, a key use case might involve the Office of Student Success seeing early warning signs (e.g. course activity, engagement) across learning management system (LMS) data for students who are at risk of dropping out. This is particularly critical for higher ed, where retention directly impacts revenue and operational stability.

In K-12, districts are looking closely at student achievement and meeting learning standards. Using Doowii to correlate data from assessments, attendance, and behavior, district coordinators can quickly spot students needing extra support. This proactive approach ensures students get the support they need before problems escalate.

Often, administrators have a hypothesis about what might be a good indicator, but there’s no easy way to pull that data. The platform now gives them direct control to explore this. For instance, they can ask for all students who haven’t looked at their syllabus within the first two weeks of enrolling in a class. This could be a strong indicator that a student might struggle for the rest of the semester. You can imagine a thousand similar scenarios that administrators or professors might want to test, and now they can do that easily with the platform.

As far as users are concerned, do you have a specific segment that you’re really looking to tap into in 2025? 

It’s both K-12 and universities. We have a good number of customers in both, but it leans more heavily toward higher education. We’ve seen faster adoption on the higher-ed side because they tend to use data more, have greater investment in data infrastructure and teams, and because data is more directly tied to how these institutions operate—particularly around retaining students and avoiding the loss of tuition dollars.

For K-12, it’s a bit more challenging and sometimes more complex. Data isn’t always a priority for districts when they have so many other needs to allocate resources to, often with very small budgets.

Part of the solution is helping districts become more efficient by using data and investing in infrastructure to create long-term savings and better resource management. We’re seeing some innovative districts pursue this and work with us to figure out how to manage the growing amounts of data being generated. Especially post-COVID, with so many digital tools in use, there’s a huge opportunity for data analytics to help improve outcomes. Right now, a lot of that data is just sitting there, underutilized, without districts fully benefiting from analytics.

What has been the hardest technical challenge around building Doowii into the platform it is today?

There’s a lot to consider. We think of each component of the platform as its own data analytics agent task. Some tasks are particularly difficult to solve. For example, one key component is the text-to-SQL functionality, which is critical for data retrieval and a core part of many analysis workflows. Ensuring that text-to-SQL is accurate is a significant challenge.

Currently, state-of-the-art benchmarks using LLMs are in the low 70% range for accuracy. That’s a considerable improvement compared to pre-LLM techniques, which were in the 40–50% range. However, 70% is still not good enough—if only two-thirds of your results are accurate, it’s not very usable for most cases.

This is one of the main reasons we’ve gone vertically focused. By building in custom context for specific systems and data models, we can push that accuracy much higher. Part of this involves building sophisticated RAG (retrieval-augmented generation) models, dynamic context management systems, and a multi-agent orchestration framework. The framework sits on top of everything, constantly evaluating how well each agent component performs.

How do you plan on Doowii progressing over the next 6-12 months? Anything specific on your product roadmap that your existing customers are excited about? 

As a startup, sales are always important, obviously. But on the product engineering side, our focus is on continuing to push the state-of-the-art for RAG models in data-specific infrastructure. Whether it’s text-to-SQL, text-to-ML, or text-to-data visualization, these processes involve working within a multi-agent framework that packages everything into a single response for the user.

The concept of agentic workflows becomes even more relevant as we improve the accuracy of returning the data users ask for. Once we’re consistently delivering accurate data, the next step is enabling users to take action on that data. For example, if the platform identifies a list of students likely to drop out, it should then recommend actions—such as reaching out to the students directly via email, contacting a professor, or other interventions.

The goal is to make the data more actionable and easier for users to leverage effectively, ensuring it delivers real, practical value. Technology alone will not improve education—people are the agents of change. To unlock data’s potential, we must break down the user barriers to data insights. Only then can we transform raw data into action and genuinely empower educators, rather than simply offering another delivery mechanism or data aggregator.

We've seen a lot of excitement around AI agents and the idea that autonomous systems will be able to complete human tasks. How are you thinking about integrating AI agents into Doowii itself? 

I think "agentic" is a bit of a buzzword. What it really comes down to is accurate tool calling or accurate function calling. For an agent to do that effectively, it needs the right amount of context about the tool it’s interacting with and the right level of access and control over that tool. You don’t want the agent producing false negatives or false positives in its behavior. It’s about having tight controls to prevent errors while minimizing friction for the user.

With Doowii, we are building API frameworks that LLMs can interact with smoothly, along with strong observability to monitor interactions. All the tooling around these agent frameworks has to strike the right balance—giving the agent enough freedom to perform its tasks while preventing rogue behavior that could lead to catastrophic outcomes.

This is especially critical when dealing with sensitive student data. For example, if the agent is allowed to send emails, it must only send them to individuals who are authorized to access that data. This means linking the agent’s actions to data access controls, which adds a lot of complexity to ensuring safe and accurate behavior.

Lastly, tell us a little bit about the team and culture at Doowii. Are you hiring, and what do you look for in prospective team members that are joining?

We are hiring and always looking for top talent. Whether you’re a front-end engineer, back-end engineer, or AI/ML engineer, if you’re highly competent, we’d love to talk to you. My general philosophy is that if you hire super competent people, it doesn’t matter if they don’t have the exact domain expertise you’re looking for—they’ll learn and adapt. That’s part of what’s necessary in a startup environment.

In terms of culture, the way I describe it is more like building a sports team than a family. Startups are sometimes described as a “work family,” but I think that analogy doesn’t quite fit. With a family, you don’t get to choose who’s in it. On a sports team, you either earn your place or you don’t make the cut, and you have to prove yourself every single day to stay on the team.

Everyone on the team is bought into a specific goal, even if it’s a moonshot. It’s like in the NFL—there are 32 teams made up of the best players in the world, and each of them has a low chance of winning the Super Bowl. But they’re all striving toward that goal with everything they have.

That’s the kind of culture I want to build. It’s fun, it’s hard, and that’s where the satisfaction comes from—accomplishing something that’s really difficult and being okay with the failures that are likely to happen along the way.

Conclusion

To stay up to date on the latest with Doowii, learn more about them here.

Read our past few Deep Dives below:

If you would like us to ‘Deep Dive’ a founder, team or product launch, please reply to this email ([email protected]) or DM us on Twitter or LinkedIn.