# Rachel Okonkwo, Principal Consultant at Okonkwo Advisory — read of Pencil (proposal-ai), May 7 2026

> 9 years in org design and change management, currently a two-person shop with one sub. I write maybe 6 proposals a month and I pick my son up at 3:30 every day, so mornings are the only real work time I have.

## How I got here

I Googled "speed up consulting proposal writing" on a Tuesday night after losing a project to someone who responded three days faster than me. I wasn't looking for AI tools specifically, I was just annoyed. This came up on page two, below a Medium post from 2019 and an older Proposify comparison piece. I clicked because the meta description mentioned "close rate" rather than "beautiful proposals" and that distinction felt honest enough to follow.

## What I clicked first

The headline: "Write proposals that close deals, not just describe work." That landed. I have used four proposal tools and none of them say that. They all sell polish. This one is gesturing at something closer to actual sales psychology, which is the problem I actually have. The "12 min avg. proposal time" number in the hero is either the most credible thing on this page or a complete lie. I don't know which yet, but it made me scroll.

## Where I paused

The feature called "Brand voice matching" stopped me. The description says: "Train Pencil on past proposals you have written and won. It learns your vocabulary, your framing, your positioning habits." I have very specific language I use around organizational readiness and change capacity. Clients have hired me partly because the proposal itself sounded different from the other three they got. If this actually works, it's the feature that matters. But I have no idea how it works, whether it's a RAG embedding over uploaded PDFs or something more shallow, and there's nothing on the page that explains the mechanism. The claim is big. The explanation is zero.

## What I distrusted

Three things.

"3.1x higher close rate." Higher than what? My current rate? The industry average? Users who signed up last month vs. last year? There is no footnote, no asterisk, no methodology. That number is doing a lot of work on this page and it has no scaffolding.

The testimonials. Marcus Delgado went from 28% to 61% close rate in three months. Dana Wu says 70% of her clients now choose the middle pricing tier. Sarah Kowalski closes at "40% higher rates." Every single person has a clean, directional, impressive number. Nobody says "it's roughly better" or "I think I close faster." The precision feels calibrated, not reported. Real people say things like "I used to lose like half my pitches, now it feels more like a third." Not "+33pt close rate improvement."

Also: "a senior copywriter who understands your industry, your clients, and what buyers need to see before they say yes." That sentence is doing the job of a claim while being completely empty. Every AI tool says some version of this. I want to know how it understands my industry. Did someone actually train on strategy consulting proposals or is "understands your industry" just shorthand for "it's a language model"?

## What would convince me

One real proposal, anonymized, that came out of this tool for someone in my vertical (org consulting or management consulting adjacent). Not a screenshot. The full document. I want to see if the scope section actually reads like a thoughtful human parsed a discovery call, or if it reads like a polished template with my client's name swapped in. That's the difference between this being useful and being a fancier version of what I already do with Notion AI.

For the close rate claim: a single sentence explaining the methodology. Even "based on users who self-reported before/after on signup and at 90 days" would be enough. I'd accept imperfect data over no sourcing.

## What I'd ask in an email reply

1. The brand voice training: does it actually learn the specific framing I use, or does it just adjust for a general tone parameter like "formal vs. casual"? Can I see what it outputs before and after training on the same brief?

2. For the proposal analytics: does the client have to open a link through your portal for tracking to work, or does it work with PDF sends? Because a lot of my clients want a PDF in their inbox, not a click-to-view portal link, and if tracking requires them to use your interface that's a hard stop for me.

3. Who built the proposal training data? Is it licensed, sourced from users, synthetic? This matters to me because I don't want my uploaded proposals being used to train outputs for a competitor's proposal a year from now.

## Verdict: on-the-fence

The page is solving a real problem and it's not leading with "beautiful AI proposals" which already puts it ahead of half the category. But I can't evaluate the core claim, which is that it actually sounds like me. That's the whole bet. I'd reply to a founder email if it was short and offered me a trial that included voice training.

---
*Memo by skeptic persona, generated 2026-05-07. Studio breaks own self-grading loop.*