Guide Fin's responses
Learn how Guidance enforces your policies and shapes how Fin responds in real conversations.
Every support team operates with shared expectations about what matters most. Some topics require precision, others demand sensitivity, and some conversations need to move quickly while others require care. Guidance is how you make those expectations explicit for Fin, defining clear rules that keep responses consistent, aligned with your standards, and scalable as volume grows.
Prerequisites
This lesson is part of the Launch Fin course. Completing the full course will give you the context you need to understand why content quality matters—and how to improve it effectively.
Before starting this lesson, make sure you’ve completed this key step:
- Add content to Fin: Add, manage, and enable content sources (such as articles, snippets, and integrations) so Fin can generate accurate, high-quality answers based on the right information.
You can learn how to do this in the Launch Fin course.
Video transcript
Knowing the right answer doesn't make you a great support agent. That's because great support is often about asking the right questions, phrasing things carefully, and knowing how to handle those sensitive situations. Now your team, they already know how to do this. They have procedures to follow and their own experience and judgment to rely on. And while we've intentionally trained Fin to be the best AI for customer service out of the box, you do still need to teach it your specific policies. And in general, how to behave in your unique environment. And that's what guidance is for. It lets you set clear guardrails for Fin so it responds consistently and behaves just like your best agents would. Guidance is organized into categories that control different aspects of Fin's behavior. And within each of these, we can create rules for Fin to follow. For example, in communication style, we could add a rule so that Fin always uses US spelling, ensuring it's consistent with our brand. For context and clarification, we could tell to clarify what device our customer is using before they answer. In content and sources then, we could say that when it comes to pricing, we always want Fin to use a specific article to answer. We can use the spam category to set boundaries, like telling Fin to ignore promotional messages so it stays focused on real conversations. And the last category is other. Now this is where we can put anything that doesn't neatly fit into the others. This covers everything from small tweaks to nonnegotiable rules that you want Fin to follow, like never ever give legal advice. Okay. So those are the guidance categories, and together, they help shape how Fin behaves in real conversations. Let's take a quick look at the product. As we start creating guidance, we're not gonna get it right on the first go, and that's fine. We don't have to. We can start by just roughly saying what we want Fin to do, and then just hit the optimize button. Now optimize helps us take that rough intent and turn it into guidance that's clearer, more precise, and easier for Fin to apply. And this becomes even more important as we build things out. As we add more rules over time, it's gonna get harder and harder to keep track of how they all interact. Now optimize acts like a safety net. It's catching overlaps and conflicts so Fin can stay consistent. Okay. There are a few more things we're showing here, and the first one is templates. Templates give us a head start. They're prewritten examples of common guidance that many teams use. We can use them as is, tweak them to fit our setup, or just use them as inspiration. They're especially helpful early on when we're trying to figure out what kind of rules make sense for our team. Next up is audience targeting. Now not every rule needs to apply to everyone. By clicking here, we can target guidance to specific audiences, like certain plans, regions, or channels. This lets us tailor how Fin behaves based on who it's talking to. Now the last thing that I wanna show you over here on the right is the preview and the event log. Now preview lets us test Fin's responses in real time. So we can actually see how our guidance actually affects the answer. And the event log shows us exactly which guidance was applied and when. So we're not guessing whether the rule actually worked or not. And that's pretty much everything you need to know. Just like training a teammate, guidance is something that you refine over time. So remember to start small, use those templates to get going, rely on that optimize button to keep everything aligned, and spot check your work with preview as you build.
Guidance writing tips
- One rule, one behaviour: Each guidance rule should focus on a single action or decision. When you combine multiple behaviours into one rule, it becomes harder for Fin to apply it consistently.
- Address Fin directly: Guidance should be written clearly and directly, and addressed to Fin. If a rule would feel vague or confusing to a team member reading it, it’s likely to be confusing for Fin as well.
- Be explicit about what good looks like: Don’t just describe what to avoid, describe the intended behaviour, give examples, clarify tone, and show Fin exactly what you expect.
- Start broad, then refine: You don’t need to anticipate every edge case upfront. Start with a small set of clear rules, then improve them based on real conversations and analytics.
Guidance is never "done"
There’s no final version of guidance. Build enough to start confidently, then iterate based on what you see in practice.
Downloadable resources:
Check out the following industry-based guidance examples to help get you started:
