Amber by inFeedo
Amber by inFeedo asks employees how they feel in a chat. HR reads group results and a list of people to meet, by name. Several AI features run on OpenAI’s GPT models, and a customer asks inFeedo by email to switch one off. The contract puts duties on the customer. Its data protection addendum covers only customers based in the EU region.
Amber asks in a chat, and HR gets a list of people by name
Amber is a chat that asks employees how they feel and reads the answers with AI. HR sees group results and a list of named people to meet.
You want listening that runs as a conversation on Teams, Slack, WhatsApp or email. You want yearly and pulse results in one record. You are ready for HR to act on a named list of people.
Every conversation must be anonymous. Amber’s conversation is confidential, and HR sees named people to meet. You also need the data protection addendum to cover your country. Its data-security terms apply it to customers based in the EU region.
Pricing is user-based. Extra users are billed in blocks of 50. Changes the customer asks for cost extra. At renewal the fee rises by 10% of the last paid fee.
Amber chats with each employee, and HR reads the results by group and by person
The way Amber asks decides what HR learns and who learns it. Amber asks each employee how they feel in a chat. One inFeedo page has the chat run inside Teams, with a link to the same chat on other channels. Its integrations page has Slack collect responses within the platform. Analytics break sentiment down by team, manager and location.
HR picks the topic and the people for a pulse. A pulse runs once or on a schedule. HR can use 26 inFeedo templates, have Amber draft questions from a brief, or upload its own. Check-ins also run at set points. They cover onboarding, tenure, exit, stay and post-exit.
Amber labels the answers positive, neutral or negative. It summarises every comment into what is working and what needs attention. Admins get daily highlights that include people marked medium or high risk of leaving. They are told to read those chats. The People to Meet algorithm analyses sentiment, behaviour and silence patterns to predict the risk of someone leaving. It also uses passive signals from existing HR data. Lens answers questions about employee data in a chat.
Each chat starts with a trust statement that tells employees who can read their answers. Nearly every customer keeps the direct manager off that list. Reports hide groups too small to protect identity. The report author chooses the minimum cut-off. One help example groups by manager and keeps groups with at least 2 responses.
Amber connects to 35+ systems through 100+ integrations. They include Darwinbox, Workday and SAP SuccessFactors. The data inFeedo receives includes full name, date of joining, email address, department and location. Other details are optional.
Our guide to employee engagement software covers how the AI in this market handles named lists.
Reach, a linked yearly survey and drafted action plans
Staff answer on their own channel
Amber reaches email, Teams, Slack, Google Chat, WhatsApp and browser links. No app is needed. SMS reaches blue-collar employees. Amber reminds people who have not answered.
The yearly survey and the pulses share one record
A yearly engagement survey can be brought in from past surveys. It links to pulse data. inFeedo’s People Science consultants help design the yearly questions. Annual results can be compared with internal benchmarks.
Action plans start with a draft
Amber drafts an action plan for HR from each eNPS detractor’s history. eNPS asks whether employees would recommend the company as a place to work. A detractor is an employee who would not. HR assigns an owner and a deadline. Amber gives up to five action suggestions per piece of feedback. Admins can edit action items and acknowledgement emails.
HR can limit what each admin sees
Admins can join access groups. The groups include demographic, manager and HR. The in-app access system can limit an admin to specific demographics or features.
Security audits are on record
inFeedo holds ISO 27001, ISO 22301 and SOC 2 Type II. It meets GDPR, CCPA and LGPD. inFeedo sets a service uptime target of 99.95%. It uses commercially reasonable efforts to meet it.
inFeedo reports results for companies from 251 to 130,000 employees
These are inFeedo’s own customer results, so read them as the vendor’s account. inFeedo Tech Inc. is incorporated in Delaware and located in New York. inFeedo counts 88 enterprises with 5,000 to 130,000 employees among its users. Its customer results also include companies of 251 and 852 employees.
Aboitiz Power, 251 employees
Aboitiz Power works in oil, gas and chemicals. inFeedo reports a 77% response rate and 3X return on investment in the first year. It also reports nearly 100% retention of at-risk employees.
