# Rachel Okonkwo, VP of People at Fieldvine (220 people, Series B SaaS) — read of Team Sentiment Analyzer AI, May 25 2026

> 9 years in People Ops, currently running a stack of Culture Amp, 15Five, and too many Notion templates. My six-year-old takes swim lessons Tuesday evenings which means I read a lot of vendor pages on my phone at the pool.

## How I got here

Googled "1-on-1 sentiment analysis tools" because our eng managers are drowning in transcripts from Otter.ai and nobody's reading them. I'd seen Lattice add something adjacent, hated it, and wanted to know if anyone else was trying this. This came up on page two. Not a sponsored result, just organic.

## What I clicked first

The line "Turn recurring 1-on-1s and pulse surveys into actionable team health insights" is actually a clear value prop. That's the job to be done. I clicked because I recognized my exact situation in that sentence, which doesn't happen often.

Then I looked for a demo or a screenshot and got "Try it Live result Before With Team Sentiment Analyzer" which appears to be a broken render of some before/after UI component. Not a great first impression for a product that's supposedly analyzing language.

## Where I paused

"Predictive Retention Scoring — Match sentiment patterns against historical attrition data."

I stopped here because this is actually the thing I want most, and they blew past it in two sentences. What historical attrition data? Mine, after I've been using the tool for 18 months? Theirs, from other customers? A model trained on public research? This is the core promise and there's no explanation of how the prediction works at all. I re-read it twice. Still nothing.

## What I distrusted

"Spot declining engagement before it turns into quiet quitting."

Quiet quitting is a 2022 LinkedIn buzzword that serious People practitioners have mostly retired. Using it in 2026 reads as copy that was written fast and not revisited. Small tell but I notice it.

The bigger trust problem: the footer says "Built by Wishdeal Studio" and below the fold there are four other unrelated AI products listed with estimated year-one revenue numbers. This is a studio portfolio page, not a product company. That doesn't mean the product is bad, but it does mean I have no idea if there's a team that actually knows anything about HR data, EEOC compliance, or the privacy considerations of running sentiment analysis on employee conversations. Those aren't small concerns. I have an employment lawyer on speed dial specifically because of stuff like this.

Also: zero pricing, zero case studies, no customer logos, no names, nothing. Not even a "used by teams at..." line.

## What would convince me

A real explanation of the data model. Specifically: show me what input you need from me (transcript format? survey platform integrations? do I have to migrate data?), show me one real output from the dashboard with fake names, and tell me who trained the sentiment model and on what corpus. HR sentiment is not the same as Twitter sentiment. The NLP model matters a lot here.

One reference from a People leader at a company my size would do more than anything else. Not a quote on a slide. A name, a company, a LinkedIn link I can click. That's it.

## What I'd ask in an email reply

1. The anonymization claim says "share aggregate findings without exposing individual names" -- but my managers are the ones doing 1-on-1s. If the system says "Eng Team 3 is at high flight risk," my manager already knows exactly which three people that means. How does this handle small-team de-anonymization? Is there a minimum group size?

2. What does "Flag individuals and teams drifting into risk zones" actually look like? Is there an alert? Who gets it? Does my manager see individual scores or only I do as the People lead?

3. Are you integrated with anything I already use, or do I paste transcripts into a UI?

## Verdict: on-the-fence

The problem they're solving is real and the core sentence landed. But this reads like a studio proof-of-concept, not a product I can trust with employee conversation data. If someone replies with actual answers to my three questions and a reference I can call, I'd take a 20-minute demo.

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