Customer Support

How to build a customer knowledge base as Small Customer Success Teams

Customer SupportFor Small Customer Success Teams2 apps12 steps~24 min to set up

Your 3-person CS team covers 250 B2B accounts, and your 'customer knowledge base' is currently a graveyard of Google Docs, a Notion wiki nobody updates, a Confluence page from 2023, and whatever lives in the head of the person who's been there longest. When a new customer asks how to set up SSO, someone spends 20 minutes hunting for the right doc before copying a Slack message they vaguely remember. When a CSM leaves, half the institutional knowledge walks out with them. You're answering the same onboarding questions through Intercom every week, and you have no single source of truth to point customers — or your own team — to.

Build this on Starch → See the apps used

Customer SupportFor Small Customer Success Teams2 apps12 steps~24 min to set up

Outcome

What you'll set up

A structured, AI-searchable customer knowledge base that surfaces the right answer from your existing docs, past tickets, and Notion pages — so your team stops hunting and starts answering in seconds

An automated intake pipeline that pulls recurring questions from Intercom and Gmail, clusters them, and flags gaps in your documentation so you know exactly what to write next

A living onboarding guide that updates itself as your product changes and can be searched by new CSMs, customers, or the Starch Customer Support Agent (coming soon) to resolve tickets without touching your team

The Starch recipe

Apps, data, and prompts

The combination of Starch apps, the data sources they pull from, and the prompts you use to drive them.

Apps used

Knowledge Management

Data sources & config

Starch syncs your Notion workspace on a schedule (pages, databases, users) and your Gmail inbox on a schedule (messages, labels, threads). Intercom and Zendesk connect from Starch's integration catalog — the agent queries them live when your knowledge base app runs to pull ticket history and recurring question clusters. Google Calendar syncs on a schedule to surface onboarding call notes tied to specific accounts.

Prompts to copy

Build me a customer knowledge base that pulls from our Notion workspace and Gmail inbox. Organize content by topic: onboarding, integrations, billing, and escalation paths. Surface the top 10 most frequently asked questions from the last 90 days of customer emails, then flag which ones don't have a matching doc.

Every Monday, scan the previous week's Gmail threads tagged 'customer' and Intercom conversations for questions we answered manually more than twice. Create a draft doc for each one and add it to the 'Needs Documentation' section of the knowledge base. Slack me a summary of what was added.

Build me an email triage view that separates inbound customer questions by urgency — renewals and escalations first, onboarding questions second, general requests third — and drafts a one-sentence summary of each thread so I can process the inbox in 15 minutes instead of 45.

Run these in Starch → or paste them into your favorite agent

Walkthrough

Step-by-step

1 Connect Notion to Starch — Starch syncs your pages, databases, and users on a schedule. This becomes the backbone of your knowledge base and the place new docs land automatically.

2 Connect Gmail — Starch syncs your inbox on a schedule. Tell Starch to look for threads where you've answered the same onboarding or product question more than twice in the last 90 days.

3 Connect Intercom from Starch's integration catalog. The agent queries your ticket history live to find recurring question patterns — no manual export needed.

4 Open the Knowledge Management starter app and tell Starch: 'Organize my existing Notion docs into four sections — onboarding, integrations, billing, and escalations — and flag anything that hasn't been updated in 90 days as potentially stale.' Starch builds the categorized view.

5 Run a gap analysis: 'Pull the 20 most common customer questions from Gmail and Intercom over the last 60 days and match each one to an existing doc. List every question that has no matching doc.' Use this list as your documentation backlog.

6 Build a weekly automation: 'Every Friday at 4pm, scan new Gmail and Intercom threads from this week. Draft a new knowledge base article for any question that came up more than twice. Add it to the Needs Review section in Notion and Slack me a summary.'

7 Set up the Email Triage app to process inbound CS email: 'Triage my inbox every morning and group customer emails by: renewal risk, onboarding question, escalation, and general request. Summarize each thread in one sentence and draft a reply for anything that matches a knowledge base doc.'

8 Build a new-CSM onboarding path: 'Create a guided onboarding sequence for new CS team members using our Notion docs. Surface the 10 most important articles in order and include a quiz question after each one based on the doc content.'

9 Wire in a customer-facing search layer: 'Build a search interface over our knowledge base that customers can use to answer their own questions. When no doc matches, add the query to a log so we know what to write next.'

10 Set a staleness alert: 'Check all knowledge base docs every two weeks. If a doc references a feature or pricing tier that no longer matches our current Notion product pages, flag it for review and Slack the team.'

11 When the Customer Support Agent launches (coming soon), wire it directly to this knowledge base so it can resolve common tickets 24/7 using your docs as the source of truth — no retraining, no separate setup.

