Internal Knowledge Search

How a 150-person accounting firm gave every employee a research assistant that never phones home.

The firm

Solstice Financial is an accounting and advisory firm with 150 staff across two offices. They advise 400+ clients on tax, audit, corporate finance, and advisory work.

Their internal knowledge is vast — 15 years of engagement files, tax research notes, client histories, precedent documents, internal policies, and regulatory briefings. It's all stored on a shared drive and in their document management system.

Nobody can find anything.

The problem

When a junior accountant needs to know:

• "How did we handle a similar R&D tax credit claim last year?"
• "What's our internal guidance on capital allowances for renewable energy installations?"
• "Has anyone in the firm dealt with a client in the dental sector acquisition scenario before?"

...they have three options:

1. Search the shared drive. Type keywords into Windows search. Get 400 results, most of which are irrelevant. Open five documents. Find what they need in the third one, 25 minutes later. Or don't find it at all.
2. Ask a senior colleague. The senior colleague spends 10 minutes explaining something they've explained six times before. The junior accountant takes notes. The notes get lost.
3. Google it. Get generic guidance that doesn't reflect the firm's specific experience or client base. Risk acting on outdated information.

The real problem isn't search — it's that the firm's collective knowledge is locked in documents that humans can't efficiently navigate.

And there's a new complication: the firm's partners have said no to using ChatGPT or similar tools for internal research. Not because they're anti-AI — because client financial data, tax strategies, and advisory work product can't go through a third-party API. The ICO would have questions. The clients would have concerns. The professional indemnity insurers would have opinions.

The cost:

• Junior staff spend 30-60 minutes per day searching for internal information
• Senior staff spend 5-10 hours per week answering questions they've answered before
• The firm re-researches things they already know because nobody can find the prior work
• Onboarding new staff takes 4-6 months before they can work independently
• Risk of inconsistent advice — different juniors find different precedents and give different answers to similar questions

What Foundry does

Foundry runs on a Mac Studio in the firm's server cupboard (it doesn't need a rack — it sits on a shelf). It's plugged into the firm's network and has read access to the document management system.

It reads everything, understands it, and answers questions — locally.

Here's what that means in practice:

1. Foundry indexes the firm's document collection — every engagement file, research note, policy document, and precedent. Not by keywords. By meaning. It understands that a document about "capital allowances for solar panels" is relevant when someone asks about "renewable energy tax relief."
2. When someone asks a question, Foundry searches the indexed content, finds the most relevant passages, and writes a clear answer in plain English — with citations to the specific documents and page numbers it drew from.
3. Nothing leaves the building. The question is processed on the Mac Studio. The answer is generated on the Mac Studio. The documents stay on the firm's network. No API calls, no cloud processing, no third-party seeing client data.
4. It learns the firm's vocabulary. When someone asks about "the Henderson matter," Foundry knows that refers to the 2024 acquisition advisory engagement for Henderson Dental Group. It picks up the firm's internal language over time.

What it looks like day to day

Tuesday, 10:15 AM

A junior accountant, Priya, is working on an R&D tax credit claim for a software company. She's never done one before.

She types into the Foundry search bar: "R&D tax credits for software development — have we done this before?"

Foundry responds in 3 seconds:

Priya reads the BrightStream summary, finds the exact approach the firm used, and adapts it for her client. She books a 15-minute call with Michael Chen to confirm her approach rather than a 2-hour "start from scratch" meeting.

Wednesday, 2:30 PM

A manager is reviewing a file and needs to check the firm's policy on independence for audit clients who also request advisory services.

He searches: "Independence rules for audit clients requesting advisory work"

Foundry responds:

He reads the policy, checks the precedent, and writes his response to the client. Ten minutes instead of an hour of searching and calling.

The numbers

Annual impact: 8,000+ hours of professional time recovered across the firm. At an average charge-out rate of £150/hour, even recovering 10% of that for billable work is £120,000/year.

Foundry cost: £999 setup + £99/month = £2,187 first year. Plus the Mac Studio if not already owned (~£5,500-6,500 for an M3 Ultra with adequate RAM).

What stayed cloud

• Email, accounting software, CRM — all still cloud-based and untouched
• Web research when staff need to check external sources (HMRC website, legislation)
• Cloud backup of the document management system (Foundry doesn't replace backups)
• Video conferencing and collaboration tools

What moved local: the AI that reads firm documents and answers questions. That's the part that was either impossible (can't use cloud AI on client data) or painfully slow (keyword search through 15 years of files).

What it doesn't do

• Does not give tax advice. It finds and summarises what the firm has done before. The accountant decides if it applies to the current situation.
• Does not access client accounting systems. It reads the firm's internal documents, not the client's live financial data.
• Does not send anything externally. All questions and answers stay on the local machine.
• Does not replace training and supervision. Junior staff still need senior review. Foundry helps them find information faster, not make decisions without oversight.
• Does not hallucinate citations. Every answer includes direct references to specific documents and page numbers. If Foundry can't find a source, it says so rather than making one up.

What the team says

"I used to spend half my morning looking for a precedent file. Now I ask Foundry, get the answer, and I'm working on the actual problem in under a minute." Priya, junior accountant (6 months qualified)

"The real surprise was onboarding. New joiners are useful in week two instead of month three because they can actually find how we do things here." Senior manager

"We banned ChatGPT for client-related research in 2025. We lost nothing — Foundry does it better because it knows our firm's experience, not just the internet's." Partner, risk and compliance

Is this right for your firm?

This setup works for:

• Professional services firms with 50+ staff and 5+ years of accumulated documents
• Accounting, legal, consulting, and advisory firms with data confidentiality obligations
• Firms where knowledge is locked in documents that keyword search can't navigate well
• Teams that need consistent advice grounded in their own precedents, not generic AI

Not a fit if you:

• Have very little accumulated documentation (a startup with no historical files)
• Are comfortable using cloud AI tools with client data (Foundry's local-first approach is unnecessary)
• Need real-time integration with live client systems (Foundry reads documents, not databases)
• Have fewer than 20 staff (the setup cost outweighs the benefit at small scale)

← Back to all case studies | Next: Code Review →