AXA Affin, 852 employees
inFeedo reports 18X return on investment for AXA Affin. It reports 78% of at-risk employees retained and an 87% response rate.
OYO, 9,000+ employees
inFeedo reports that OYO retained 75% of the at-risk employees it identified. It reports the response rate rose from 25% to 80%.
Genpact, 130,000 employees
inFeedo reports Amber deployed across 130,000 Genpact employees. It reports that employees who engage are twice as likely to stay.
Named lists, outside AI models and contract duties need a decision
Three things each need a decision before staff are asked to answer. The first is the named list. The second is the AI models that read the answers. The third is the duties the contract puts on the customer.
Silence can flag a person
Amber flags an employee who skips two or more check-ins as potentially disengaged. inFeedo adds that silence may be workload, burnout or disengagement. It adds that Amber does not assume intent. An anonymous pulse cannot mark a person silent, because people cannot be uniquely identified. Tell staff before they start whether skipped check-ins are tracked.
Names go to a model
AI Chat Summary sends the employee’s full name and the whole chat to a model. It also sends the company name. Chat Sentiment Analysis sends the questions and responses, plus the company name if a question contains it. Names, email IDs and phone numbers are not shared there. That feature runs on GPT 3.5. inFeedo uses OpenAI and Azure AI as sub-processors for its AI features, with 30-day retention for abuse monitoring. OpenAI does not train its models on inFeedo customer data.
Switching an AI feature off is a request
A customer asks inFeedo by email to turn off an AI feature. Customers have a right to opt out of features that may share personal information. inFeedo’s trust centre sets out the optional AI features. inFeedo calls AI output advisory and reviewed by customer administrators before action.
Sentiment and the EU ban
The EU AI Act bans inferring emotions at work, except for medical or safety reasons. Amber analyses every response for sentiment, and Chat Sentiment Analysis reads written answers. In its guidelines the Commission places emotions and sentiments inferred from non-biometric data outside the ban. The Commission calls the guidelines non-binding. inFeedo puts its AI systems in the limited-risk group under the EU AI Act.
Summaries on anonymous pulses
Its pulse page has Amber summarise every comment into what is working and what needs attention. Its help centre puts the AI Summary on tenure and non-anonymous pulses only. The AI Summary needs at least 5 answers per element. The admin must ask the customer success manager to enable it.
The customer carries staff duties
The customer warrants it has taken permission from its employees to transfer their personal data. The customer also covers inFeedo against claims by those employees about the use and processing of that data. inFeedo treats the customer employer as the data controller for employee data.
Data region and contract scope
The data protection addendum applies only to customers based in the EU region. inFeedo puts the AWS region as the customer’s choice on its platform page. On its trust centre it lets customers state a preference. inFeedo uses AWS data centres in the USA, India, Ireland and Indonesia.
inFeedo gives several language counts
inFeedo gives 30 languages for Listening on its pricing page. It gives 100+ languages elsewhere on that page. Its product pages give 34 languages for employee answers. In its plan comparison, three plans show English plus 10 popular languages.
Pricing is per user, and the terms set the dates after you sign
The price is set per user, and the costs that matter come after you sign. Pricing is user-based. Extra users are billed in blocks of 50. Changes the customer asks for cost extra. Customers can start with one of inFeedo’s three parts and add others. Consulting bundles from its People Scientists include manager effectiveness design and cultural diagnostics.
Three dates and sums cost money after signing. These product terms apply to order forms dated 31 August 2022 onwards. At renewal the fee rises by 10% of the last paid fee. The customer must give at least 15 days’ notice to renew, or access is revoked at expiry. The service is suspended immediately when an invoice is unpaid after 30 days, until payment.
Some terms are worth raising before signing. The customer must provide at least two testimonials for inFeedo to publish on social media, always including LinkedIn. A new sub-processor counts as approved when the customer has not objected in writing within 15 business days of notice. Each party’s liability is capped at the amount paid or payable for the service giving rise to it. The cap covers the twelve months before the first incident. Exceptions cover the customer’s restrictions and responsibilities, confidentiality and IP infringement.
Export your data before the term ends. Under inFeedo’s data protection addendum the customer is responsible for exporting it. When the contract ends, confidential information and data are returned to the customer, or deleted with an undertaking. inFeedo also deletes data within 30 days of an authorised request.
