I Gave Claude AI My Business Documents — Here's What It Found
I've been using AI tools for writing and research for months. What I hadn't done was turn them on my own business.
One afternoon I gathered everything: three client contracts, my standard terms of service, two proposals I'd sent recently, and the SOPs I use internally. I uploaded them all to Claude and asked a simple question:
"Read all of these documents and tell me what you find. Be honest. What's inconsistent, what's unclear, what's missing, and what could cause me problems?"
What came back was useful, occasionally uncomfortable, and genuinely better than I expected.
What I Uploaded and Why
Three client contracts — Documents I'd signed with current clients. All adapted from a template I found years ago and modified over time.
Standard terms of service — The T&Cs on my website. Written a couple of years ago when I was setting up, not updated since.
Two recent proposals — Live proposals I'd sent to prospects in the past month.
Three internal SOPs — Process documents describing how I handle client onboarding, project delivery, and invoicing.
My expectation: Claude would find some grammatical inconsistencies and maybe flag a missing clause.
What Claude Found
In the Contracts
Inconsistent payment terms. Two of the three contracts specified payment within 30 days. The third — my largest client — said 14 days. I'd adapted the template differently and never noticed. If I ever chased the large client for late payment using the standard 30-day language, I'd have been contractually wrong.
Missing scope limitation clause. None of my three contracts had a clear clause limiting what happens when the scope changes. They described the initial scope but had no language about what constitutes scope creep or how additional work gets priced. Claude flagged this as a potential dispute risk.
Liability cap ambiguity. My liability clause was present but worded in a way Claude described as "potentially unenforceable in several UK jurisdictions due to ambiguous language." I'm not a lawyer and neither is Claude, but this was a flag worth taking to one.
Termination inconsistency. One contract required 30 days' notice to terminate. The other two said "reasonable notice." Not a legal problem, but inconsistent enough that it could cause confusion.
In the Terms of Service
No mention of AI tools. My T&Cs said nothing about my use of AI in delivering work. Given that I now use AI tools regularly in my workflow, this is a gap — both for transparency and for managing client expectations.
Data handling vague. The T&Cs had a paragraph about data confidentiality but nothing specific about how client data was stored, for how long, or what happened to it at the end of the relationship. Post-GDPR, this is worth addressing.
Price revision clause missing. No language about how and when prices could be revised. If I want to increase fees for an existing client, there's nothing in the T&Cs governing how that works.
In the Proposals
Conflicting timelines. Both proposals mentioned a timeline in the introduction and a different (slightly longer) timeline in the project breakdown. Not intentional — I'd edited one section and not updated the other.
Missing approval process. The proposals described what I'd deliver but not the approval process — how many revisions, what counts as approval, what happens if the client doesn't respond.
No reference to T&Cs. Neither proposal mentioned or linked to my terms of service. Technically this means the proposal might not incorporate the T&Cs by reference.
In the SOPs
This was the most useful section.
The onboarding SOP described a step that no longer existed. I'd changed my client questionnaire format months ago but hadn't updated the SOP. If someone else was following it (or if I referred back to it), they'd follow an outdated process.
The invoicing SOP had no escalation process. What do you do when an invoice is 15 days late? 30 days? 60 days? My SOP had no answer.
Knowledge gap: The delivery SOP assumed knowledge I have but that anyone new to my process wouldn't have. Claude flagged three steps that referenced systems without explaining how to access them.
What I Did With This
I spent about three hours working through Claude's findings.
Some I fixed immediately (the proposal timeline inconsistency, the SOP steps). Some I noted as things to review with a professional (the liability clause, the GDPR language in the T&Cs). Some I added to a future list (price revision clause, AI disclosure in T&Cs).
I didn't act on everything. But I now have a list of the gaps in my business documentation that I'd been blissfully unaware of.
The Uncomfortable Finding
The AI tools disclosure gap.
I use Claude, ChatGPT, and other AI tools regularly in my client work. My contracts and T&Cs say nothing about this. Whether that's a problem depends on the client and the nature of the work — but it's something I should address explicitly rather than leave as an undisclosed assumption.
I've since added a short section to my standard client agreements covering AI tool usage and how client data is handled in that context.
How to Do This Yourself
Step 1: Gather your key business documents — contracts, T&Cs, proposals, SOPs.
Step 2: Upload them to Claude (claude.ai — you can paste text if you don't have the PDF upload feature).
Step 3: Ask: "Read all of these documents. Find inconsistencies, ambiguities, missing clauses, potential risks, and anything that could cause problems. Be direct."
Step 4: Work through the findings. Some you'll act on immediately. Some need professional advice. Some you'll disagree with.
Important: Claude is not a lawyer. Its findings are a starting point for your review, not legal advice. Anything involving contract enforceability, liability, or GDPR compliance should be verified by a professional.
The Verdict
This experiment took one afternoon and surfaced issues I'd been carrying in my business documents for years.
The ROI is obvious. The time investment was low. The value was high.
More importantly: I would not have done this audit manually. The friction of reviewing my own documents — documents I'm too familiar with to see clearly — was too high. Giving them to an AI to read with fresh eyes removed that friction.
I now do a version of this quarterly.
Watch the Process
I walked through exactly how I uploaded and queried these documents on the Automate Lane YouTube channel: Automate Lane on YouTube
Automate Lane covers how AI tools can improve the parts of your business you've been too busy to look at. Subscribe for weekly practical experiments.