Plenty of small teams run reviews with a Google Form per reviewer, a Sheet to collect the answers and ChatGPT to turn them into a summary. It is nearly free, you control every question, and for a small team doing one cycle a year it can be enough. It breaks in three places: nothing remembers last cycle, nobody chases late reviewers, and pasting employee feedback into a general-purpose AI tool is a data-protection decision you have to make each time. This page is honest about where the DIY route wins.
| review.center | Google Forms + ChatGPT | |
|---|---|---|
| Cost | Public, per user: from €1.99/month, up to €7.49 with everyone in one cycle (by automation level; Continuity included from Enhanced; 10–20% volume breaks above 250 users) | Forms and Sheets are free with a Google account. ChatGPT Business is $20/user/month billed annually ($25 monthly) with a 2-seat minimum, and only the people who write prompts need a seat |
| Setup | We set up your company, people, managers and work profiles; billing starts when that is done | You build the forms, the sheet and the prompt, and rebuild or copy them each cycle |
| Scoring against a competency framework | Yes: a work profile of categories and items every reviewer scores | Only if you build and maintain the structure yourself |
| Chasing late feedback | Cycle status is tracked and the compiling manager can nudge outstanding feedback | Manual: you check the sheet and message people |
| Who sees what | Access follows roles inside the product, scoped to your company | Anyone with access to the response sheet sees every response; you manage sharing yourself |
| Memory across cycles | Each cycle's improvement points are kept, and next cycle's reviewers answer a focus question per open point | None. Each prompt starts from whatever you paste in |
| AI and employee data | Anthropic's Claude API under its commercial API terms (inputs and outputs not used for training), EU hosting on AWS eu-west-1, three listed sub-processors, human sign-off available; a Classic cycle runs no AI automatically | You paste feedback into ChatGPT. OpenAI says Business data is not used for training by default and offers a DPA. Its 2025 data-residency announcement lists Enterprise, Edu and the API, not Business; personal accounts have different controls, so check yours |
| Consistency between managers | The same questions, ratings and rules for everyone in a role | Depends on who writes the prompt and the form that cycle |
If you have a handful of people and one review a year, use the form and the sheet, and leave this page. The case for switching is a second cycle. Once you are comparing against last time, losing track of who has not answered, or reading the same complaints twice, you are spending manager hours on the parts a tool does for you. On AI, whichever route you take, decide what feedback text may leave your company, and check the account type and terms before pasting anything in.
It depends on the account and your agreements, and this is not legal advice. Employee feedback is personal data under GDPR, so you need a lawful basis and, for a processor, a data processing agreement. OpenAI says business-plan data is not used for training by default and offers a DPA for ChatGPT Business; personal accounts have different controls. OpenAI's November 2025 data-residency announcement lists Enterprise, Edu and the API Platform, and does not list Business. Check the current terms before you rely on it.
Usually at the second cycle, or around 25 to 30 people. Typical signals are late reviewers you chase by hand, the same improvement points reappearing, and managers spending days compiling answers. Below that size the free route is fine, and review.center is probably not worth it.
No. It calls Anthropic's Claude API, plus Voyage AI for embeddings of the improvement-point wording with names removed. Both are listed with AWS as sub-processors on the Trust page. A Classic cycle runs no AI automatically; summaries and improvement points start at the Light automation level and AI rating suggestions at Enhanced.