Sift AI
Book a demo
Workflows

Describe the job. Agents and workflows run it.

Route conversations to the right queue, run steps when something matches, and describe outcomes in plain language with AI Goals. Every run is logged.

Book a demo Read the docs
AutomationsWanderly · example
When: matches saved search “Refund requests”Trigger Trigger
Then: move to Billing, tag Refund, notify #billing3 steps Steps
AI Goal: make sure every VIP gets a reply within an hourNatural language AI Goal
Last run: completed in 1.2sRun log OK

AI Goal ran on 14 conversations: 11 acted on, 3 abstained with a reason.

TriggersCondition, saved search, schedule
Webhook events9 types
RetriesUp to 24 hours
Run logEvery run

What it does

Triggers

Run on a condition, on a saved search match, or on a schedule in your timezone.

Steps

Tag, add a note, assign, move queues, set fields, close, notify Slack or email, send a webhook, or apply a saved response.

AI Goals

Describe an outcome in plain language. The AI agent acts, or abstains and says why.

Queues and routing

Queues come from saved searches, with priority order, auto-assign, routing skills, workload-aware assignment, and SLA targets.

Action rules

Auto-close inactive conversations, lock closed threads, set close reasons, and choose which ones send CSAT.

Signed webhooks

Send events like action created or status changed to your systems, with retries and a delivery log.

Rules for the predictable work

Pick a trigger, add steps, and let it run. Queues come from saved searches, so routing follows the same filters your team already uses.

Workflows
  • Tag, assign, move queue, set fields, close
  • Notify Slack or email, send a signed webhook
  • Apply a saved response

AI Goals for the judgment calls

Describe the outcome you want in plain language. The agent acts on matching conversations, or abstains and records why, so you can see every decision.

Workflows
  • Natural-language goals
  • Act or abstain, with a reason
  • Outcomes visible in the AI Goals tab

Where agents act, and where you decide.

Rules run the predictable work. AI Goals handle the judgment calls and log every decision.

  1. Agent
    Match

    A condition, saved search, or schedule starts the run.

  2. Agent
    Act

    Steps run in order. A failed step doesn’t stop the rest.

  3. Agent
    Judge

    AI Goals act or abstain, with the outcome recorded.

  4. Person
    Review the log

    Every run shows as started, completed, or failed.

Connects to your tools

All integrations →
SlackCustom API

Questions

Do I need code?

No. Workflows are built in Settings, and AI Goals are written in plain language. Webhooks are there if you want to connect your own systems.

What if a step fails?

The rest of the run continues, and the run log shows what failed.

More of the platform

AI triageScore, summarize, tag, and filter every message.AI repliesDrafts in your voice, grounded in your knowledge.Knowledge baseThe facts your agents answer from.SiftGPTAsk your data, get answers with sources.InboxEvery channel in one keyboard-first inbox.AnalyticsTopics, sentiment, SLAs, and custom dashboards.ListeningBrand mentions you weren’t tagged in.PeopleAdvocates, detractors, and who’s going quiet.ModerationSpam, policy, and blocked users.Quality & CSATQA agents, reviews, and customer scores.Security & adminRoles, audit log, and agent access controls.

Describe a goal. Watch it run.

We’ll set up a routing rule and an AI Goal on your own conversations.

Book a demo