Code and Understanding: How Patrick Dajos and Hyperbound Are Building the Next Generation of Intelligent Software

What if the software you use every day could not only follow instructions, but also adapt in real time and even anticipate your needs? That idea, once confined to science fiction, is now becoming reality. Intelligent software does more than execute commands: it learns from context and collaborates with humans in solving complex problems. This evolution is raising urgent questions about how people and machines will work together in daily life and business.

'AI and LLMs are still quite a new thing. They are not as smart as people would like them to be,' says Patrick Dajos, founder of Avalonia, a premium software development agency with global clients. 'In order to get the best results, we need to give AI a sort of tunnel vision on one concept or one niche. Most companies fail because they go too broad, and the result is poor quality or outright hallucinations.' In his current work as a distinguished founding engineer at Hyperbound, a San Francisco-based Series A startup focused on building intelligent software for sales teams, he focuses on narrow, purposeful AI applications to real-world products to help usher in the next wave of innovations in intelligent software.

From Language Models to Human Collaboration

The arrival of powerful large language models transformed how machines interpret human language. Dajos points to OpenAI's work with ChatGPT as an example of this shift. 'They're saving entire conversations people have with ChatGPT and using them to power memory, so the system understands you better. That's extremely powerful because it's no longer just keyword matching—it's about full conversational context.' While the potential is immense, the obstacles are equally significant. Current systems remain limited by their small context windows, which prevent them from processing longer and more nuanced conversations. For developers like Dajos, the task is to design around these limitations while pushing forward. 'The most successful companies right now aren't trying to replace developers or designers,' he says. 'They are trying to power them up. We are not building sales agents to talk with people. We're using AI to make salespeople better at what they do.'

Real Call Scoring and the Psychology of Conversation

This philosophy comes to life in Hyperbound's flagship feature: real call scoring. Initially built to evaluate AI roleplay sessions for training sales teams, the system now analyses live conversations between salespeople and clients. The innovation lies in its adaptability. Teams can customise scorecards to align with their unique sales strategies and cultural contexts. Whether one team prefers skipping permission-based openers or an enterprise team values structured greetings, the system adapts accordingly. 'Generalised feedback isn't impactful. The real value is in feedback people actually want to listen to,' Dajos says.

The tool also delves into the psychology of dialogue. It detects objections, tracks engagement, measures filler word use, and even gauges reactions to authenticity. These insights help teams understand not just what was said, but how it was received, a crucial distinction in sales. 'So much of a conversation is about flow, tone, and subtext,' he says. "You need to capture that if you want actionable feedback.'

The feature faced issues at launch. Early iterations struggled under the weight of hundreds of calls daily, with unstable systems and inconsistent feedback. Dajos spent two weeks refining the product through iteration, and it paid off. The result was a stable, reliable system that now processes hundreds of thousands of calls for Fortune 100 and Fortune 10 companies. 'I'm so proud of it,' he shares. 'It started as an experiment, and now it closes deals for some of the biggest companies in the world.'

Building the Future Responsibly

Human collaboration is at the heart of progress, and Dajos believes this will remain true even as AI capabilities expand. For him, the real opportunity lies in steering intelligent systems toward narrow, high‑impact use cases that empower professionals rather than replace them. The success of features like real call scoring illustrates how these targeted innovations can transform industries while keeping humans firmly in the loop. 'Right now, AI is best as a sidekick at your side, helping you be more productive,' he says. And as the next wave of intelligent software emerges, Hyperbound's work shows how precision and practicality will shape the future of AI.

Connect with Patrick Dajos on LinkedIn for more insights on the future of intelligent software.