Light evidence (grade D: anecdotal)Short social post by the company's CEO; metrics aren't defined and couldn't be checked.

A SaaS Team Took Intercom Fin From 20% to 85% Resolution by Rebuilding Its Knowledge Base

HeyReach's new support lead rebuilt team processes, rewrote the knowledge base with 30+ new articles in two months and widened what Intercom Fin resolves alone. Three months on, the CEO reports resolution up from 20% to 85% and CX score from 60% to 100%.

Who this is about: LinkedIn outreach SaaS (HeyReach; 7,500+ customers per its CEO's bio)

Intercom Fin StartupsAutopilot

Get the prompt

SupportAdded Oct 16, 2026Source: x.com

What it takes

Cost: PaidFin is billed per outcome on Intercom plans, per Intercom's pricing page; the post gives no costs.
Setup time: A few weeksOur estimate from the source, for one business.
Skill needed: No codeApps and settings only.

Explain it simply

A software company was paying for an AI support agent but barely using it. A new support lead rewrote the help articles the AI learns from, changed how it handles tickets and let it solve more on its own. Within three months it was solving most conversations and customer ratings went up.

New to AI? See how AI works

What was done

When a new lead took over support in May, Fin sat at about 20-25% resolution, satisfaction was 60%, the team was demotivated and two senior people had left.

He set new team processes and encouraged the team to use AI without fear.

He rewrote the whole knowledge base, adding 30+ articles in two months, changed how Fin handles tickets and expanded what it resolves on its own.

How it went

Three months later, CEO Nick Velkovski reports resolution at 85% (from 20%) and CX score at 100% (from 60%), with more of support automated and the team no longer buried in tickets.

Not promotional: the CEO is praising his own team, and HeyReach sells LinkedIn outreach software, not support tools.
A short social post: how resolution and CX score are measured isn't stated, and he gives the starting rate as both ~20-25% and 20%.
The team changed at the same time (new lead, two senior departures), so the gain isn't down to Fin alone.

Prompt

PromptWritten by our editors from the source

Here are last month's support conversations that our AI agent handed to a human: [export].

Group them by reason (missing article, outdated article, needs account data, needs an action, policy question). For each group, give the count and two examples.

Then draft outlines for the 10 help articles that would let the agent resolve the most of these on its own. Flag anything that needs a human decision instead of an article.

Setup notes

Start with the knowledge base: list the questions the AI hands off and write or fix the articles behind them.
Widen what the agent may resolve step by step, and watch satisfaction as you go.
Agree with the team that the AI is there to help them, not replace them; HeyReach's turnaround started with the team.

Business fit

Who this helps, and how:

SaaS support teams: get more out of an AI agent you already pay for by fixing the content it answers from.
Startups
Stores with a help center: rewrite stale help articles so the chat AI resolves more on its own.
Shopify & ecommerce

Similar tools

Other tools for this job, and how they differ. Chosen by our editors, never paid for.

Gorgias
Ecommerce helpdesk with an AI Agent billed per resolved conversation; the closer fit for Shopify stores.
Grok Bot
Drafts replies and acts after your approval instead of resolving tickets on its own.
See it in use: Run customer support with a bot that asks before refunding
Zendesk AI agents
AI agents built into Zendesk; the comparable option if your team already works in Zendesk.

“Light evidence” (grade D) means: A brief, early or in-progress claim. Treat it as an idea to test, not proof. How we grade. Spotted a mistake? Suggest a correction.