# Marcus Delgado, Director of Client Services at Hargrove & Vann CPAs — read of Account Health Predictor AI, May 25 2026

> 14 years in public accounting, currently managing client relationships across 220 active business clients at a 9-person firm in Tucson. I coach little league on Saturdays and I am always behind on email.

## How I got here

I was searching "AI tools for CPA client retention" after we lost two of our better clients this quarter with almost no warning. Found a Reddit thread that linked to something called the Wishdeal Factory. Clicked through a catalog page, saw this product, figured I'd spend five minutes.

## What I clicked first

The hero says "Account Health Predictor AI" and then immediately tells me "This product page is being finished." So the first sentence under the product name is an apology. That's an odd way to open. I did not click the 30-second explainer because the embed wasn't loading when I landed.

## Where I paused

The scoring panel stopped me cold. Not because it was impressive. Because it said "$-34,700 Year-1 take-home (Fermi)" with a negative sign in front of it and "1 in 8 meaningful-success odds" and they just... left that on the page. Voluntarily. That's either the most honest thing I've read from a software-adjacent product in years or it's some kind of reverse-psychology trust play. I genuinely couldn't tell. I sat with that for a minute.

## What I distrusted

I still have no idea what this product actually does. Not in any concrete sense. "Account health predictor" for which accounts? My firm's client accounts? Customer churn prediction? AR aging flags? It could be three completely different products depending on who's reading. The related products in the sidebar are Bookkeeper AI and Tax Planning AI, which suggests this is aimed at accounting operators, but the product itself never says that.

"credibility: 10/10" is listed as a strongest axis. Credibility of what? The page has zero customer quotes, zero screenshots, zero output examples, and the product page is literally unfinished by the builder's own admission. A 10/10 credibility score on a page that hasn't been completed yet is the kind of thing that would get flagged in our firm's vendor review process.

## What would convince me

One real walkthrough. Not a demo video with a narrator voice and placeholder data. An actual firm owner, named, on video, showing me what the dashboard looks like and saying "here is the client I almost lost, here is the flag the system surfaced three weeks before they churned." That specific. Ideally someone with a book of business comparable to ours, not a 2-person solo shop and not a 400-person national firm.

The Fermi math being negative is fine if they explain the assumptions. What's the revenue model? Are they assuming I'm charging clients for this as a feature, or is it purely internal efficiency? If it's internal, show me the time savings in hours per week per staff member. If it's a client-facing product, show me one pricing conversation that worked.

## What I'd ask in an email reply

1. When you say "account health," whose accounts are you predicting health for -- my firm's client accounts (are they going to churn me?) or my clients' customer accounts (are their customers going to churn them)?

2. What data sources does this actually connect to -- QuickBooks, Karbon, something else -- and does it work with existing data we have or does it require a behavior change to generate useful signals?

3. The year-one take-home estimate is negative. I get that you're being honest. But walk me through the assumptions. Are you assuming I'm building this as a standalone product I sell, or as an internal tool that saves labor cost?

## Verdict: on-the-fence

The radical honesty about the negative financials and the unfinished state is disarming in a good way. I've never seen a product page say "1 in 8 odds" and mean it. But I'm on the fence almost entirely because I do not know what this thing does yet. The transparency is doing a lot of lifting for a page that hasn't explained the product.

---
*Memo by skeptic persona, generated 2026-05-25. Studio breaks own self-grading loop.*