In this day and age of automation and machine learning (ML), many have learned to enjoy the convenience of having AI do most things for them. In most cases, this could lead to skill degradation and over-reliance on AI. On the opposite side of the spectrum, many people are wary of AI, for fear of it taking their jobs from them.
People like Mohammad Adnan tread the middle ground and seek to make AI even better, but not at the expense of leaving people without jobs, but making sure that AI serves to help people do what they do even better.
Mohammad Adnan contemplates the future of human-AI interaction and uses it to build AI systems that understand people. Adnan has previously spearheaded the development of foundational ML products at Amazon Web Service, including AWS Bedrock, Amazon Q for business, and the AWS’s Titan model.
Helping Learning Machines Unravel the Human Behavioral Puzzle
When people interact with AI, they are engaged in a sophisticated dance of meaning. The machine processes our words and also attempts to decipher our intentions. This is what technologists call “intent recognition,” and it is transforming how businesses engage with customers. According to recent data, 92 percent of businesses using AI report improved response times, while 83 percent say it makes handling customer requests easier.
Mohammad Adnan has a background in developing AI systems that serve millions, and understands that intent recognition is not merely a technical challenge but a deeply human one. He has worked on creating systems that respond to explicit requests and also anticipate needs based on context, history, and subtle linguistic cues. “AI needs to understand not just what words people use, but what they mean by them,” Adnan says.
Beyond Basic Understanding
Traditional search systems relied on keywords—matching the right terms and getting the right results. But human communication is messier, more nuanced. The AI systems Mohammad Adnan helped develop at Amazon moved beyond simple keyword matching to understand the underlying intent.
Typically, modern AI systems recognize four primary types of intent, including Informational intent, which seeks knowledge or answers, like when someone asks about the weather or a recipe; they are looking for facts, not actions; navigational intent, which aims to reach a specific destination, like when a user says. “Take me to the Portal’s help page”, which the AI interprets as not a request for information but a desire to go somewhere in the digital space.
There is transactional intent, which focuses on completing an action, like making a purchase or booking a service, and finally, support intent seeks assistance with a problem or question, often in customer service contexts. The ability to distinguish between these intents—and respond appropriately—marks the difference between frustrating AI interactions and helpful ones.
How the Human Element Plays into Machine Understanding
Mohammad Adnan’s approach also leverages generative AI (genAI) and ML for automation and enhanced user experience, particularly for small businesses handling payroll challenges. Many neuroscientists observe that the human brain is its own universe, making it distinct from how AI processes information. AI does not think like humans, as it lacks consciousness, understanding, biology, self-awareness, emotions, moral sentiments, and agency.
Yet, sophisticated pattern recognition and training on vast datasets of human communication can create an impression of understanding. Mohammad Adnan’s work assists in this aspect as his innovations help machines link human expression and machine comprehension.
What is the Next: Retrieval-Augmented Generation
One of the most promising approaches to improving AI understanding is Retrieval-Augmented Generation (RAG), which combines real-time data retrieval with AI content generation. This approach allows AI to gather relevant, up-to-date information when responding to queries, rather than relying solely on its training data. Industry experts believe that the future of RAG will likely include more sophisticated retrieval algorithms, specialized knowledge bases, seamless integration with other AI systems, and more intuitive user interaction.
Mohammad Adnan’s work had already factored this in. By creating systems that can retrieve relevant information and generate helpful responses, he is helping to build AI that truly understands what users need. This is an immense help for businesses and services that are largely reliant on understanding the needs of their consumers.
The Question Of AI Autonomy
As AI becomes more autonomous, there is a risk that it could disturb creative intent and negatively affect individual learning and development. Many industry applications reduce humans to uncritical users of AI tools with pure application skills. Mohammad Adnan’s work is proving to be essential, as his approach, informed by his experience in AI development, emphasizes keeping humans in the loop. His work focuses not on replacing human decision-making but on enhancing it through AI-powered insights and automation.
This approach aligns with what experts call “reflective practice”, the ability to think critically about one’s actions and adjust accordingly. As AI continues to evolve, the challenge of intent recognition will only grow more complex. Future systems will need to understand what users say and what they mean in increasingly nuanced contexts.
