Using AI Helper

AI Helper is the fastest way to create a flow in JAMS Incident Management. You describe what you want to monitor in plain English; JAMS Incident Management drafts the flow and asks you to confirm any missing details before saving.

It's especially useful when you're not sure which trigger type or settings to pick β€” the AI handles the structural decisions.

Two modes. The AI Helper page has a toggle at the top:

  • Build a flow β€” draft a flow from a description (the rest of this article).
  • Ask a question β€” ask about JAMS Incident Management and get answers grounded in this Help Center. See Ask a question below.

Plan availability: AI Helper is included on the Professional plan and up. See Plans & Billing.

Before you use it

AI Helper needs an AI provider API key configured for your tenant β€” either OpenAI or Anthropic. Without one, the page falls back to "heuristic mode" β€” basic pattern matching that catches the common cases but isn't anywhere near as good.

Set the key once at Dashboard β†’ User menu β†’ AI Provider Key. It's a tenant-wide setting, not per-user. Costs a few cents per flow drafted β€” typically under $1/month for most teams.

You don't pick a provider β€” the key does. Paste an OpenAI key (sk-...) or an Anthropic key (sk-ant-...) and JAMS Incident Management works out which vendor it belongs to and tells you what it detected before you save. Switching vendors later is just pasting the other key; there is no setting to change and nothing to restart. One key per tenant β€” saving a new one replaces the old.

Why isn't this just "use JAMS Incident Management's key"? Each tenant pays for their own AI usage so we don't have to charge a flat monthly markup and we don't see your prompts. The key is encrypted at rest before storage.

Opening AI Helper

Click AI Helper in the sidebar (or hit the "+ New flow" button on the Flows page and pick AI Helper).

You'll see a chat-style interface. The first message asks you to describe your alert in plain English.

Prompts that work well

The more specific you are, the less back-and-forth you'll have. Two examples that work cleanly:

Email-triggered alert:

Alert me by SMS and call when an email arrives in monitoring@mycompany.com with "CRITICAL" in the subject. Send to the on-call group. Escalate after 2 minutes.

JAMS job failure:

When a JAMS job called "Nightly ETL" fails, call my phone and SMS the data team. If no one acknowledges in 3 minutes, also email the team lead.

JAMS Incident Management parses the prompt into a draft flow β€” trigger, recipients, channels, escalation β€” and shows you what it built. If anything is missing (an email connection you haven't created yet, a Team Group that doesn't exist), it'll ask for it inline.

Prompts that won't work yet

AI Helper can't yet do things JAMS Incident Management doesn't support natively. Some examples:

  • Conditional logic across flows β€” "if A fails AND B is offline, alert; else just log." Each flow is independent.
  • Time-based scheduling of the alert itself β€” "every day at 2am, send a status SMS." Flows react to triggers, not the clock. Use a cron job that calls your webhook for this.
  • AI summarization of payload β€” "include a 1-sentence summary of the error in the SMS." The message body comes from the trigger payload as-is.

After the draft is created

AI Helper takes you to a review screen showing the flow it built. Before saving:

  • Verify recipients β€” most common edit point. AI Helper picks reasonable defaults but might not know your team structure.
  • Verify escalation timing β€” defaults to 5 minutes; tighten to 1–2 minutes for production-critical alerts.
  • Check the trigger details β€” for webhook flows, JAMS Incident Management shows the URL + API key after save. For email flows, confirm the right mailbox is selected.

Save β†’ the flow is live. Test it (see Getting Started step 5).

Editing flows AI Helper created

Once saved, flows from AI Helper are no different from manually-built flows. Edit them at Flows β†’ click flow name β†’ Edit. The AI doesn't "own" them β€” they're regular flows.

If you want to recreate something similar with a tweak, easiest path is copy a working flow from the Flows list (Actions menu) and edit the copy.

When AI Helper says "Heuristic mode active"

That's the fallback β€” your AI provider couldn't be used. See troubleshooting for the fix (usually means the key isn't set, isn't valid, or hit a rate limit).

The heuristic mode catches the most common patterns (webhook + SMS, email + call), but anything beyond that needs a real LLM. Set the key and try again.

Ask a question (Help Center Q&A)

The Ask a question toggle on the AI Helper page turns it into a documentation assistant. Ask something like "How do I set up the JAMS connector?" or "Which recipients can escalate?" and it answers from this Help Center, linking the articles it drew from so you can read the full detail.

  • Answers are grounded in the Help Center β€” it won't invent features that don't exist, and it cites the docs it used.
  • It uses the same tenant AI provider key as Build mode. Without a key, Ask mode returns a short pointer to the Help Center instead of an AI answer.
  • Ask mode is read-only β€” it doesn't create or change anything. Switch back to Build a flow when you want to make something.

Tips

  • Be specific about channels. "Notify me" defaults to SMS. If you want a phone call, say so explicitly.
  • Name recipients by group, not individual emails. "On-call team" or "data team" works if you have those Team Groups; raw emails work too but are less reusable.
  • Set up prerequisites first. If your flow needs an email connection or a Team Group, create those at Connections and Team Groups before running AI Helper. Otherwise it'll keep asking inline.