Article
CommunityJev Integrations & Adapters
Smart Conditions and AI Assignment in There There
A follow-up on There There's Jev use: workflow conditions described in plain language become a Boolean question in the same inbound classification request, and Jev picks a ticket's assignee from a curated list.
- Format
- Article
- Reading time
- 3 min
- Published by
- Community
- Author
- Freek Van der Herten
- Last verified
- Oct 2026
- Languages
- PHP
- Capabilities
- Requires API key
Highlights
- A smart condition turns its plain-language description into a Boolean question about whether the incoming message matches it.
- The condition's question joins the spam and auto-reply questions in the same inbound request, and the workflow reads the stored probability later.
- The editor previews a sample email or an existing ticket before a workflow is enabled; only incoming email is checked, not the full ticket history.
- Assignment is a Choice question over the selected people plus a None of these option; Dries, Jef, Jimi, and Zuzana each have a short description.
- Jev only assigns on a clear favorite, disabled members and people without channel access are removed first, and team and person assignment stay separate.
Watch out
Production notes from one team, with screenshots rather than benchmarks: the post gives no accuracy figures and does not say what happens when the top-choice probability is flat. Both features depend on There There's workflow editor, and the final decision stays in application code.
Questions about this entry
Short answers drawn from the linked project or page.
A workflow condition you describe in plain language, such as the customer asks for a refund or to cancel a payment. That description becomes a Boolean question in the inbound classification request, and the workflow reads the stored probability as its match score.
The incoming message matches this workflow description, followed by your description, with true and false descriptions explaining what fitting and not fitting look like. The answer is stored on the message so later runs need no new Jev call.
You select the people Jev may choose from and add a short description of what each handles; Jev answers a Choice question with those candidates plus a None of these option. Only a clear favorite results in an assignment; otherwise the action makes no change and another workflow can pick it up.
Disabled members and anyone who cannot access the ticket's channel are removed before the question is asked, and spam and automatic replies are skipped before either the team or person assignment action runs.
In the same inbound classification request as the spam and auto-reply questions. The answer is stored with the message, and the workflow reads its probability from there rather than calling Jev again when the workflow runs.
The editor lets you try a sample email or an existing ticket and see the score before you enable the workflow. Only incoming email is checked; Jev receives the subject, sender, and message text, not the whole ticket history.
It gives no benchmarks, accuracy, or latency numbers for the new features, and it does not describe fallback behavior when probabilities are close. The approach still keeps the final decision in code: the workflow decides what happens next based on the answers.
From the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.