12 Review the weekly Slack summary every Monday. Your documentation backlog is auto-prioritized by question frequency, so you always know which article will save the most time to write next.

See this running on Starch

Connect your tools, describe what you want, and the agent builds it. Closed beta is free.

Try it on Starch →

Worked example

Q1 2026 Onboarding Backlog Cleanup — 'TechFlow CS Team'

Sample numbers from a real run

Recurring questions answered manually in Q1	47
Questions matched to existing docs	19
Net-new articles drafted by Starch automation	28
Hours saved on manual email replies (est.)	34
Intercom tickets deflected after knowledge base went live	61

TechFlow's CS team of 3 was covering 250 accounts heading into Q1 renewals. Their Notion wiki had 40+ articles, but half were outdated and none were connected to what customers were actually asking. After connecting Notion and Gmail to Starch and running the gap analysis prompt, they found that 28 of their top 47 recurring customer questions had no matching doc — including 'how do I add a second admin user,' which had been answered manually 11 times in January alone. Starch drafted all 28 articles in a single Friday afternoon run, pulled from context in past Gmail threads and existing Notion product pages. Within two weeks of publishing the updated knowledge base to customers, Intercom ticket volume for onboarding questions dropped by 61 tickets in the first month. The team used the time saved to run QBR prep for their 40 enterprise accounts instead of answering the same SSO question for the twelfth time.

Measurement

How you'll know it's working

Time-to-first-response on inbound customer questions (target: under 2 hours for non-escalations)

Knowledge base deflection rate: percentage of Intercom/Zendesk tickets resolved without CSM involvement

Documentation coverage: percentage of top-50 recurring questions with a live, non-stale doc

New CSM ramp time: days from start to handling solo account load

Weekly repeat-question volume: how many customer emails this week asked something we've already documented

Comparison

What this replaces

The other ways teams handle this today, and how the Starch version compares.

Notion alone

Great for writing docs, but has no connection to your actual ticket or email data — the gap analysis, staleness detection, and auto-drafting have to be done manually every time.

Guru or Tettra

Purpose-built knowledge base tools with good search, but they don't connect to HubSpot, Gmail, or Intercom — you still have to manually identify what to write based on a gut feel rather than actual question frequency.

Gainsight or Catalyst

Full CS platforms that include knowledge management features, but they start at six figures and require a CS-ops person to configure — not realistic for a 3-person team.

Intercom Articles

Covers customer-facing docs well and integrates with Intercom tickets, but it's siloed from your internal Notion wiki, your Gmail history, and your HubSpot account data — you can't build a unified knowledge surface across all three.

On Starch RECOMMENDED

One platform — knowledge management, founder inbox all running on connected data. Setup in plain English; numbers stay current via scheduled syncs and live agent queries.

Try it on Starch →

FAQ

Frequently asked questions

We already have a Notion wiki. Does Starch replace it or connect to it?

Starch connects to it. Starch syncs your Notion pages, databases, and users on a schedule — your docs stay in Notion. What Starch adds is the automation layer on top: gap analysis, staleness detection, auto-drafting based on email and ticket patterns, and a searchable surface you can expose to customers or new team members.

Can Starch pull ticket data from Intercom to figure out what we're missing?

Yes. Connect Intercom from Starch's integration catalog and the agent queries your ticket history live when your knowledge base app runs. You can ask Starch to cluster recurring questions, match them against your existing docs, and generate a ranked backlog of what to write next.

What about Zendesk — we use that instead of Intercom.

Same story. Connect Zendesk from Starch's integration catalog; the agent queries it live. You'd use the same prompts — just swap Intercom for Zendesk.

Will the Customer Support Agent automatically use this knowledge base to answer tickets?

That's the plan. The Customer Support Agent is currently in development — you can request beta access now. When it launches, it will use your Starch knowledge base as the source of truth to resolve common tickets 24/7 without involving your team. Building the knowledge base now means you'll be ready to wire it in the day the agent is available.

We're not SOC 2 certified as a company — is Starch?

Starch is not SOC 2 Type II certified today. If your enterprise customers require vendor SOC 2 before you can share account data with a third-party tool, that's worth factoring in. For most early-stage CS teams covering 250 accounts, this isn't a blocker — but it's honest to name it.

What happens when a CSM leaves and their Gmail history goes with them?

If you've been running Starch's weekly automation — scanning Gmail threads and drafting knowledge base articles from repeated questions — most of the institutional knowledge buried in that inbox has already been extracted into docs before they leave. You won't capture everything, but you'll capture the patterns that matter most.

Can I expose the knowledge base directly to customers, or is this only internal?

You can build a customer-facing search interface in Starch by describing it in natural language: 'Build a search interface over our knowledge base that customers can use to answer their own questions.' Whether you embed that externally depends on your setup, but the surface itself is buildable today.