Loading...
Loading...
When no SOAR playbook matches an alert, it takes the platform's default path. Cortex XSOAR types it "Unclassified," Tines emits it labeled "No match," and Google SecOps documents a catch-all playbook that enriches it and asks an analyst, "Is this a True Positive?" A person investigates it, or an AI SOC Agent such as Simbian's does.
Your SOAR renewal deck will count what the playbooks automated. It won't count the alert no SOAR playbook was written for.
Follow one. It arrives, nothing matches, and on Google SecOps it lands in a catch-all that enriches it and asks a person one question. "Is this a True Positive?"
How many of your alerts take that road? No vendor can tell you. It matters, because SOC staff get to 49% of the alerts they are supposed to review in a working day, according to a 2023 Morning Consult survey of 1,000 SOC team members published by IBM.
When no SOAR playbook matches an alert, it goes to a default path and no use-case playbook runs on it. A SOAR playbook is a workflow written in advance for one kind of alert, such as phishing triage or disabling a compromised account. Four SOARs document when a playbook runs. Three say where everything else goes.
| SOAR | A playbook runs when | What happens to the rest |
|---|---|---|
| Cortex XSOAR | The classifier maps the event to an incident type with a default playbook | Unmapped events get the "Unclassified" incident type unless you choose another default |
| Splunk SOAR | The event's label matches the playbook's label | Playbooks "run only on containers that match their label." The docs don't say what happens to a label with no playbook |
| Tines | A Condition path matches | Unless you set emit_no_match, the event is still emitted, labeled "No match" |
| Google SecOps SOAR | An event-specific or product-specific playbook triggers | A lowest-priority catch-all, "a fallback for new or uncategorized alert types," enriches the case and asks an analyst |
None of the four documents dropping it. Unless you build more automation, it waits for a person.
You cannot automate when you are playing by someone else's rules, because there will always be new events you didn't prepare for.
More alerts miss every SOAR playbook than the renewal deck shows, because most teams write few playbooks. Forrester observed in 2022 that teams "often implement a maximum of five to 10 playbooks in total over the first several years of adoption," often kept alive by one analyst.
In the SOAR migrations we run, we typically see teams automating around 25% of their alert volume before they switch.
The playbooks you do have will drift. Rename one field upstream and an alert your playbook was written for can stop matching, often without an error, and land in the same bucket as a brand-new alert type. Quietly.
Your own number beats ours. Count 90 days of "Unclassified" incidents, "No match" events, unmatched Splunk labels or catch-all tags, and divide by total alerts. Then pull 20 and ask how many a playbook was meant to catch.
On the default path, SOAR automation stops at enrichment at best, because enrichment can be scripted in advance and investigation can't. Enrichment tells your analyst whether an indicator is known bad. Investigation tells them what happened on the machine, and each step depends on the last answer.
Google's catch-all playbook example enriches every entity, raises priority on a high threat score, then offers the analyst three answers. One is "Needs More Investigation."
That's good triage. Our CTO, Alankrit Chona, calls enrichment the fact sheet about the nouns, and says of it, "Almost everybody does this." Is this normal for this user? Has this IP shown up on other hosts? Did something you dismissed last month just become relevant? Every one of those questions goes back to your analyst, with the enrichment attached and the hard part still to do.
A playbook is an L1 with a runbook. Machines are better at following instructions than humans, and that's exactly what a playbook is good at.
The unmatched alert needs an L2, and in the same IBM study, 81% of SOC team members said manual investigation of threats slows down their overall response.
AI SOC Agents investigate an unmatched alert the way an L2 analyst does, choosing each next query from the evidence. That's automated alert investigation. Simbian's AI SOC Agent runs a query, reads the result and picks the next one, until it reaches a verdict or can say what it couldn't resolve. No playbook has to match first.
A global manufacturer's APAC engineering teams ran custom in-house applications that set off the EDR constantly. Almost every alert looked unique, so the working assumption was that most were true positives, and no analyst had written a rule or profile for them. Simbian's AI SOC Agent built one from region, team, a file path in a temp directory, a successful MFA login and no data leaving. After a couple of working sessions, it was identifying about 91% of the deployment's alerts as false positives.
An alert that fit only part of that profile and carried something new, such as a zero-day file or Microsoft infrastructure acting as command and control, stayed open. The Agent kept digging.
Every investigation ends in a Case Report that names what the Agent couldn't resolve, next to its step-by-step analysis and MITRE ATT&CK mapping. Then it proposes a response. We investigate and respond to every threat, on average under 4 minutes from detection to response, and 95% of the responses we propose are approved by the customer's own analysts. Their call, not ours.
When an analyst corrects a verdict, the correction can become a proposed change to how the Agent investigates, and a person approves it before it applies. That loop is part of what Simbian calls Self-Improving Defense. We never train on your data.
Today, every time you investigate a threat, you incur some cost. With Simbian, every time you investigate a threat, your cyber program becomes a bit more resilient.
No. AI SOC Agents work alongside the SOAR playbooks that match, which stay good at fixed, audited response steps like disabling an account. Start the AI SOC Agent on the alerts no playbook runs on today, the ones your SOAR sends to its default path.
Take your count and your 20-alert sample into the renewal, and propose a pilot on those alerts. Have every AI SOC vendor you evaluate, your SOAR vendor's own AI Agent included, investigate the same sample against your analysts' conclusions. Same alerts, same yardstick. Simbian's 90-day pilot validates against your senior analysts' verdicts before scaling, so start it at least 90 days before the renewal date.
Once the verdicts hold, our AI SOC vs SOAR comparison for SIEM alert triage covers what your playbooks become.
If the number comes back small, good news: your playbooks are doing their job, and that's a fine answer to take into the room.
Large is the interesting case. You've found the queue your renewal never priced. Count it, then book a demo and bring the number.
Q: What is a SOAR playbook? A SOAR playbook is a predefined workflow that runs a fixed sequence of steps when an alert matches the type, label or condition it was written for. Common SOAR use cases and SOAR playbook examples include phishing triage, indicator enrichment and disabling a compromised account.
Q: Why do SOAR playbooks break? A SOAR playbook matches on fields and conditions written in advance. When a log schema renames a field or a new tool starts sending alerts, the playbook stops matching without an error, and the alert goes quietly to the default path.
Q: Should I use a SOAR platform or an AI SOC platform? Use an AI SOC platform, such as Simbian's AI SOC Agent, for alert triage and investigation, and keep SOAR for the fixed, audited response steps you want run the same way every time, such as disabling an account. Whatever your SOAR sends to its default path is investigation work no playbook covers today.
Q: Do SOAR vendors' own AI Agents investigate unmatched alerts? Some do, with documented limits. Google SecOps' Triage and Investigation Agent works only on alerts ingested through Google SecOps SIEM. Splunk's ES 8.7 AI SOC Analyst needs ES Premier on Cloud and, per Splunk's setup documentation, is limited to about 200 findings per tenant per day.