AI has the potential to transform industries, but trust remains a major challenge. Many AI systems function as black boxes, making it difficult to ensure accuracy, reliability, and transparency. While AI-powered chatbots and language models are impressive, their unpredictable nature makes them risky for enterprise use.
In this episode of Semaphore Uncut, Ken Kocienda, co-founder and CTO of Infactory, joins Darko Fabijan to discuss how his company is building AI systems that are both useful and dependable. Ken, a veteran software engineer known for his work on Safari, iPhone auto-correct, and Apple’s UI innovations, shares his insights on how AI can move from a black-box guessing machine to a fully transparent and enterprise-ready system.
From Apple to AI: Ken’s Journey
Ken has spent over 25 years shaping the way we interact with technology. At Apple, he played a key role in developing Safari and WebKit, then later joined the original iPhone team, where he worked on foundational UI elements like touchscreen scrolling and auto-correct.
After years of working at the intersection of human-computer interaction and software, AI was the next natural step. However, Ken saw a major problem — AI’s lack of predictability and transparency. It became clear that businesses couldn’t fully trust AI-generated responses without a way to verify them. This realization led him to co-found Infactory, a company focused on delivering AI-powered systems that produce consistent, fact-based answers rather than unpredictable outputs.
The Problem with AI Today
One of the biggest concerns with modern AI is that it behaves inconsistently. You can ask the same question multiple times and get different answers. While this might be acceptable in casual conversations, it’s a dealbreaker for businessesthat rely on accurate, traceable, and secure data.
AI’s black-box nature makes it difficult for companies to understand how a response is generated. In many cases, it pulls from vast datasets without explaining why it reached a particular conclusion. This lack of control raises major concerns for industries where accuracy is essential — such as finance, healthcare, and enterprise software.
Ken believes that AI doesn’t have to be a guessing game. At Infactory, his team is focused on building AI that delivers factual, structured, and transparent responses — something that today’s AI models often lack.
How Infactory is Changing AI Development
Instead of relying on AI to generate free-form answers, Infactory structures AI development in a way that ensures consistency and transparency. Their system works by first connecting to structured datasets and then allowing developers to build reliable, repeatable queries.
Rather than a model generating its own response, Infactory’s AI converts natural language requests into deterministic, structured outputs. Developers can then deploy these as secure API endpoints, ensuring full traceability and compliance.
This approach removes the uncertainty that plagues traditional AI models. Enterprises don’t have to worry about hallucinations, unreliable answers, or security risks. Every interaction with InFactory’s system is logged, making it easy to audit and refine responses as needed.
Bringing AI to the Enterprise
Many businesses are excited about AI but hesitate to fully integrate it due to security and compliance concerns. Ken points out that companies in finance, healthcare, and enterprise software can’t afford unreliable AI responses. A miscalculated financial report, incorrect medical recommendation, or misleading customer support response can have serious consequences.
By ensuring that AI responses are deterministic and verifiable, Infactory provides companies with the tools to safely integrate AI without the risks of unpredictable or unverifiable outputs.
What’s Next for Infactory?
Since launching in 2024, Infactory has been working with early enterprise customers to refine its platform. Ken and his team are focused on scaling their pilot programs, hiring backend and ML engineers, and bringing their AI-powered infrastructure into real-world deployments.
His vision is clear: AI should be as transparent and reliable as any other software system. For AI to be truly useful, developers must have full control and visibility over how it works — something Infactory is making possible.
Final Thoughts: AI for Developers
“I want to build AI tools that I would actually use. If we solve this for ourselves, we solve it for the industry.”
Ken believes that developers shouldn’t have to gamble with AI outputs. By focusing on deterministic behavior, security, and enterprise reliability, Infactory is making AI more trustworthy and practical for real-world applications.
Follow Ken & Infactory
🌐 Website: Infactory.ai
🐦 Twitter (X): @infactory_ai
🐦 Ken’s Twitter: @kocienda
Originally published at https://semaphore.io on March 18, 2025.
