In my last post about plain language automations I mentioned you stay in charge of how much latitude Flow by ServiceNow has. Today I’m covering what that means: how Flow answers questions, and who decides what information it works from.
“What are the company holidays?” The right answer is whatever your policy says, and Flow can only know that if someone connects to where it is documented. Flow draws on the best practices and policies, often known as knowledge, that you connect. Flow can also be configured to pull from general publicly available knowledge, where available. More to come on that. Your teams guide how it uses them, and when the answer isn’t in those knowledge sources, Flow says so and helps take next steps.
Flow works from the knowledge you already have
Flow can draw on Confluence, SharePoint, Google Drive, and ServiceNow Knowledge Base. You don’t need a dedicated knowledge base to begin. Flow works from the content your organization already maintains, and it syncs on a schedule or on demand.
Administrators connect knowledge sources and tie them to specific teams, so each one's Flow works from the content it needs. HR gets HR articles. IT gets IT’s. When a source contributes to an answer, Flow shows source cards beneath it. Open one and you land on the document itself, so anyone can verify the answer. These are the receipts.
Search respects the access of the person asking. Connecting a source doesn’t make every document visible to everyone. The ServiceNow Knowledge Base connection shows how this works: Flow uses supported ServiceNow user criteria to identify who can search an article. If Flow can’t safely determine who may read an article, it doesn’t make that article available.
An answer is only as current as the last sync of its source, which is one more reason to keep your documents up to date.
Some questions need a source and some don’t
Not every question deserves the same treatment. Here are a few examples:
| Question | Needs a documented source? | Why |
| “What are my company holidays?” | Yes | The answer is company policy, found in the employee handbook |
| “Am I eligible for a company phone?” | Yes | Varies by individual or role, so Flow won’t assume it unless it’s in the knowledge |
| “How does VPN work?” | No | General knowledge, not company policy |
| “How do I change my Zoom background?” | No | General how-to that doesn’t depend on your documents |
Different questions call for different answers. A question about company policy needs a documented source behind it. A general how-to may not. Neither is wrong. The questions are just different. And Flow automatically determines the right path using the configuration settings and context clues. If Flow isn’t giving the right answer, go back and update the source document.
Here’s what setting up knowledge behavior in Flow looks like:
Teams guide how Flow uses knowledge
Teams can do this through Agent Behavior, where they publish instructions that guide how Flow uses knowledge in conversations. Instructions can be set for the whole organization or for a single team, and they take effect once published. For instance, you can tell Flow to only use your company documents and not apply general or publicly available knowledge. HR policy questions are a good fit for this; HR policies often require a knowledge source, so public knowledge should be disabled for HR questions.
When Flow can’t confirm an answer, it says so
Flow doesn’t automatically know every company policy, and it doesn’t pretend to. If a policy isn’t documented, sits outside the asker’s access, or is missing from your connected sources, Flow can’t confirm the answer. It doesn’t treat an undocumented policy as fact.
That’s the point. A policy question deserves an answer someone can stand behind, and a confident wrong answer about parental leave is worse than “I can’t confirm that yet.” A “can’t confirm” also tells you where a document is missing. The answer lives with the policy owner, and once it’s documented, Flow can confirm it for future questions. The more complete your connected documents are, the more Flow can confirm. Flow offers next steps if relevant, such as escalating and alerting the Flow agent. After resolution, Flow offers an automation setup to the agent answering the question, so no human involvement is needed next time around.
What Flow does when it works from your documents
| If someone... | Flow should... |
| Asks something your documents answer | Answer, and point to the document |
| Asks something your documents don’t cover | Say it can’t confirm, not invent an answer, and offers next steps, if relevant |
| Asks something again after a fact in a document has changed and the source has synced | Give the updated answer |
That last row is why a knowledge base is worth maintaining. Flow follows your documents.
Flow will help you build your knowledge over time
You choose the documents Flow draws on, your teams guide how it uses them, and Flow is upfront when a document is missing. You can start without any knowledge sources and add them as you go.
For the latest product updates and direct access to the team building Flow, visit the Flow Community.
FAQ
Does Flow need a knowledge source to get started?
No. Knowledge sources are optional and can be configured later. Adding them lets Flow respond with information specific to your company and processes.
Can different teams use different sources?
Yes. Knowledge sources are bound per team, so HR and IT can each work from their own content.
What does Flow do when it doesn’t know?
It says it can’t confirm the answer. Flow doesn’t treat an undocumented company policy as fact. Once the policy is documented in a connected source, Flow can confirm it for future questions.
What if my documents are out of date?
An answer is only as current as its source’s last sync. Syncs run on a schedule or on demand.
Can Flow search the web?
Optionally. Web search is an option, where available for your deployment, and it’s separate from your connected sources.