You could also run survey steps with agents built on your own systems. We have mapped what that looks like. The HR Agent Map
Thirteen questions to put to inFeedo
- How does an employee learn, before the first answer, that Amber can put their name on a list? Good answer: A plain message shown before the first question, in the employee’s language.
- Who can open the People to Meet list, and can we limit it? Good answer: Named roles that we set, with the direct manager off the list.
- What is the smallest group a report shows, and who sets it? Good answer: A number we set in writing, so no report goes below it.
- Which AI features are on for us, and how do we turn each one off? Good answer: A list of features, and a written way to switch each one off.
- Will your data protection addendum cover our country? Good answer: Yes, with the country named in the order form.
- Which AWS region holds our data, and is it written in the order form? Good answer: One named region, written in the order form.
- Can you put the price per user, the first term and the renewal rise in writing? Good answer: All three on the order form, before we sign.
- Can the People to Meet flags be switched off, and how? Good answer: A written yes, with the setting named and who can change it.
- Which AI summaries run on anonymous pulses? Good answer: A written list of summaries for each pulse type.
- Which languages are in our plan? Good answer: A named list of languages in the order form.
- Does the Slack conversation run inside Slack, or does it open a link? Good answer: A clear answer for each channel we use.
- Can you narrow the staff-permission promise and the claims clause? Good answer: A written change to both clauses in the order form.
- How do we export our data before the term ends, and in what format? Good answer: A file format and a date written into the contract.
Alternatives to inFeedo, including buying nothing
Windmill and Qualtrics each answer one point where Amber stops, and buying nothing is a third route.
Look at Windmill if you want AI summaries of anonymous results. The Report Studio AI Summary at inFeedo works only for tenure and non-anonymous pulses. On Windmill’s anonymous pulses the AI summary leaves out respondent names, so it covers them. Windmill leaves anonymous pulses out of its Pulse stats.
Look at Qualtrics if you want your own admin to switch generative AI on and off. At inFeedo a customer asks by email to turn off an AI feature. Qualtrics Brand Administrators can turn generative AI features on or off for the whole organisation, and new ones start off. Non-generative AI features, including Text iQ sentiment, are controlled only by user permissions.
If an HR system runs your surveys, read the guide for HR system owners.
Buying nothing means no staff-permission promise and no claims clause. Under inFeedo’s product terms the customer warrants it has taken permission from its employees. The customer also covers inFeedo against their claims. You would go without the Amber chat on Teams or WhatsApp and the daily highlights. You would also go without Anonymous Ally, inFeedo’s 24x7 channel for anonymous feedback.
https://www.infeedo.ai/listening · inFeedo Listening · read 27 Oct 2026
https://www.infeedo.ai/enps-survey · inFeedo eNPS survey · read 27 Oct 2026
https://www.infeedo.ai/pulse-surveys · inFeedo pulse surveys · read 27 Oct 2026
https://www.infeedo.ai/annual-engagement-surveys · inFeedo annual engagement surveys · read 27 Oct 2026
https://www.infeedo.ai/employee-experience-platform · inFeedo platform · read 27 Oct 2026
https://www.infeedo.ai/pricing · inFeedo pricing · read 27 Oct 2026
https://www.infeedo.ai/integrations · inFeedo integrations · read 27 Oct 2026
https://support.infeedo.com/en/articles/8069493-tenure-drivers-summary · inFeedo help article, tenure drivers and sentiment · read 27 Oct 2026
https://www.infeedo.ai/lens-ai-people-scientist · inFeedo Lens · read 27 Oct 2026
https://www.infeedo.ai/employee-engagement-surveys · inFeedo engagement surveys · read 27 Oct 2026
