Please introduce yourself and tell us about what you’re building at ProtonLabs?
I’m Emmanuel Olarewaju David, a software engineer, AI and machine-learning engineer and co-founder and CTO of ProtonLabs Technology. I hold a master’s degree in Artificial Intelligence from the University of Manchester. My work has included applying AI in logistics, healthcare and education, but the common thread has always been productivity: how can we use intelligent systems to remove repetitive work, make complex information easier to understand and help people make better decisions?
At ProtonLabs, I lead our AI and software engineering work. We are building a portfolio of practical AI products, including AskYourPDF for document intelligence, Cowriter.ai for AI-assisted writing and deep research, Detect.ai for AI-content detection, and Simplicity.AI for intelligent document automation. I am especially interested in the last mile of AI: turning a capable model into a reliable, accessible product that people can trust in their everyday work.
ProtonLabs’ AI products have attracted more than 5 million users. What do you think has been the key to that growth?
The key has been the combination of choosing a real problem, moving quickly and taking distribution seriously. AskYourPDF started with a very clear pain point: people had valuable information trapped in long documents and did not have an efficient way to find and use it. We released an early product, observed how people actually used it and kept shortening the loop between user feedback and product improvement.
ProtonLabs is bootstrapped, so we have had to be disciplined about the difference between attention and value. Sign-ups are useful, but customer retention, willingness to pay and word of mouth tell you whether a product is genuinely helping people.
I have also learnt that technical founders must learn to sell. A technically strong product does not distribute itself. We made our products available through the web, ChatGPT, mobile applications, Chrome, Zotero and APIs, meeting users where they already work rather than requiring them to develop completely new habits. Product quality creates retention, distribution creates discovery and speed of learning connects the two. I believe our growth has come from combining all three rather than relying on any one of them.
AskYourPDF has become one of the most widely used productivity tools on ChatGPT. Why do you think it has resonated with so many people?
I think it has resonated because it solves a universal job rather than serving only one profession by helping people find, understand and use information quickly. A student reviewing lecture notes, a researcher comparing papers, a lawyer reading a contract and an investor analysing reports all have different contexts, but they share the same underlying problem.
When we built the Research Assistant, we did not predict either its scale or the full range of use cases people would discover. It works inside a familiar ChatGPT experience and provides access to more than 400 million research papers from sources such as PubMed, Nature and arXiv. It can help users discover literature, analyse documents, generate summaries, organise references and produce well-cited research.
In 2025, GOBankingRates included AskYourPDF Research Assistant among three ChatGPTs highlighted for investors in an article that was later republished by Yahoo Finance. We did not build it specifically as an investment product, and that was an important lesson for us. When you solve a broad research problem well, users will adapt the product to valuable needs you may never have anticipated.
What is the biggest misconception businesses still have about AI?
The biggest misconception is that adopting a model is the same as having an AI strategy. Many businesses still treat AI as plug-and-play. In reality, the model is only one component. The difficult work is selecting the right workflow, preparing the data, integrating existing systems, deciding what happens when the model is wrong, evaluating quality, controlling cost and helping employees and customers adopt the new process.
Do you think the AI industry is becoming too focused on hype rather than solving real-world problems?
There is certainly too much hype in parts of the industry, but hype and genuine progress can co-exist. AI capability and adoption are advancing rapidly, while many organisations are still finding it difficult to turn experiments into reliable, organisation-wide value.
A test I like is to remove the word “AI” from the pitch. Is the value proposition still compelling? Can the company demonstrate time saved, errors reduced, revenue created, decisions improved or users retained? If not, AI may be functioning mainly as a marketing label. When the answer is yes, AI is doing what good technology should do by being almost invisible while the customer outcome becomes obvious.
What separates successful AI startups from those that struggle to gain traction?
Successful AI startups turn probabilistic technology into a dependable user experience. They begin with a specific problem that customers experience frequently, stay very close to their users, measure where the system fails and continuously improve the product based on how people actually use it.
On the other hand, many startups struggle because they build technology before validating the market. They may have an impressive demo but no clear customer, and no sustainable business model. In the end, success isn’t about having the smartest AI system. It’s about building a product people keep coming back to, are willing to pay for, and recommend to others.
Many companies are rushing to add AI features to their products. What should founders consider before doing so?
Founders should begin by asking whether the customer problem actually requires AI. A rules-based system or simpler form of automation may be cheaper, faster and more reliable. They should then consider the full operating reality. What data will the system access, and does the company have permission to use it? How will privacy, security, bias and regulatory obligations be handled? What is the consequence of a wrong answer, and where is human review or escalation required? Customers should clearly understand when they’re interacting with AI and what its limitations are.
What are you most looking forward to seeing in US50 entrants this year?
I am looking forward to seeing companies that make complexity feel simple. I want founders who can explain, in plain language, exactly who the customer is, what painful problem they are solving and why their approach is meaningfully better than the current alternative. From a technical perspective, I will also look at how teams think about reliability, data, security and economics. I am especially drawn to founders who can build great products, communicate their vision clearly, and learn quickly.
What excites you about the US startup ecosystem in 2026?
What excites me most in 2026 is the shift from AI as a standalone chat interface to AI embedded in real workflows. We are beginning to see systems that do more than generate an answer. Instead, these systems can retrieve information, use tools, complete parts of a process and hand work to a human at the appropriate moment. That opens significant opportunities across document operations, healthcare, logistics, education, finance, legal services and industrial work.
The US is uniquely positioned because it brings together frontier research, cloud and computing infrastructure, deep capital, experienced operators and a large base of customers willing to try new products. Those ingredients create very fast feedback loops and allow small teams to pursue global-scale problems.
At the same time, funding and attention are highly concentrated around AI, so competition is intense and the standard is rising. I see that as healthy. The next generation of winners will need not only impressive technology, but measurable value, customer trust and sustainable economics.
Any advice for US startups entering the TechRound US50 this year?
Be concrete and make it easy for the judges to understand your company. In the first two sentences, tell us who you serve, the exact problem you solve, the outcome you create and what makes your approach different. Do not lead with a long history of the market or a list of technologies.Then support the story with evidence. Use numbers that provide context such as retention, revenue, customer outcomes, growth and time saved. Explain why now is the right time, why your team is suited to solve the problem and how the current product can expand into a larger business.
My final advice, especially to technical founders, is to explain distribution as clearly as architecture. Show not only that you can build the product, but how customers will discover it, trust it, pay for it and continue using it.
