From Saheed Azeez’s YarnGPT to Ijemma Onwuzulike’s Igbo Speech, artificial intelligence has seen some degree of development in Nigeria; a fraction of what has been achieved in Western countries.

Besides hobbyists like  Azeez and Onwuzulike, there are very few true AI companies in Nigeria. Instead of full-fledged AI companies, the country has fintechs or insurance companies that embed AI to detect fraud or automate payouts — most of these solutions leverage existing AI infrastructures built by AWS, Microsoft, or ChatGPT.

However, there’s a Nigerian startup called Autogon AI that wants to do something no African startup has done. It wants to build from the ground up, an AI infrastructure that African and even foreign businesses can integrate into their systems at a fraction of the cost and time it would take to code into the systems.

Think of Autogon AI as a drag-and-drop builder for AI that businesses can use to create AI-powered tools, like fraud detection or virtual assistants.

Meanwhile, building an AI infrastructure from scratch is no joke. Autogon AI’s Co-founder and CEO, Obi Ebuka David, is not new to this, he built Identity Pass (now Prembly). However, in a conversation with Techpoint Africa, it was clear that Autogon AI was a completely different ball game.

Building Autogon AI

Before DeepSeek came onto the scene, building AI infrastructure — especially foundational AI (LLMs like ChatGPT) — has been perceived as a billion-dollar project.

Chinese-owned DeepSeek built an AI model comparable to OpenAI’s GPT-4 but with far less funding compared to OpenAI’s billions. They achieved this by optimising infrastructure, leveraging Chinese AI expertise, and efficiently training models.

According to David, Autogon AI is following the same model. “DeepSeek had fewer resources but still made groundbreaking advancements. That’s the same mindset we have at Autogon AI, building powerful AI solutions without needing billions in funding.”

David was inspired to build Autogon AI after seeing how difficult and expensive it was for businesses to integrate AI into their operations. As a Co-founder and former CTO of Prembly, he experienced first-hand how expensive and time-consuming it was to hire skilled AI engineers.

Even with AI training programs, he noticed that many people struggled to build real-world AI solutions due to the steep learning curve.

This led him to ask: “How can we shorten the timeline for businesses to implement AI without needing deep technical expertise?”

His answer was a no-code AI platform that allows businesses to build and deploy AI models in minutes without writing a single line of code.

“Companies don’t need to spend millions or hire an army of engineers just to use AI,” David explained. “With Autogon AI, even someone with no AI background can create powerful models in a few clicks.”

However, simplifying AI on this level required a lot of work.

“We worked on this for a year and a half. It took a team of super smart engineers.” Getting these engineers was not a walk in the park. David had to look for people who were already extremely smart and train them to do the kind of work he wanted.

According to David, to hire someone exactly like him would cost $15,000 to $25,000 a month, so he had to be creative in sourcing talent.

What makes Autogon AI special?

Without technical expertise, it can be difficult to understand the work that went into creating Autogon AI, but David tried his best to simplify things.

Only 2% of Autogon AI relies on existing infrastructure; 98% was built from scratch, meaning David and his team had to develop custom AI algorithms, model training systems, and real-time APIs from the ground up.

“The hardest part was making something this complex feel simple,” David said. “It took us over a year of deep engineering work to get it right, and every piece had to be custom-built to handle different AI use cases.”

The different AI use cases include fraud detection, customer behaviour analysis, predictive analysis, and risk management.

Giving an analogy of how it works, David said a fintech can easily de-risk transactions happening on its platform by making sure they follow compliance rules, or understanding user behaviour and creating an alert system when there are oddities in usage.

There is also another use case that allows businesses like loan companies to predict what their customers will do. David said these companies can create a predictive model that can help them reduce risk.

“You can predict whether a customer is going to default on a loan and the best amount to give them.  You can bring in your existing data sets, and the AI understands your user behaviour.”

David pointed out that as simple as these things sound, they would have normally taken a machine engineer about 2,000 lines of code to create. However, cramming 2,000 lines of code into three clicks was not easy.

 ”Imagine dragging and dropping a flow, and those multiple lines of code have been abstracted.”

David and his team had to build and patent algorithm structures, which he said have been used in medical research to develop AI models for tuberculosis detection, brain haemorrhage analysis, skin disease identification, and even drug discovery, helping researchers generate novel treatments for diseases like pancreatic cancer and COVID-19.

He also said they built low-level gen AI models, which they were able to connect together with the help of AWS.

The startup also had to consider and build for companies that might have very large data sets.

However, he pointed out that the startup had some help. They built the low-level Gen AI with Google’s attention architecture.

How Autogon AI works

The first step for businesses that want to use Autogon AI is uploading their datasets. If businesses don’t have their own data, they can integrate Autogon AI’s ready-to-use APIs into their existing systems.

“You don’t need to know how to train an AI model,” David said. “You simply tell Autogon AI what you want to build, and the system does the rest.”

Once the data is uploaded, Autogon AI’s machine learning engine processes it, trains a model, and optimises it for use. The system automatically generates APIs and SDKs that allow companies to integrate AI into their applications, websites, or business tools without any coding required.

“Normally, integrating AI into a product requires a dedicated development team; with Autogon AI, businesses get a fully functional AI solution that they can connect to their operations with just a few clicks.”

For example, if an online store wants to implement an AI-powered virtual assistant, they can generate the chatbot using Autogon AI and deploy it on their website in minutes without any technical setup.

How Autogon AI makes money

Autogon AI operates on a subscription-based model, with businesses paying monthly fees to access its AI tools. The company also generates revenue through custom AI model development and enterprise partnerships, especially in the financial services sector.

David said less than $150,000 has been raised, but the startup can sustain itself based on the revenue it makes. While he didn’t give revenue numbers, he said that the company is looking to hit $150,000 to $200,000 in monthly recurring revenue.

However, the company has a billion-dollar cash cow — its tuberculosis detection, and brain haemorrhage analysis model. David said that if the drug discovery goes live, the company can sell it for $500 million.

Till then, the startup has to figure out how to get more businesses and even the government interested in what it is doing because Autogon AI’s growth is dependent on the growth of AI development on the continent.