# Marcus Delgado, Independent Strategy Consultant — read of LexiRisk, June 17 2026

> 11 years consulting for early-stage companies, currently juggling 4 active clients and signing or reviewing roughly 2-3 contracts a month. Coach my daughter's U10 soccer team Saturdays so Tuesday mornings are my only real admin time.

## How I got here

I was in the Indie Hackers Slack and someone dropped a link with the note "anyone tried this for NDAs?" No strong endorsement, no hate, just a link. That's actually how I trust links more than LinkedIn ads, so I clicked. Landed on the homepage around 9pm after bedtime routines, which means I gave it maybe 4 minutes before I'd close the tab.

## What I clicked first

The comparison table. Specifically the "$9-$49 vs $300-$1,000+" line. That's a legitimate delta and I've definitely paid $400 to have a lawyer glance at a 6-page services agreement and tell me "looks standard." The "30 seconds vs 3-7 days" row also hit because the real cost isn't always the money, it's the waiting.

The line that actually got me was: "flags buried terms that usually hide on page 12." That's specific. That's the kind of specificity that makes me think someone actually read contracts to write this page, not just built a landing page template.

## Where I paused

The "~94% of high-risk clauses" claim in the FAQ. That number is doing a lot of work. 94% out of what? Their own test set? Self-reported? "Benchmarked against human contract reviews" is vague in a way that could mean anything from a rigorous third-party audit to two interns spot-checking 20 contracts. I sat on that for a minute. If it's their own training data they're testing against, that number is nearly meaningless. If it's blind holdout data reviewed by actual attorneys, I want to know that.

## What I distrusted

"Join 2,000+ founders and freelancers." That number isn't signed or dated. Could be 2,000 free trial signups from a Product Hunt launch 18 months ago, with 40 of them converting. It's the kind of social proof number that means nothing without a conversion qualifier. I've learned to discount these automatically.

Also: the negotiation tips feature. "For each flagged clause, we suggest common counter-terms." I'd want to know if those counter-terms are generic boilerplate or actually context-sensitive to the contract type. Because a counter-term that makes sense in a services agreement could be noise in an investor side letter. That claim could be the best feature on the page or total nonsense depending on execution, and the page doesn't show me an example.

## What would convince me

One real contract walkthrough. Not a demo video with a fake NDA someone wrote themselves. Actual output from a real vendor agreement or a SaaS service contract, with the flagged clauses visible and the plain-English explanations shown side by side. I want to see what a "High risk" score actually looks like versus a "Medium." If the AI is calling a standard limitation-of-liability clause "High risk" because all liability language looks scary to a model, that's a false positive problem that would make me stop using it fast.

A lawyer's name somewhere on this page would also help. Not a compliance boilerplate disclaimer, but an actual attorney who consulted on what gets flagged. That would do more work than the 94% number.

## What I'd ask in an email reply

1. The 94% accuracy claim -- what was the test set, who reviewed the contracts it was benchmarked against, and do you have a breakdown by contract type? I sign vendor agreements and SOW retainers mostly, not M&A docs.

2. What happens when the AI is wrong? If it misses a clause that ends up costing me, what's the liability position? I'm not asking because I expect compensation, I'm asking because how a product answers that question tells me a lot about whether the founders actually thought through the risk model.

3. Is the plain-English output generic or does it reference the specific contract terms? Like if a non-compete says "18 months in North America" does the summary say that, or does it say "this contract contains a non-compete clause which may restrict your activities"?

## Verdict: on-the-fence

The page communicates the product clearly, which puts it ahead of 80% of the AI tools I've looked at this year. But the social proof is thin and the accuracy claim needs unpacking before I'd pay $9 for a scan, let alone $29/month. One real example output would probably push me to try the free tier.

---
*Memo by skeptic persona, generated 2026-06-17. Studio breaks own self-grading loop.*