Despite some recent media headlines to the contrary, Agentic AI is harder than it looks. So what should you be seeking in a custom AI agent development shop these days? As Chief Architect of Level Up, I’m not objective, but I have done my homework, and I personally think it looks a lot like what Level Up is already doing:
1) People
Unlike a lot of those social media accounts with hype-filled “insane”, “mind-blowing” demos on your feed and mine, Level Up has a proven track record of turning upstream open source projects into successful, scalable, and secure enterprise open source solutions, for years now. AI agents are really just more of a continuation of long-term innovation trends for our architects, and I think it shows in the quality of our work.
2) Processes
This is what has been called “the middle to middle” problem in AI workflows. While many ISV’s and dev contractors claim to be offering AI agents that will run “end to end”, in reality far too many are forcing customers to restart their traditional workflows at new, awkward points. They also make clients deal with abrupt technical endpoint handoffs that don’t really make much “business sense” for them. The hard limit of any custom AI agent project is going to be the limit of the external software engineer’s ability to understand your company’s unique, internal processes, metrics, and goals. Beware the ISV that tries to tell you they have a one-size-fits-all AI agent.
3) Technologies
With something that’s still fairly new like Agentic AI, perhaps the single most important thing that anyone offering to build you your first agent should be able to say is, that the same base AI agent is already working for them internally today. We’re very proud of the fact that the custom agents we’re bringing to clients now have already shown real-world value for our own consulting team in the first place. No guessing games, no “solutions in search of problems”. Just practical, field-tested results.
These AI agents include:
- Backlog Zero: Instantly integrates your service desk (e.g., Jira), GitHub, and Red Hat Ansible Automation Platform (AAP) to automatically create testable AAP job templates. User experience is primarily web UI.
- Worklog Hero: Customizable, truly agentic workflows for service desk task entry automations. User experience is primarily CLI.
- QandAgent: 100% internal Q&A docs agent for teams, as well as organizational insights for the C-Suite. User experience is primarily Slack, but customizable for almost any chatbot.
Notably, our AI agents are designed to take advantage of free open source language models and 100% local model serving options, such as Red Hat® AI Inference Server. Without question, open source AI is winning right now, and as a Red Hat Premier Partner, we applaud Red Hat’s ongoing leadership in the AI/ML space. And leveraging both open source models and local inference servers can substantially lower your total costs in successfully adopting AI internally… vs. cloud chatbots, whose per-token API pricing seems to be continually skyrocketing for increasingly frustrated users, almost weekly, as of this writing.
Thanks for reading! Message the Level Up team today if we can help you take next steps on your own AI Automation journey: arc@levelupla.io.
–Daniel Goosen, Chief Architect @ Level Up