Customer Support

How to build a customer knowledge base as Small HR Teams

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

Your 150-person company has HR knowledge scattered across a Notion that was last updated in Q3 2023, a Slack channel where someone pinned the PTO policy six months ago, a Google Doc with the onboarding checklist, and your own head. Every week you answer the same questions: how does the 401k match work, what's the parental leave policy, who do I contact about a benefits change, how do I submit a reimbursement. You answer them in DMs, in #general, on calls. There's no single place employees can go to find authoritative answers, and when the policy changes — because Paylocity updated the accrual rules or you switched dental carriers — you have to hunt down every place it was written and update them all, one by one.

Build this on Starch → See the apps used

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

Outcome

What you'll set up

A searchable company knowledge base that pulls from your Notion, Gmail, and connected HR systems — so employees find answers themselves instead of Slacking you

An AI-powered document that detects when HR policies go stale relative to your actual payroll or benefits data, and flags what needs updating

An onboarding path builder that generates role-specific reading lists and checklists for new hires, so you stop rebuilding the same Google Doc every time someone joins

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 Email Agent

Data sources & config

Starch connects directly to Notion (scheduled sync) and syncs your pages and databases on a schedule so the knowledge base is always pulling from current docs. Gmail is also connected via scheduled sync so email threads with policy context can inform the knowledge base. Slack is connected from Starch's integration catalog; the agent queries it live to deliver staleness alerts. BambooHR and Rippling are connected from Starch's integration catalog; the agent queries them live to cross-reference employee data when building onboarding paths. Paylocity syncs directly to Starch on a schedule for payroll and benefits data.

Prompts to copy

Build me a company knowledge base connected to our Notion workspace. Organize content into sections: Benefits & Compensation, PTO & Leave, Onboarding, Performance, and Company Policies. Surface an AI search bar so employees can ask questions in plain English and get answers pulled from our actual docs.

When an employee asks a benefits question over email, draft a reply that pulls the answer from our knowledge base and flags me if the question doesn't match anything in our documented policies.

Build me a staleness detector: every Monday, compare our Notion HR policy pages against the last-updated date and flag any page that hasn't been edited in 90+ days with a Slack notification to me.

Create an onboarding reading list generator: given a new hire's role and start date, pull the relevant policy docs from our knowledge base and output a structured 30-day reading plan.

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

Walkthrough

Step-by-step

1 Connect Notion to Starch (scheduled sync) — Starch will pull all your HR pages and databases automatically. This takes about five minutes and requires no API key; you authenticate through Notion's standard OAuth.

2 Connect Paylocity to Starch (scheduled sync) — Starch syncs employee records, benefits enrollments, and payroll data on a schedule so your knowledge base can reference live policy data, not a screenshot from last year.

3 Connect Slack and BambooHR (or Rippling) from Starch's integration catalog — the agent queries both live when building onboarding paths or pushing policy alerts.

4 Open Starch and describe your knowledge base: tell Starch the sections you want (Benefits, PTO, Onboarding, Policies), the Notion databases to pull from, and what tone to use in AI-generated answers. Starch builds the surface.

5 Run the Knowledge Management starter app as your starting point — fork it, rename the sections to match your actual Notion structure, and add your HR-specific categories. You don't start from a blank page.

6 Set up the staleness detector automation: describe the rule to Starch (any Notion HR doc not updated in 90 days gets flagged), set the schedule to Monday morning, and tell it to Slack you a list with direct links to the stale pages.

7 Build the onboarding path generator: describe it as 'given a new hire's role from BambooHR and their start date, pull matching policy docs from Notion and output a 30-day reading plan grouped by week.' Starch assembles this as a custom app.

8 Wire the Email Agent to your Gmail: when benefits questions come in, the agent drafts replies sourced from your knowledge base and flags any question it can't answer confidently so you review before sending.

9 Test the AI search bar with the ten questions you get asked most often — 401k match structure, PTO carryover rules, reimbursement process, parental leave eligibility. Tune the source documents until answers are accurate.

10 Share the knowledge base link with your team and pin it in Slack. Tell employees to search there before DMing you. Track whether inbound HR DMs drop over the next 30 days.

11 Schedule a monthly 15-minute review where you check the staleness alert log and update whatever flagged. The system tells you what's overdue; you don't have to remember to check.

12 As policies change — new dental carrier, updated accrual rules, revised parental leave — update the Notion source doc and Starch re-indexes it automatically on the next sync cycle. One edit propagates everywhere.

