All articles
Founder Posts

"I Can Just Use Claude or ChatGPT for That" — And Why That Answer Is Costing Subcontractors More Than They Think

Charles Inokon
June 15, 2026
9 Min Read
Founder Posts
Breva Blog Post
Quick Answer A general-purpose AI subscription — whether that's Claude, ChatGPT, Gemini, Grok, or Copilot — is not the same category of tool as a purpose-built construction finance platform. The difference isn't the price. It's the data protection, the domain context, and the connected workflow.

Running a subcontracting firm means making good decisions fast with incomplete information. So when Claude, ChatGPT, and Gemini became genuinely useful a couple of years ago, a lot of sub owners and controllers did exactly what I'd expect a resourceful operator to do: they started using them. Typing in cash flow questions. Pasting in WIP schedules. Asking about lien waiver language. Getting answers in seconds instead of waiting on a call with a CPA or attorney.

I get it. I'm a CPA. I've watched smart operators make pragmatic technology choices my whole career. And I'm not here to tell you general-purpose AI is useless — it isn't. Claude is an excellent research and writing tool. ChatGPT is genuinely useful for drafting. Gemini integrates well with Google Workspace. These are real products with real utility.

But I am going to make a case that for your financial operations specifically — your pay-app workflow, your WIP, your cash position, your bonding exposure — the "I'll just use one of those" answer carries risks most subcontractors haven't fully priced in.

What Actually Happens to the Data You Type Into a Consumer AI

Let's start with the thing most people don't check: the terms.

When you open Claude Pro, ChatGPT Plus, Gemini Advanced, or Grok on a personal account and paste in your Schedule of Values, your draw request, or your aging receivables — that data doesn't disappear when you close the tab.

The pattern is consistent across every major platform: consumer tiers default to using your conversations for model training unless you manually opt out. The opt-out exists on all of them. Most people never find it.

ChatGPT
Free and Plus plans use conversations for training by default. Opt out via Settings → Data Controls. Team and Enterprise exclude training by contract — but that's $30/user/month, not the $20 plan most people use.
Claude
Consumer plans on Claude.ai use conversation data for training by default unless you opt out in account settings. The API is excluded by default. The consumer opt-out works going forward — it doesn't retroactively remove data already used.
Gemini
Consumer plans allow data to be used for model improvement unless settings are changed. Gemini Enterprise provides strong data protection, but that's enterprise-tier pricing.
Grok
Trains on conversations by default, with an opt-out in settings. Also has access to your entire X (Twitter) activity — every public post you've ever made. Under active regulatory scrutiny
Copilot
Personal Copilot may use conversation data to improve Microsoft products unless you opt out via the Microsoft Privacy Dashboard. Microsoft 365 Copilot enterprise users have strong contractual protections — again, an enterprise product.
The through-line The protection most business users assume they have is the protection that comes with a product they're not paying for.

That distinction matters when what you're pasting in is your firm's WIP schedule. Your retainage exposure. Your cash position going into a slow payment cycle. That data is a map of your business — who owes you, how much, and how far behind they are.

There's a second risk beyond training data. Nearly one in five organizations has reported a breach tied to unsanctioned AI use. The reason usually isn't a hacked platform — it's sensitive data ending up in a conversation history, a shared account, or a browser extension with overpermissioned access. For a sub running a $3M book of work, your lien rights, bonding relationships, and lender trust all depend on your financial data staying where it belongs.

The Accuracy Problem Nobody Talks About

Data privacy is the obvious risk. The accuracy risk is quieter but just as costly.

General-purpose AI tools are trained to be broadly useful across millions of use cases. Ask any of them about your cash flow and you'll get a confident, fluent, well-formatted answer. What you won't get is an answer that accounts for how construction firms actually get paid.

They don't know what a Schedule of Values is in the context of a draw cycle. They don't know the difference between a conditional and unconditional lien waiver. They don't know that your 10% retainage won't release until final completion — and that "final completion" means something specific in your contract. They don't know your state's prompt payment statute.

Hallucination is the technical term for when an AI produces a confident but incorrect answer. In financial contexts, that's material. A misquoted accounting standard. An incorrect interpretation of AIA contract terms. A cash flow projection built on assumptions that don't match how your draws actually clear. The more you're relying on the output, the more the accuracy gap costs.

What "Purpose-Built" Actually Means for Your Workflow

The category difference between a general-purpose AI and a vertical AI platform isn't about which one has a better underlying model. It's about what surrounds the model.

A general-purpose AI starts from scratch every time you open a chat window. You have to teach it every time. At Breva®, we built the platform around the full operational workflow a subcontractor runs to win jobs and get paid.

1
Takeoff V1 Live
AutoTakeoff™ is V1 — it works, and we're actively improving it. The workflow starts here, not at the invoice. No general-purpose AI connects your quantity takeoff to what happens downstream.
2
Estimates
Quantities flow from takeoff into estimates. That connection — between what you measured and what you priced — is where most general-purpose tools break down entirely.
3
Schedule of Values
Your SOV should follow from your estimate. When it does, your pay-app is defensible. When it doesn't, you're negotiating against your own numbers.
4
Pay Applications
This is where the cash cycle lives. A general-purpose AI can help you draft language. It cannot process your actual pay-app data, connect it to your WIP, or flag when your billing doesn't match your cost-to-date.
5
Quotes
Winning the next job while waiting to get paid on the current one is the structural challenge every sub lives with. Your quote process needs to be connected to your actual cost data — not a chat window where you re-enter numbers each time.
6
Prequalification
Your bonding capacity, your Breva Score, how your numbers look to a GC's risk team — all of it is a direct function of how disciplined your financial operations are upstream. Ask Bre™ coaches you on exactly this.

V1 of every stage is live. None of it is perfect — we're shipping improvements to each component. But the architecture — the connection between these stages — is what makes the platform categorically different from typing questions into a chat window.

Why the Platform Runs on SOC 2 Type II

SOC 2 Type II is an independent audit of security controls — verification that your controls actually operated effectively over a sustained period. Not a self-report. An auditor's confirmation.

Breva® is SOC 2 Type II certified, hosted on Microsoft Azure. What that means practically:

What SOC 2 Type II means for your data
Your financial data is not used to train AI models
Not accessible to other users or stored in a shared chat history
Security controls independently verified, not self-reported
Hosted on Microsoft Azure enterprise infrastructure

Claude Pro, ChatGPT Plus, Gemini Advanced, and Grok subscriptions don't come with that. They come with terms of service you probably haven't read and privacy toggles you probably haven't set.

The Real Cost Comparison

The "it's cheaper" argument makes sense if the comparison is subscription price versus subscription price. On that math, yes, a consumer AI costs less.

The comparison that actually matters: what does it cost to use the wrong tool?

  • Every hour your controller spends reformatting SOV data that should flow automatically from your estimate
  • Every pay app that goes in late or incomplete because there's no workflow connecting billing to cost tracking
  • Every time a lender or GC asks for your financials and you're pulling numbers from three different places
  • The retainage that sits uncollected because nobody tracked the release conditions
  • The job you didn't bid because your prequalification package wasn't ready

None of those costs show up on a subscription invoice. They show up in your work-to-cash cycle, your bonding capacity, and your ability to grow.

The Bottom Line

Use general-purpose AI tools for what they're good at — drafting emails, researching regulations, summarizing documents. Claude, ChatGPT, Gemini are genuinely useful for that work. We use them ourselves.

Don't run your financial operations through them. They're the wrong category of tool. They don't know construction. They don't protect your financial data the way a SOC 2 Type II platform does. They don't connect your takeoff to your estimate to your SOV to your pay app to your funding.

That's what Ask Bre™ is built for. Not as a replacement for your CPA or your attorney — Ask Bre™ does not provide legal or financial advice — but as an AI financial coach that knows construction, knows your numbers, and operates in an environment built to protect them.

See how the full workflow — from takeoff to funded — works for subcontractors.

See How Breva Works
Cadence Financial Group, Inc. DBA Breva® | This post is for informational purposes only and does not constitute financial, legal, or tax advice. Consult qualified professionals for guidance specific to your situation.
Charles Inokon
Co-Founder & EO

Get the Work to Cash newsletter.

One email a week. No fluff.

Like this? You'll love what Breva® does for contractors.