Hey everyone, been following this subforum while trying to learn more about modern support tools. All the talk about AI-assist for agents is super interesting, but honestly, it's hard to cut through the hype.
I'm looking for tools that actually help in 2026, not just promise. What are you all using that *works*? I'm especially curious about:
* Tools that integrate directly with ticketing systems (like Zendesk or Jira Service Desk).
* Ones that can reliably pull from internal docs or a knowledge base to suggest answers.
* Anything that helps with triage or routing based on the customer's problem.
Bonus points if it's something that can be self-hosted or run in a container. I'm comfortable with Docker, so if you have a compose example for a local test setup, that would be amazing. For example, I've seen open-source tools like DeepPavlov mentioned, but no clear setup guides.
What's giving you the best results without getting in the agent's way?
Containers are magic, but I want to know how the magic works.
Great question. Cutting through the hype is the hard part.
From what I've seen, most of the big, reliable tools for Zendesk and Jira are SaaS-only now. The self-hosted or containerized options tend to be fragile, and the setup and maintenance time often kills the ROI. You're basically building your own support team for the AI tool.
If you're set on a local test, check out Rasa's latest open-source framework. It's built for that, and their docs have a Docker guide. But temper expectations - getting it to reliably pull from your internal KB and not hallucinate is a multi-month project.
For actual production use, we went with a vendor. The one that works for us had to pass a simple test: did it actually save agent time on mundane replies, and did the agents *use* the suggestions without being forced? Most fail that second part.
Trust but verify.