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

Open Enrollment 2026 — Benefits FAQ Overload

Sample numbers from a real run

Benefits questions via Slack DM (2-week enrollment window)	47
Average minutes to answer each manually	8
Total HR time spent answering repetitive benefits questions	376
Questions answered by knowledge base AI without HR involvement	39
Questions escalated to HR (nuanced / edge cases)	8
HR time spent after knowledge base went live	64

During October open enrollment, your two-person HR team fielded 47 Slack DMs over two weeks — almost all of them variations of the same six questions about the new dental carrier, the HSA contribution limits, and whether part-time employees qualify for the 401k match. Before Starch, you spent roughly 8 minutes per message finding the right doc, pulling the accurate number, and writing a response that wouldn't create a compliance problem. That's 376 minutes — nearly two full work days — on questions that were all answerable from the benefits guide you'd already written. After setting up the knowledge base with your Notion benefits docs and Paylocity benefits enrollment data wired in, 39 of those 47 questions were answered directly by the AI search bar or the Email Agent draft. The 8 that reached you were genuinely edge cases: a contractor asking about COBRA eligibility and two employees with mid-year life events that changed their tier. You spent 64 minutes instead of 376. The staleness detector also caught that your HSA contribution limit doc still showed 2025 IRS limits — you updated it before anyone cited the wrong number.

Measurement

How you'll know it's working

Inbound HR DMs per week (target: reduction over 60-day baseline after knowledge base launch)

Percentage of benefits/policy questions self-served vs. escalated to HR

Number of stale policy documents flagged and updated per month

Time-to-answer for new hire onboarding questions (days from start date to first resolved question)

Onboarding completion rate for 30-day reading plan (tracked per new hire cohort)

Comparison

What this replaces

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

Notion alone (no AI layer)

Notion stores your docs but can't answer employee questions in plain English, can't detect stale content automatically, and can't pull from Paylocity data — so the maintenance burden stays entirely on you.

Guru or Confluence

Purpose-built knowledge bases with good search, but they don't connect to your Paylocity or BambooHR data, don't auto-detect policy staleness against live HR data, and require a separate tool budget on top of what you already pay for Notion.

Lattice or 15Five knowledge modules

Tied to your performance management workflow, not your full HR policy surface — great for review documentation, poor for benefits FAQs, PTO rules, or cross-system onboarding paths.

ChatGPT or Claude with uploaded docs

Fast to spin up for one-off questions, but there's no scheduled sync to keep answers current, no Slack or email integration, no staleness detection, and you have to manually upload new versions every time a policy changes.

On Starch RECOMMENDED

One platform — knowledge management, email agent 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

Does Starch replace Notion, or does it sit on top of it?

It sits on top. Starch syncs your existing Notion pages and databases on a schedule and builds an AI search layer over them. You keep editing in Notion the way you always have — Starch just makes that content findable and actionable in ways Notion can't do on its own.

What happens when a policy changes in Paylocity — does the knowledge base update automatically?

Paylocity data syncs to Starch on a schedule, so any changes to employee benefits records or payroll rules will be reflected in the next sync cycle. Your Notion docs still need a human edit when the policy itself changes — Starch will flag them as stale if they haven't been updated recently, but it won't rewrite policy docs on its own.

We use BambooHR, not Paylocity or ADP. Can Starch still connect to it?

Yes. BambooHR is available through Starch's integration catalog; the agent queries it live when your apps need employee data. It won't have the same deep scheduled sync as Paylocity or ADP, but for building onboarding paths or cross-referencing employee records, live queries work fine.

Is Starch SOC 2 Type II certified? We have to be careful about what systems touch employee data.

Not yet. Starch is not currently SOC 2 Type II certified, which is worth knowing if your company has strict data-handling policies around employee PII. This is an honest limit. If SOC 2 is a hard requirement for your security team today, factor that in before connecting Paylocity or BambooHR data.

The Customer Support Agent app sounds like exactly what we'd want for employee self-service HR questions. Can we use it?

Customer Support Agent is currently in development — it's coming soon, and you can request beta access to get notified when it launches. Today, the Knowledge Management app plus the Email Agent covers most of the same ground for internal HR use: AI-powered search for self-service and draft replies for questions that come in over email.

How do we keep the knowledge base from becoming the same stale Notion that nobody reads?

The staleness detector automation is the answer to this. You describe the rule to Starch — any HR policy page not updated in 90 days gets flagged — and it runs on a schedule and Slacks you a list every Monday. You spend 15 minutes reviewing the list and updating what's actually changed. The system tells you what's overdue; you don't have to hold it in your head.

Can employees access the knowledge base directly, or does everything go through HR?

Employees can access the knowledge base directly — that's the point. You share the link, pin it in Slack, and employees use the AI search bar to answer their own questions. The goal is to get you out of the loop for the 80% of questions that are already documented somewhere.