How Fern Group Made Claude the Way the Whole Firm Works

Most small firms are experimenting with AI. Very few have turned it into institutional capability — tools every person uses the same way, with guardrails built in. That's what Fern Group set out to do, and it took eight weeks.

Where Fern started

Fern Group, a boutique Australian advisory firm, was already using AI — but adoption lived with individuals. Each person had their own prompts and habits, results were inconsistent, and the firm's capability walked out the door whenever attention shifted. More critically, there were no guardrails: no data classification policy, ad-hoc decisions about what client information went into AI tools, and SharePoint permissions that couldn't stop sensitive data flowing into AI via the Microsoft 365 Connector.

A sprint, then a roadmap — governance first

We began with a five-day AI Growth Sprint: an AI maturity assessment, target operating model, technology recommendation and phased roadmap. The assessment found Claude was the team’s strongest tool — outperforming Copilot across every use case Fern tested — and set three imperatives: make Claude the firm’s primary intelligence layer, fix the governance foundations before building further, and sequence builds by time-recovery value.

We then executed the roadmap in three phases:

1. Data Guardrails. A three-tier classification framework (AI-Unrestricted / AI-Restricted / AI-Prohibited) with every client classified, firm-wide PII rules — passports, police checks and bank details never enter any AI tool — SharePoint permission hardening with Fern’s IT provider, and a Claude Enterprise evaluation.

2. The Fern Plugin. One installable Claude Plugin bundling Fern’s tone and style guide, a shared prompt library, branded presentation and Word document creators, and a proposal-generation workflow — with the data guardrails built in. Install it, and every team member has identical, production-grade tools from day one.

3. AI Philosophy. A leadership workshop defining what Fern will use AI for, and the capabilities to invest in.

The results

Client proposals that used to absorb around 10 hours of drafting now take about 20 minutes — on-brand, in Fern’s tone, built by the plugin’s proposal workflow. Across the team, that and the document tools free up around five hours per person per week.

Beyond the numbers: 100% of Fern’s clients are now covered by the classification framework, the full team works from one production-grade plugin, and AI capability now belongs to the firm rather than to individuals. The strongest signal came unprompted — on the strength of the work, Fern referred us to another firm, which became a new engagement.

Propel didn’t just show us what AI could do — they made it how our firm works. Every one of us now produces client-ready documents with the same tools and the same guardrails, and we trust what goes into AI because the governance was built first.

— Daniel Khong, Co-founder, Fern Group

The pattern that works

Fern’s journey — sprint, guardrails, then institutional tooling — is repeatable for any professional-services firm that wants AI to be how the firm works, not just what a few keen people do. If that’s the journey you’re on, talk to us at propelventures.ai/contact.

Propel Ventures is a partner in Anthropic’s Claude Partner Network.