Setting your success benchmarks after onboarding to Intercom
Getting Fin live is one thing—keeping it performing is another. Set a benchmark from your own data and follow a 90-day plan to keep improving.
Getting Fin live and keeping it performing well are two different challenges. How do you know when it’s actually working? What follows is a practical benchmark you can set from your own data, a 90-day plan to reach it, and the patterns that separate support teams that hit their stride from those that stall.
Prerequisites
To get the most from this lesson, a couple of things should be in place.
- Fin is deployed: Fin AI Agent should be live and responding to conversations on at least one channel. If you haven’t set this up yet, the Setting up your AI-first helpdesk course covers the full setup.
- Key metrics familiarity: This lesson uses Automation Rate, Involvement Rate, and Resolution Rate throughout. You should have a working understanding of each. The Monitoring Fin’s performance help article covers all three.
Where to begin
Your benchmark starts with your own baseline. After Fin has been live and involved in conversations for about a week, open your Performance dashboard (Fin AI Agent > Analyze > Performance) and record your Automation Rate.
That number is the baseline you’ll measure all progress against, and your goal is sustained, meaningful improvement from it over the next 90 to 120 days.
What is Automation Rate? The share of conversations Fin is fully resolving without human involvement, and it’s the clearest signal of whether your AI-first operation is working.
Why might my rate look low? Higher-volume operations will naturally see lower percentages because of conversation complexity: more edge cases, more languages, more queries that genuinely need a person. That’s expected, and it doesn’t mean something is wrong. What matters is that your rate is climbing steadily over time.
[IMAGE PLACEHOLDER: Performance dashboard showing Automation Rate, Involvement Rate, and Resolution Rate. Path: Fin AI Agent > Analyze > Performance]
Why day 120, not day 30?
It’s tempting to check your numbers in the first week or two and draw conclusions, but early data can be misleading.
Automation rates take time to settle. Week-one data reflects the novelty of Fin encountering your conversations for the first time, and month-one data is better but still early. By day 90, you have enough volume, enough improvement cycles, and enough content refinement to see a real signal, and day 120 confirms whether that signal holds.
What you do in the first 90 days largely determines where you land, so the work is front-loaded even though the measurement needs more time to catch up to the effort. Give yourself that runway before you benchmark.
What sets successful operations apart
Across thousands of Fin deployments, a clear pattern emerges. The teams that reach strong automation rates do a handful of things differently from those that plateau early.
| Markers of success | What tends to hold teams back |
|---|---|
| A broad content library at launch: articles, snippets, and external sources all available to Fin | A handful of articles and nothing else for Fin to draw from |
| Fin actively involved in conversations, not just turned on | Fin is live but barely touching conversations because it hasn’t been deployed across the right channels or audiences |
| Escalation rules configured before launch so hard-stop topics always go to a human | Fin attempting everything, including topics that should be routed to a person |
| Channels expanded progressively: Messenger first, then email, then others as confidence grows | All channels activated at once, or only one channel with no plans to expand |
| Weekly reviews of what Fin couldn’t handle, with content fixes published in response | Performance dashboard checked once a month, or not at all after launch |
| Resolution Rate and Involvement Rate checked separately to identify the right fix | Adjusting Guidance when the real gap is a content or coverage problem |
| Multiple teammates active in the workspace and reviewing performance | One person configures Fin and nobody else engages with it |
None of these are complicated on their own, and the pattern that cuts across every team that stalls is simpler than any single row in that table: the improvement loop stops running. Teams configure Fin, launch, and treat the setup as done.
The three levers (and when each one matters)
There are three things that move your Automation Rate, and they matter in a specific order.
Content breadth is your first lever, and it matters most in your first 30 days. The more Fin has to draw from, the better it performs, so a broad library of articles, content snippets, and external sources from the start gives Fin a real foundation to work with. If Fin can’t find the answer, it can’t resolve the conversation.
Fin’s reach is your second lever, and it matters most between days 14 and 60. The most common way to stall is having Fin turned on but not involved in enough conversations because it hasn’t been deployed across the right channels, audiences, or conversation types. Your Involvement Rate (Fin AI Agent > Analyze > Performance) tells you whether Fin is actually in the game, and if it’s low, the fix isn’t better content, it’s broader deployment.
Resolution quality is your third lever, and it matters most from day 30 onward. By month 2, the question shifts from “is Fin reaching conversations” to “is Fin resolving them,” and your Resolution Rate tells you what share of conversations Fin successfully handles when it does get involved. That’s where reviewing what Fin couldn’t resolve pays off, because each content gap you fix is a set of conversations Fin can now handle instead of escalating.
