TL;DR - Business Process Management and Case Management have a first-order place in the AI Agent ecosystem
I’ve been spending the last week working with Anthropic’s Claude. I started with Claude Code but then discovered that Opus 4.1 does a great job creating a CVs targeted toward a specific job listing. I’ve started spamming the AI ASTs with CVs to test the results.
I’m a big fan of repeatable processes, though, so as part of our work Claude and I drafted a process document and I used a Claude project to ensure the LLM reviews the process document. Sounds great right?
Except, after a dozen or so iterations, I noticed that Claude was skipping parts of the process I hadn’t explicitly asked about, especially proof-reading and fact-checking a draft CV.
Realization: When building processes, it’s advantageous to think of LLMs more as wetware than as software -- they’re fallible, random, and somewhat lazy.
Implication: The same audit and review processes are necessary on LLM-based processes as were necessary on people-based processes. Cross-LLM platform quality checks, “four-eyes” reviews, and auditable step completions all have a place in the LLM universe.
Fortunately, in the universe of Business Process Management (BPM) and case management tools, we have the infrastructure to leverage such processes as may already exist and to create new, optimized processes.
What’s the action item? Look at your existing well-documented business processes and find the low-hanging fruit for LLM utilization. Maybe it’s a document creation process that the current teams hate. Perhaps it’s a major Googling exercise. Find the mind-numbing human tasks an LLM could do and free up the creative humans for the tasks that could really matter.