https://support.infeedo.com/en/articles/7883826-mood-score-across-demographics · inFeedo help article, mood score across demographics · read 27 Oct 2026
https://support.infeedo.com/en/articles/8069483-drivers-across-demographics · inFeedo help article, drivers across demographics · read 27 Oct 2026
https://www.infeedo.ai/silent-employees · inFeedo silent employees · read 27 Oct 2026
https://www.infeedo.ai/action-planning · inFeedo action planning · read 27 Oct 2026
https://www.infeedo.ai/predictive-people-analytics · inFeedo predictive people analytics · read 27 Oct 2026
https://www.infeedo.ai/employee-anonymous-feedback · inFeedo Anonymous Ally · read 27 Oct 2026
https://www.infeedo.ai/continuous-listening · inFeedo continuous listening, Genpact · read 27 Oct 2026
https://www.infeedo.ai/subprocessors · inFeedo sub-processors · read 27 Oct 2026
https://www.infeedo.ai/privacy-policy · inFeedo privacy policy, including the EU AI Act notice · read 27 Oct 2026
https://www.infeedo.ai/data-security-terms · inFeedo data-security terms · read 27 Oct 2026
https://www.infeedo.ai/terms-and-conditions/2022 · inFeedo product terms, order forms dated 31 Aug 2022 onwards · read 27 Oct 2026
https://www.infeedo.ai/hubfs/Data%20Protection%20Addendum%20(1).pdf · inFeedo data protection addendum · read 27 Oct 2026
https://trust.infeedo.ai/ · inFeedo trust centre · read 27 Oct 2026
https://support.infeedo.com/en/articles/7887670-why-am-i-receiving-daily-highlights · inFeedo help article, daily highlights · read 27 Oct 2026
https://support.infeedo.com/en/articles/8069484-tenure-ptm-summary · inFeedo help article, People to Meet · read 27 Oct 2026
https://support.infeedo.com/en/articles/8177272-what-are-the-openai-powered-features-on-the-dashboard · inFeedo help article, OpenAI-powered features · read 27 Oct 2026
https://support.infeedo.com/en/articles/8177247-how-infeedo-uses-openai-gpt-api · inFeedo help article, how it uses the OpenAI API · read 27 Oct 2026
https://support.infeedo.com/en/articles/8094484-elements-summary · inFeedo help article, AI Summary in Report Studio · read 27 Oct 2026
https://support.infeedo.com/en/articles/10057104-silent-people-across-demographics-pulse · inFeedo help article, silent people · read 27 Oct 2026
https://www.infeedo.ai/success-stories/how-aboitiz-power-achieved-3x-roi-retained-at-risk-talent-with-amber · inFeedo customer results, Aboitiz Power · read 27 Oct 2026
https://www.infeedo.ai/success-stories/how-axa-affin-achieved-18x-roi-saved-78-at-risk-employees-with-amber · inFeedo customer results, AXA Affin · read 27 Oct 2026
https://www.infeedo.ai/success-stories/how-oyo-saved-75-at-risk-employees-scaled-culture-across-7-countries-with-amber · inFeedo customer results, OYO · read 27 Oct 2026
https://efts.sec.gov/LATEST/search-index?q=%22infeedo%22 · SEC EDGAR search, Form D filed by inFeedo Tech Inc. · read 27 Oct 2026
http://publications.europa.eu/resource/cellar/dc8116a1-3fe6-11ef-865a-01aa75ed71a1.0006.03/DOC_1 · EU AI Act, Regulation (EU) 2024/1689, Article 5 · read 27 Oct 2026
https://ai-act-service-desk.ec.europa.eu/sites/default/files/2025-08/guidelines_on_prohibited_artificial_intelligence_practices_established_by_regulation_eu_20241689_ai_act_english_ied3r5nwo50xggpcfmwckm3nuc_112367-1.PDF · Commission guidelines on prohibited AI practices · read 27 Oct 2026
https://digital-strategy.ec.europa.eu/en/library/commission-publishes-guidelines-prohibited-artificial-intelligence-ai-practices-defined-ai-act · European Commission page on the guidelines · read 27 Oct 2026
https://help.gowindmill.com/features/pulse-surveys/anonymity-and-privacy · Windmill help centre, anonymous pulses · read 27 Oct 2026
https://help.gowindmill.com/features/pulse-surveys/notifications · Windmill help centre, pulse notifications · read 27 Oct 2026
https://www.qualtrics.com/support/survey-platform/sp-administration/artificial-intelligence-ai-administration/ · Qualtrics support, AI administration · read 27 Oct 2026