The progression is worth internalizing: build the knowledge base so Fin has something to work with, then make sure Fin is reaching enough conversations, then refine so Fin resolves more of what it touches.
What stalling looks like
Three patterns to watch for, each pointing to a different fix:
- The content gap shows up when Fin doesn’t have enough to work with. Resolution Rate is low because answers don’t exist yet, not because Fin is doing something wrong, and the fix is to review conversations Fin escalated, identify the missing topics, and publish the content to fill those gaps.
- The involvement gap shows up when Fin is on but barely touching conversations. Automation Rate is low because Fin isn’t involved rather than because it can’t resolve, and the fix is to check your Involvement Rate and expand Fin’s reach across channels, audiences, and conversation types.
- The loop stopping shows up when everything was set up well but nobody is reviewing or refining. Automation Rate climbs in the first few weeks and then plateaus, and the fix is to reestablish a weekly review cadence, even if it’s just 30 minutes reviewing escalated conversations and acting on the top gaps.
Knowing which pattern you’re in tells you where to focus, and checking Resolution Rate and Involvement Rate separately is what gets you there.
Your 90-day milestone plan
Days 1 to 7: Observe
Your only job this week is to let data settle, so resist the urge to optimize before you have a clear picture.
- Confirm Fin is live on at least one channel
- Pin the Fin AI Agent inbox view and check it daily (Fin AI Agent > Inbox)
- Review a sample of conversations Fin escalated to start spotting common content gaps
- At the end of the week, once the Performance dashboard has populated, record your baseline Automation Rate, Involvement Rate, and Resolution Rate (Fin AI Agent > Analyze > Performance)
Data from the first few days is too early to act on, so let it settle before drawing any conclusions about how Fin is performing.
Days 8 to 30: First improvement cycle
Content is your fastest lever at this stage.
- Publish fixes for the top 3 to 5 content gaps you spotted in escalated conversations. If you have the Pro add-on, Recommendations (Fin AI Agent > Analyze > Recommendations) will surface these automatically.
- Review your Escalation Rules and confirm hard-stop topics are routing to a human (Fin AI Agent > Train > Escalation)
- Check Resolution Rate at the end of each week to confirm it’s moving upward
- Activate a second channel if applicable, since email is typically next after Messenger
Days 31 to 60: Coverage expansion
Now it’s time to build depth and widen Fin’s reach across your operation.
- Confirm you have a broad content library with articles, snippets, and external sources enabled for Fin
- Review Involvement Rate and confirm Fin is active across all eligible channels and conversation types
- Establish a weekly review rhythm of 30 to 45 minutes. With the Pro add-on, Recommendations and Topics Explorer surface gaps and trends automatically. Without Pro, review a sample of escalated conversations each week and look for repeated topics Fin couldn’t handle.
- Record a CSAT baseline for comparison at Day 90 from your reports or CX Score if you have access to Pro add-on.
Days 61 to 90: Benchmark approach
Your metrics should be trending in the right direction, and this phase is about confirming that trajectory and making sure the improvement loop is sustainable.
- Confirm Automation Rate is trending upward from your Week 1 baseline
- Confirm Involvement Rate is healthy, meaning Fin is reaching a meaningful share of your conversations
- Confirm Resolution Rate is healthy, meaning Fin is resolving a good share of what it touches
- Make sure your are running your review cadence as regularly as possible to achieve the best results
Day 90: Benchmark check
Run your Performance report for the full 90-day period (Fin AI Agent > Analyze > Performance).
- Compare your current Automation Rate to your Week 1 baseline
- If you’re not where you want to be, identify which lever needs work by asking whether the gap is in Involvement Rate or Resolution Rate
- A low Involvement Rate means Fin needs broader reach, so expand channels, routing, and conversation types or more audiences
- A low Resolution Rate means Fin needs better content, so continue reviewing escalated or non resolved conversations and pub
- lishing fixes
- Schedule your day 120 check, since day 90 is the early signal and day 120 confirms whether it holds
After day 90
The 90-day plan ends, but the improvement loop doesn’t.
Your operating rhythm from here is to review what Fin escalates on a weekly basis, act on content gaps before they accumulate, and check Involvement Rate quarterly as your support operation grows. Day 120 is your first formal benchmark check, and then quarterly after that.
The support teams that reach strong automation and keep it there all share the same habit: they keep showing up, reviewing what Fin couldn’t handle, fixing the content, and expanding coverage. The product does more of the work over time, but only when someone is steering.
