From ticket to knowledge base
The AI drafts the answer from the wiki, the agent sends it, and the article is written for next time.
Service Desk Wiki Automations CDP
Explore use case Marketplace · added to your suite
Automations
An event on any entity of the platform starts a chain: the AI reads what happened and classifies the event — and from there the system does the work along the scenario you built. No integrations, because all the data is already here.
A trigger on any entity The AI classifies the event rather than just writing text Runs and the success rate are visible
app.sintora.ai / automations / LIVE support bug triage
← Automations LIVE support bug triage System Enabled Test Check Save
Add a node
TRIGGER team · created
AI You are a support triage as…
ACTION create_project_task
+ −
TRIGGER Node settings
Entity type
Team
Team
test
Event
Created
Type
Request
Request form
Any form
Fires when a task or request is created in the selected team.
Automations — a marketplace module: it is added to a suite you already use, and is not sold on its own.
The key difference
A trigger can be put on anything
A standalone automation tool first has to connect to your systems, and every new entity is a new integration for somebody to maintain. Here there is nothing to integrate with: a deal, a ticket, a document, a call and an employee are records in one data model. So a trigger sees all of them from day one.
A deal in the CRM
created · stage changed · won · lost
A request in Service Desk
created · assigned · SLA breached · closed
Document
submitted for approval · approved · rejected · signed
Call
ended · summary produced · decision extracted
An employee
hired · changed role · left · on leave
A task or a project
created · status changed · overdue
Team
request created · new form · new member
Any record
created · updated · deleted
And the actions likewise. An automation that began with a support request creates a task in a project, assigns an owner from the HR structure and replies to the customer in their own channel. Three modules, one automation, zero integrations.
What it is made of
Five node types — and nothing else to learn
An automation is drawn on a canvas. Each node is configured in the panel on the right, and data from earlier steps is inserted through “Insert data”.
Trigger
Where it all starts: an entity, an event and a filter. Exactly one per automation, so it is always clear what set it off.
Entity and eventFilter by typeOne per workflow
AI
It reads the event and returns a structured answer. Not prose for a person but JSON, which the outgoing branches then test.
A prompt with the event dataAnswer schemaBranching on the answer
Action
It does the work inside the platform: creates a task, assigns an owner, replies to the customer, changes a status, posts to chat.
6 kinds of actionFields from the event dataSeveral actions in a row
Condition
A branching point. Usually it is easier to hang the condition on the connection itself, but when there are many branches it becomes a node of its own.
A condition on the connectionA separate nodeSeveral branches
Webhook
A signed POST to an external URL. This is how an automation reaches the systems that stayed outside the platform.
A signed requestAn external URLThe event data in the body
AI node
An AI that classifies the event rather than writing prose
The difference matters. An assistant writes a draft for a person. Here the AI returns a structured answer — the severity, whether an escalation is needed — and the workflow reads its fields and carries on along the scenario a person built and switched on. The AI classifies the event; what happens next with that answer is set by whoever wrote the branch.
- The prompt sees the event. Trigger data is substituted straight into the text: {{trigger.request_data}}.
- Answer schema. JSON with the fields the AI must return, so the branches after it have something to test.
- Branching on the answer. The outgoing branches read the answer fields: escalate = true takes one path, false the other.
AI assists. Humans remain accountable.
The node classifies the event. What to do with each answer — which branch to take and which action to run — is set by the person who built the scenario and switched it on; so are the limits on what the automation may do.
app.sintora.ai / automations / AI node
← Automations LIVE support bug triage System Enabled Test Check Save
Add a node
TRIGGER team · created
AI You are a support triage as…
ACTION create_project_task
+ −
TRIGGER Node settings
Entity type
Team
Team
test
Event
Created
Type
Request
Request form
Any form
Fires when a task or request is created in the selected team.
What an automation can do
Actions run inside the platform, so any field from the event can be substituted into them: the task title, the assignee, the labels, the text of the customer reply. A webhook adds everything else — a signed POST to your URL.
Create a project task
Link entities
Reply to the customer
Post a message to chat
Change the request status
Assign an owner
What it looks like in practice
Support
A customer complaint → a task for engineering
A request in the support team starts AI triage: it classifies the bug by severity, flags whether an escalation is needed, and creates a task in the project labelled «bug, urgent, from a customer».
Sales
A deal moves to won → a project
A won deal creates the project, carries the scope across from the proposal, assigns the team and posts to the channel that delivery is starting.
Documents
A contract is signed → an invoice and a folder
A signed document changes the deal status, creates a task to raise the invoice and links the document to the counterparty.
People
A new hire → access and the first week
A hire in HR creates the first week’s tasks and notifies the manager and buddy in chat.
When there are a hundred of them
An automation is easy to build. Living with it is the hard part
Six months in, the question is not how to create a workflow but which one has been quietly failing for three weeks. Which is why the run statistics sit on the card rather than in a separate report.
app.sintora.ai / automations
Automations
Event-based automations, AI triage and webhooks
Automations Runs
Search workflows, tags… ActivePausedDraftError
#ai#e2e#regression#support#triage
Support automations 2
E2E: support bug → task Paused
team.2bbd84…
✓13% ◐87% 8 runs · 7 in 24 h
#e2e#regression
You 20 hours ago
LIVE support bug triage Active
team.2bbd840c.cr…
✓97% ◐3% 136 runs · 0 in 24 h
#support#triage#ai
System 4 days ago
You can see what is breaking
Every automation carries its own success rate, run count and runs per day. A workflow at 13% success is visible before anyone complains about it.
A test before it goes live
The Test and Check buttons run the automation on real data with no consequences and show where it will fail.
Order among a hundred workflows
Groups, tags, state filters and search. Draft, paused and error are distinct states rather than a guess from the name.
Pause instead of delete
Any automation can be switched off with one toggle, keeping its run history. System ones are protected from accidental changes.
What stops being necessary
Vendor 1 and Vendor 2 · Workflow automation between your own systemsConnectors somebody has to maintainTriaging incoming requests by handReminders to “remember to create the task”Scripts running in the shadowsPaying per workflow step
The permissions you gave it
An automation acts within the permissions you gave it, and every step it takes is written to the audit log alongside people's actions. You can see exactly what the system did, when, and on the basis of which answer from the AI node.
Runs on the shared memory — Synapse
Every run leaves a trace: what fired, on which data and why. That becomes context for the automations after it. This is the same memory Synapse answers from and Command Center runs on. The memory comes with every paid plan, and its size is set by the plan you are on.
Bring the process you are tired of
We will take a routine your team does by hand every day and build it on the canvas during the demo — with a trigger on your entity and your own actions.
Use cases
Where Automations does its work
None of these workflows lives inside a single product, and that is why they work.