Tell it in a conversation or fill it in

You write the text of your review in one of two ways: in a conversation with the Feedbax assistant or yourself in the form. Both ways lead to the same sections that make a Feedbax review meaningful, and you can switch at any time. This article shows how both work and which way suits when.

At a glance

• Where: you choose the way on the first page of the review. The writing happens in step 4 of 7, see Reviewing a provider: the 7 steps.

• "Tell it in a conversation", marked "Recommended": the assistant asks, you answer. Then the AI arranges your sentences.

• "Fill it in yourself", marked "Classic": nine fields with hints. Your texts stay as you write them.

• Required in both ways: starting point, approach and outcome. Everything else is optional.

• Switching: at any time at the top of step 4 under "How you write", without losing any text.

• Stars: in both ways only you give them. The AI never suggests a rating.

How it works

Both ways write into the same nine sections. They also appear in this order in the published review: starting point, goals, choosing the provider, approach, team, outcome, collaboration, impression and "What could have gone better".

In the conversation, the sections are grouped into six topics:

Only the text that you accept in the draft and later approve is published. The chat history itself does not appear anywhere. Where the finished review appears is shown in Where will my review appear?

Tell it in a conversation

At the top of the chat there is a note: you are talking to the Feedbax assistant, an AI. It asks questions and takes notes. Nothing is published without your approval.

How to have the conversation

1. Read the assistant's question. It always asks exactly one question, in at most two short sentences.

2. Write your answer in the "Answer here" field. The Enter key sends, Shift and Enter start a new line.

3. Click "Send". While the assistant phrases the next question, it says "The assistant is thinking".

4. If you would like to talk about a particular topic, tap it in the list "Topics of the conversation". The assistant then asks about it next.

5. As soon as you have said something about all three required topics, "Create draft" becomes active. Before that, the note next to it reads "First say something about the required topics."

The list "Topics of the conversation" appears above the chat on a smartphone, and on the right below the preview on larger screens. Required topics carry the label "Required" until they are covered. If the assistant thinks everything important has been said, you see "Everything important has been said. You can now create your draft." If a draft already exists, the button is called "Continue to the draft".

How the assistant asks

• It asks about one topic after the other, first about the topic you tapped, otherwise about the first open required topic.

• If an answer is very general, it asks once for a concrete example and then moves on.

• If you rated a point low, it asks factually what was behind it. It does not judge, does not play things down and does not name a number of stars.

• It never asks whether you would recommend the provider. Your stars say that.

• It does not ask leading questions and does not put words in your mouth.

• "What could have gone better" is optional. It asks about it on its own at most once and does not follow up.

• It does not repeat names of individual people and does not ask for them. It refers to the role, for example your contact person at the provider. Company names are allowed.

Privacy in the conversation

Below the input field it says: "The AI assistant processes your answers. Learn more in our privacy policy." The link opens the privacy policy in a new tab. How Feedbax handles your details is also described in Confidentiality: who sees your data

Fill it in yourself

Under "Tell us about your project" there is the hint "Write in your own words. Short sentences are enough." The three required fields are at the top, the optional ones below. Each field has a question and a grey example text that shows what kind of answer fits.

1. Fill in the fields "Starting point", "Approach" and "Outcome".

2. Add the optional fields you would like to say something about.

3. Click "Continue to the draft". The button becomes active as soon as the three required fields are filled in.

Each field holds up to 4,000 characters. Saving happens after every short pause in typing.

What the AI does in the draft

In the next step, "Your draft", the two ways differ. The draft itself is described in Reviewing a provider: the 7 steps.

After the conversation, the content check also checks whether every sentence is backed by your answers. If it cannot find a sentence in your answers, a note next to it says: "We cannot find this sentence in your answers. Check that it is right, or delete it."

Switching the way

1. In step 4, click "Conversation" or "Form" at the top under "How you write".

2. The other way opens with the same texts.

What you said in the conversation only becomes text with the draft. If you switch to the form before the first draft, its fields are therefore still empty. Your answers remain in the conversation and are not lost.

If the assistant does not answer, it says below the chat: "The assistant is not answering right now. Try again or switch to the form." Your answer stays in the input field.

Which way suits you

• Conversation: you would like to finish quickly, do not like phrasing things or are not sure what readers are interested in. The assistant asks about what counts.

• Form: you have already sorted your thoughts, would like to decide every sentence yourself or have a text, for example from a project report, that you would like to use.

Limits

• In the conversation you give at most 40 answers per review. After that it says "The conversation is long enough. Create your draft now."

• A single answer in the chat holds up to 4,000 characters, and so does a field in the form.

• The AI writes the draft at most five times per review, including the first one.

Frequently asked questions

Does the AI make anything up?

It is not supposed to, and the content check marks every sentence it cannot find in your answers. Still, read the draft carefully. You can change or delete every section before you accept it. At the end you approve the text yourself, see Confirm and approve your review, and the Feedbax team checks it before publication, see After approval: the Feedbax check.

Does the AI suggest stars?

No. You give the stars yourself in step 3. The assistant does not name a number of stars or a score in the conversation either.

May I also express criticism?

Yes. "What could have gone better" is a section of its own, and readers value it. It is optional. Factually phrased criticism helps other companies more than pure praise. The content check marks words that could come across as hurtful. It must be honest: with the approval you confirm that you really worked with the provider, see Who may review a provider?

Can I switch the way afterwards?

Yes, at any time in step 4 via "How you write". Both ways work with the same texts.

What happens to my answers in the chat?

They are saved with your draft and processed by the AI assistant. Only the text you approve is published. How to continue writing later is described in Continue a review later.

Why should I not mention names?

Feedbax does not show names of individual people publicly. If you mention a name, the draft suggests the person's role instead, for example the project lead. The "Replace" button applies the suggestion. Company names may remain.

Related articles

• Confidentiality: who sees your data

• Client dashboard: the overview

• What is the Feedbax Score?