KP Journal Applied research
in AI agent systems and automation engineering

"We came in expecting the audit to confirm what we already planned to build. It didn't. James walked through our document ingestion pipeline and identified two steps we'd assumed were automatable but actually required judgment calls that no current LLM handles reliably at our error tolerance. We deprioritised both. That reallocation probably recovered four to six weeks of engineering time we'd have spent on the wrong thing."

Marcus Theron

Head of Engineering, Relatico (B2B SaaS, Series A)

"KP Journal is one of the few places writing about AI agents for people who are actually building them, not evaluating whether to. The article on context window management was the one that hit hardest — specifically the distinction between gradual quality degradation and silent tool-result dropout. We were monitoring for the first and missing the second entirely. We added validation checks on tool-result carryover across turns in our categorization pipeline, and caught two failure patterns that had been running undetected in production for weeks. That's a direct operational outcome from reading an article."

Priya Sandra

ML Engineer, Series B Fintech — transaction intelligence team

"We commissioned a white paper on agentic document processing for a technical buying audience — architects and senior engineers who would immediately notice if something was abstracted beyond usefulness. The KP Journal team got the architecture section right on the first draft, including the retrieval boundary and audit trail framing. We did one revision round on the compliance language, which they turned around in under 48 hours without losing the technical register. We've used other content vendors for this kind of material."

Oliver Bitton

VP Product, B2B software company — February 2025