Objections and questions — Automation Orchestrator
What this audience actually asks, and answers grounded in what the repo really does — including the questions where the honest answer is "not yet".
Rules for using this:
- Answer the question that was asked, then stop. A long answer to a short question reads as evasion.
- If the answer is "it doesn't do that", say so first, then say what it does do. Never lead with the workaround.
- Everything here is checkable in a public repository. If you are not sure, say "let me check" and check — you can, live, in front of them.
"AAP already has workflows and approvals. Why this?"
AAP workflows chain job templates on success or failure — linear or branching, but always defined in YAML with a fixed graph. AO adds:
- A visual canvas — drag nodes, connect them, run the result
- Conditional logic nodes that branch on a previous step's output
- Approval nodes with configurable policies — who approves, how many, what happens on timeout
- A Temporal-based execution engine — workflow state is durable and survives pod restarts
- Cross-platform potential — the node type catalog includes HTTP request, script, and integrations beyond AAP
"If your workflows are linear chains of job templates, AAP's built-in workflow visualizer is enough. AO earns its place when the orchestration needs conditionals, human gates, or when the audience wants to see the sequence before trusting it."
Sources: GA blog, Add AAP to workflows, Node type catalog
"Do we have to migrate our job templates?"
No. AO connects to AAP as an integration and sees every job template the connected credential can see. On sandbox, that is 33 templates — every one this repo defines. No migration, no duplication. Your templates stay in AAP, where they are version-controlled and tested.
"AO orchestrates your existing templates. It does not replace them or copy them."
Source: GA blog
"Is it GA?"
Yes. Version 2026.8, generally available as an add-on to AAP 2.7 or later.
The May 2026 Red Hat Summit press release said "technology preview". The August GA release superseded that. If someone cites the press release, point them to:
- Release notes
- GA blog (2026-08-21)
"What does it cost to run?"
Measured on sandbox (cluster-kbjvc): 1.91 vCPU and 2.47 GiB of requested resources. Nine AO pods (backend x2, UI x2, worker x2, background-worker, temporal, redis) plus one PostgreSQL instance with a 10Gi PVC. That is roughly the footprint of two medium demo VMs.
Source: playbooks/install_ao.yml, measured in
#141
"Where does its data live?"
Its own CloudNativePG PostgreSQL cluster — three databases:
| Database | Purpose |
|---|---|
ao_backend |
Application state |
ao_temporal |
Workflow execution engine |
temporal_visibility |
Temporal's visibility store — fixed name, not documented in the CRD |
Deliberately not in AAP's database. AAP's postgres is owned by the AAP
operator with blockOwnerDeletion, so databases added to it live inside
something another operator recreates at will. Temporal is write-heavy, and
putting that load on the database the whole demo platform depends on trades a
working AAP for a working AO.
Source: playbooks/install_ao.yml header comment,
#141
"Where are the credentials?"
"Four credentials, all created by automation, none in a tracked file."
| Credential | Where it lives | How it is created |
|---|---|---|
| AO admin password | Kubernetes Secret ao-admin-password |
install_ao.yml — seeded from aap_password, so the two match |
| OIDC client | AAP OAuth2 application named "Syntara" | configure_ao.yml — POST setup_aap_oidc |
| AAP integration credential | AO's database | configure_ao.yml — POST /credentials |
| MCP server token | Operator's local Claude config (not tracked) | make-ao-mcp.sh — gateway personal access token |
Source: playbooks/configure_ao.yml, playbooks/install_ao.yml
"Is it really idempotent?"
The install and configure playbooks are. The one exception: POST
setup_aap_oidc returns 502 if the OAuth2 application "Syntara" already exists
on AAP. The playbook checks for an existing identity provider first and skips
the call — so a second run is safe, it just cannot self-heal a half-created OIDC
setup.
"If the OIDC setup is partially broken, delete the 'Syntara' OAuth application in AAP's gateway UI and re-run the configure playbook."
Source: playbooks/configure_ao.yml header, lines 29-31
"What happens when it breaks?"
The most common failure is the SSRF allowlist. AO uses langchain_core's
SSRF protection to block private IP addresses on integration URLs, and the AAP
hostname resolves to a private IP from inside the cluster. The configure
playbook patches APP_INTEGRATION_URL_ALLOWED_HOSTS on the ao-backend
deployment. If the operator reconciles and drops it, re-run the configure
playbook.
| Symptom | Fix |
|---|---|
| AO login page loads but "Log in with AAP" fails | Re-run configure_ao.yml — the SSRF allowlist was reconciled away |
| AO shows 0 job templates | Re-run configure_ao.yml — the integration credential or allowlist is stale |
| AO is unreachable (503 / no Route) | Re-run AAP Ecosystem - Deploy Automation Orchestrator — it converges |
| Database gone | oc delete namespace automation-orchestrator and re-deploy from scratch |
"Who can launch this?"
Anyone who can launch a workflow in AAP. The AAP Ecosystem - Deploy Automation
Orchestrator workflow needs the Sales Demos - Vault and Sales Demos - Env
Secrets credentials, which are scoped to the IT Service Automation
organization.
"Can I have it?"
"Yes. The playbooks are in a public repository."
The only prerequisites are an AAP 2.7 or later subscription and an OpenShift
cluster with the Automation Orchestrator operator available in the catalog.
playbooks/install_ao.yml and playbooks/configure_ao.yml are the whole
install. The operator is separately subscribed — "Red Hat Ansible Automation
Orchestrator" — but installs under the environment's existing pull secret with
no extra credential.
Questions to ask them
After the canvas beat:
- "What does your approval process look like today — Slack messages, email threads, or something else?"
Before the close:
- "Is the next conversation about orchestrating your existing AAP workflows, or about bringing in systems AAP does not manage today?" — this reveals whether the next meeting is technical (more integrations) or organizational (adoption). They are very different meetings.
Things not to say
- Do not say "replaces AAP workflows" — it orchestrates them, it does not replace them
- Do not say "AI-powered" unless they ask about the LLM integration — it is not wired up in this demo
- Do not compare to ServiceNow Orchestrator, Camunda, or Temporal directly — let them make the comparison
- Do not promise EDA integration — it exists in the interactive demo but is not wired up here
- Do not claim the login page can be branded — it cannot;
#426confirmed AO has nocustom_login_infoorcustom_logoequivalent