# Rachel Mendez, Creative Director and Co-Founder at Threadline Studio — read of Pencil, May 5, 2026

> Nine years running a 5-person UX and brand studio in Portland. I write or review every proposal we send. My kid is four and goes down at 8pm, which is when proposal season really starts.

## How I got here

Someone dropped this in the Bureau of Digital Slack with the message "anyone tried this thing?" I clicked at 12:40 on a Tuesday while eating leftover rice. I had just spent three hours the night before on a proposal that the client hasn't opened. I was primed.

## What I clicked first

The hero subheadline: "Write proposals that close deals, not just describe work." That's a clean line. It's the actual problem. Most tools lead with "save time" and stop there, but "describe work" vs "close deals" is a real distinction and they said it without four paragraphs of warmup.

The "12 min avg. proposal time" stat also stopped me. I didn't believe it, but I wanted to keep reading to find out if they'd explain it.

## Where I paused

The brand voice matching section. "Train Pencil on past proposals you have written and won. It learns your vocabulary, your framing, your positioning habits." I actually read that twice. That's the only feature on this page I haven't seen before, framed this way at least. Every AI writing tool says it "sounds like you" but they mean "just adjust the tone slider." The specificity here, vocabulary, framing, positioning habits, is doing real work. I don't know if it actually delivers that, but they at least articulated something real about what "voice" means in a proposal context.

## What I distrusted

"3.1x higher close rate." No source, no methodology, no sample size. Is that across all their users? Their best users? That one guy Marcus Delgado who they quote right below? The Marcus quote is also too clean. "The difference is that the proposal actually sounds like I understand the client's problem, because it does." That's a great line. It's also suspiciously great. It reads like someone workshopped it, not like someone typed it into a review form.

The "4,200+ proposals sent" counter is also small. If this product has been live more than a few months and has consultants writing 5 to 25 proposals a month, that number should be much higher. It makes me think they're newer than the page implies, or the adoption is thinner than the testimonials suggest.

I also noticed the FAQ answer about ChatGPT was the most honest writing on the page. "ChatGPT starts from zero every time and requires careful prompting." True. The fact that the FAQ sounds more credible than the hero copy is a little backwards.

## What would convince me

One real, traceable case study. Not a quote. A mini teardown. Something like: "Here's the brief James Torres fed in. Here's what Pencil drafted. Here's the proposal he sent. Here's what the client said." Even a before/after on scope language, what the AI generated vs what a generic template would have produced, would do more than three star testimonials with headshots that all have the same background tone.

On the voice training feature specifically, I want to see failure cases. What happens if I upload three proposals that don't sound like each other? What if my "voice" is actually inconsistent? If they've thought about that edge case, I'd trust the feature more.

## What I'd ask in an email reply

1. The "3.1x higher close rate" -- where does that number come from? What's the baseline they're comparing against and what's the sample?
2. On brand voice training: when the draft comes out and it doesn't sound like me, how do I tell the tool what's wrong? Is there a feedback loop or do I just keep re-generating?
3. We sometimes write proposals collaboratively, where one person writes scope and another handles pricing. Does the Agency plan actually support that workflow or is it just multiple seats logging in to the same queue?

## Verdict: on-the-fence

The problem framing is genuinely good and a few features, tiered pricing especially, are specific enough to be interesting. But I'd need one real example and a straight answer on that close rate stat before I handed over a credit card.

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