Optimizing Fin AI Agent is key to maintaining peak performance and ensuring faster, more accurate AI resolutions. Watch the video for a quick overview of the top ten ways to optimize Fin then read below to put them into practice.
1. Use the same language as your customers
Go to AI & Automation > Content gaps to see which topics Fin tends to route to your team. Click any topic to find individual customer queries that Fin could not resolve.
The Content gaps feature groups unresolved customer questions into topics. These topics can be organized by language, volume, abandonment rate, or routed-to-team rate.
Check if customers are using the same language in their questions as you do in your help content. Make sure you're using similar words and phrases so Fin can be sure it’s on the right track and answer these questions effectively.
2. See which content Fin used unsuccessfully
Click on any unresolved question group and check the Related content tab.
Examine the content Fin referenced when answering queries on this topic. Add relevant details, clarify ambiguous text, and ensure images or other rich media are explained verbally as well. If the topic isn’t covered at all, you can create a new article, snippet, or Custom Answer to fill the gap so Fin can handle these questions without help from your team.
Focus on the most specific question groups, e.g. "Project duplication issues" would help highlight specific issues customers are having with this feature, vs "Task management" which could be a broad area and involve more manual investigation to drill down into conversations and find content gaps.
3. Dive into Fin conversations in the inbox
In the Inbox, you can click Improve Answer on any AI-generated response to review the content Fin used to create it.
Even if the response successfully resolved the issue, you can enhance it by adding more information or clarity. This helps Fin perform better and deliver more accurate answers in the future.
4. Foster a QA culture among your team
Consider establishing a feedback loop that allows your support team to flag inaccuracies, suggest improvements, and request new content to address knowledge gaps.
One way to achieve this is by creating a dedicated Back-office ticket type, enabling teammates to submit these requests directly from the inbox. If Fin provides an unhelpful answer, teammates can review the source and open a ticket to update the content without leaving the inbox or losing valuable time.
Another option is to ask your teammates to write snippets immediately to address content gaps or update information. Your knowledge manager or team can then review these snippets regularly, incorporating them into main articles and other resources.
These approaches not only streamline the process but also enable teammates to proactively contribute to improving content and Fin’s performance.
5. Check your Fin involvement rate to see how often Fin takes part in conversations
To see a Fin AI Agent report, go to Reports and create a new report, then select the Fin AI Agent template to get started quickly.
The Fin AI Agent involvement rate chart shows the percentage of your total conversations where Fin participated.
You can boost this rate by getting Fin involved in more conversations. Review your Fin workflows to ensure you're reaching a broader segment of your customer base across all channels, including email. This helps increase Fin's impact, ensures more customers benefit from AI assistance, and drives ever greater efficiency for your team.
Custom reports also allow you to track key metrics such as “Fin AI Agent unresolved conversations” and segment data by "Topics". This is useful if you’ve already set up conversation topics and you want to segment your reports by these to get insights into specific product areas.
Segmenting conversations by topics helps you to understand where you can make the most impactful improvements by filling content gaps and using Fin to handle more of these topics. If you haven’t set up topics yet, consider creating them based on FAQs or specific product areas to improve categorization and management.
Use the chart drill-in feature and export chart data to assess conversation transcripts for deeper analysis.
6. Keep a close eye on your resolution rate
The Fin AI Agent resolution rate chart indicates how often Fin’s answers lead to a resolution, offering insight into the overall health and accuracy of your help content.
If this rate is steadily increasing, you're on the right track. However, if you notice a decline, it might be time to conduct a knowledge audit to identify gaps or inaccuracies and improve the quality of your content.
7. Review negative Fin CSAT feedback
It's encouraging to see positive feedback from customers about their experience with Fin, but you can often learn more from negative feedback. Drill-in to the 😠 Terrible and 🙁 Bad remarks to identify where Fin's interactions went wrong and address the root cause.
This is an effective way to pinpoint issues that might not be related to content, but rather to Fin's settings or workflow setup, such as routing or targeting problems. By addressing these areas, you can improve Fin's overall performance and customer satisfaction.
Note: You'll need to have Fin AI Agent CSAT enabled to see this reporting.
8. See which content Fin is using but is not driving resolutions
Within the Fin AI Agent legacy report, you can view Content performance and check the resolution rate of your content. The "involved" column shows how many times each piece of content was used by Fin. The "resolved" column shows how many times each piece of content has effectively answered customer questions.
Note: The content performance chart is currently only available in the legacy Fin AI Agent report. This will be made available in new Intercom reporting soon.
Filter by Fin AI Agent involved conversations to identify the content referenced most frequently. If the number in the Fin AI Agent resolved conversations column is significantly lower for a piece of content, it means the information Fin is using is not providing satisfactory answers to customers.
Examine the underperforming content for missing, inaccurate, or ambiguous information and revise where necessary, empowering Fin to tackle customer issues more effectively.
9. Use Custom Answers when exact phrasing is critical
Sometimes you want Fin to provide a carefully prepared response rather than generate its own answers. That’s where Custom Answers come in.
When a customer asks a question for which you have a Custom Answer, Fin will always deliver the Custom Answer, bypassing any AI-generated response. This ensures consistency in how Fin handles certain queries.
Here are a few examples of topics where Custom Answers can be useful:
Legal disclaimers: Terms and conditions, privacy policies, and liability disclaimers where deviation in language could create loopholes or expose a company to legal risks.
Health information: Instructions regarding medicines, supplements, or allergies should be provided consistently to avoid misinterpretation.
Safety instructions: Guidelines for operating products or handling some materials must be provided exactly as specified to prevent accidents or injuries.
Product warnings and usage instructions: Warning labels on products, especially those involving electrical equipment or hazardous chemicals, need to be clear and precise to ensure consumer safety.
10. Check which Custom Answers Fin is not using
Under AI & Automation > Custom Answers > Live, you will find a list of your live Custom Answers. Filter by Sent, and find which Custom Answers are not being sent to customers at all. They might need more training examples so Fin knows when to use them.
Additionally, check the Underperforming tab to see which Custom Answers Fin is using but that rarely result in an AI resolution. Ensure the information is clear and up to date, or create new answers to cover customer needs.
Follow these tips to keep Fin optimized and drive positive results for teams, Fin, and customers alike. 😃
Need more help? Get support from our Community Forum
Find answers and get help from Intercom Support and Community Experts