“Road work should happen on the weekend. I mean, start Thursday evening and take Friday and Saturday…”
Massive AI voice interviews on WhatsAppAn MIT spinoff
AI-powered collective intelligence, at scale.Uncover the what, why, and what to do.
Crowd Voice interviews massive populations by WhatsApp voice, in their own language, with follow-ups that get to the why and what to do. AI-analytics turns the massive qual + quant data into dashboards, reports and AI tools your team & AI agents can act on.
“I’d ask them, please, to come and take a look at the school, so they can see how dirty everything is…”
“Look, I don’t have a magic wand. But I think the consequences of not vaccinating are the best, in quotes, advertising for vaccination…”
“Welcome talent and brilliant minds with open arms, and don’t make experience a basic requirement…”
Heard for
To date
- 100k+people interviewed
- 8+countries
- 10+languages
The what, the why, and what to do
Riyadh residents told us, by Arabic voice note, how the city’s infrastructure works affect them and what would help.
51%
say the works affect their daily life. Yet 88% are satisfied with safety on site, and 88% with the information they get.
55%
mention peak-hour congestion. Then closures and detours, and night-time noise that costs them sleep.
44%
of their ideas: work off-peak, and finish faster. Many also ask to keep part of the road open, with clear detours and early notice.
Tap any highlighted person to hear them.
«مثلاً الحفر أخاف الأولاد الصغار يقعوا فيها…»
The holes, for example: I’m afraid little children will fall into them…
Anonymized voice note
«بعض الأماكن اللي بيتم فيها حفر أو بنية تحتية ما بيكونش عليها علامات إرشادية أو ضوئية بالليل، فبيتم حوادث كتيرة في الأماكن دي…»
Some dig sites have no guide signs or lights at night, so there are a lot of accidents there…
Anonymized voice note
«الضوضاء والزحام. الضوضاء وخصوصا في الصباح الباكر، ولدي أطفال، وزحام الطرق، وتأخير عن عملي…»
The noise and the traffic. Noise especially early in the morning, and I have children. Jams, and I’m late for work…
Anonymized voice note
«تأثير إزعاج بحيث لما نجي ننام مثلا يكون صوت العمال مزعج…»
It’s disturbing: when we try to sleep, the workers’ noise keeps us up…
Anonymized voice note
«يكون العمل بالمساء أو أوقات غير الدوام…»
Do the work in the evening, or outside office hours…
Anonymized voice note
«يتم العمل في الأوقات غير الرسمية لخروج الموظفين وخروج الطلاب من المدارس وأوقات الذروة من الازدحام المروري في أوقات في الليل إلى الصباح مثلا أو أوقات العطلة…»
Work outside the hours when employees and students head out, away from rush hour: overnight, or on holidays…
Anonymized voice note
«زيادة المسارات في الطرق، عدم العمل في وقت الدوامات…»
Add more lanes, and don’t work during commuting hours…
Anonymized voice note
In Voices for Urban Infrastructure (Riyadh, 3,434 residents), 51% say the works affect their daily life. The main impacts are peak-hour congestion (55%), road closures and detours (24%), noise and sleep (22%) and dust (7%). Residents’ own fixes: work off-peak and finish faster (44% of ideas), keep roads partly open with clear detours (24%), and quieter, daytime-only noisy works (13%).
Why Crowd Voice
The depth of a conversation,at the scale of a survey.
Surveys give you the what. Focus groups give you the why, but only for a few people. Crowd Voice interviews thousands by voice note and follows up on every answer, so you get the what, the why and what to do, from everyone.
-
4×
lower cost than field surveys plus focus groups
-
4×
faster to full results: about 3 weeks per cycle, not 3 months
-
50×
more qualitative data on the why and what to do
The intelligence engine
AI-analytic products to exploitmassive qual + quant data.
Beyond static reports: powerful, flexible AI products to exploit every study, in your team’s hands and inside your own AI tools.
Every engagement closes with a knowledge-transfer workshop, so your team can take it from there.
- Peak-hour congestion 55%
- Road closures and detours 24%
- Noise and sleep 22%
- Dust 7%
- Risks to vulnerable residents 5%
- Vehicle damage 4%
95% confidence interval
Latest voice notes
0:27 “Add more lanes, and don’t work during commuting hours.”
0:29 “The noise and the traffic. Noise especially early in the morning, and I have children. Jams, and I’m late for work.”
- Claude
- ChatGPT
- Gemini
- Copilot
Your AI assistant · connected to Crowd Voice
We’re considering closing a main road in full for six weeks. How would residents react? Make me a 5-slide deck for Thursday’s steering committee.
-
crowdvoice · research_analyze_topicsClosures and detours: 24% of residents -
crowdvoice · research_get_testimonials12 quotes on closures -
crowdvoice · research_answer_questionCited to 4 voice notes
Expect resistance to a full closure. Closures and detours are already the second-biggest impact, named by 24% of residents. Their own fixes favour keeping part of the road open with clear detours, and working off-peak. Here’s the deck:
Road-closure-what-residents-say.pptx5 slides
Executive report · AI draft
Voices for Urban Infrastructure
Summary
Half of residents (51%) say the works affect their daily life, above all through peak-hour congestion. Satisfaction with safety and information is high, at 88%. Residents’ own top fix: work off-peak and finish faster.
Key findings
- 55%mention peak-hour congestion
- 24%mention road closures and detours
- 22%mention noise that affects sleep
Recommendations
- Schedule works outside commute peaks
- Keep part of each road open, with clear detours
- Announce works early, including on map apps
51%
- Peak-hour congestion55%
- Road closures and detours24%
- Noise and sleep22%
- Dust7%
- Work off-peak, finish faster44%
- Keep roads partly open, clear detours24%
- Quieter, daytime-only noisy works13%
Example studies
Voices for…
-
Voices for Urban Infrastructure
-
Voices for Diriyah’s Retail
-
The Voice of Students
-
Voices for Heritage
-
Voices for Urban Childcare
-
The Voice of Fintech Users
-
The Voice of Iraqi Youth
-
Voices for Child Vaccination
How it works
3 technologies in one platform.
One platform runs the whole study. It interviews, it brings the right people in, and it turns every answer into evidence your team can use.
- Conversational interviews on WhatsApp, by voice note or text
- Smart follow-up questions in real time
- Active listening: every voice note understood before moving on
- Moderation for relevance, depth, respect and authenticity
- Answers by voice, text, choice, NPS, CSAT, photo or video
- Any language, with AI translation to and from any other
- Describe your goals and the AI drafts the interview; test it on WhatsApp first
- External or internal target groups, people or businesses, at large scale
- Sociodemographic and geographic targeting
- Invitation links and QR codes per channel: SMS, email, social media, websites, physical locations
- Reminders and AI-checked incentives
- Live monitoring of completion, drop-off and sources
- Statistical representativity and weights
- Every voice note transcribed and translated, and kept with its audio
- Every answer coded into topics and subtopics, with AI summaries
- Qualitative and quantitative together: KPIs × segments × time, with weights and confidence intervals
- Personalized classifiers, such as sentiment, and AI segmentation
- AI Q&A cited to quotes, and an MCP server for your AI agents
- No personal identifiers, deep anonymization, hosted in-country (PDPL and GDPR)
- On WhatsApp: nothing to install
- Active listening: voice note understood
- Smart follow-up, written in real time
- Moderation: relevant and respectful
- Arabic ⇄ English, or any language
Messages illustrative. Voice notes are real, from Voices for Urban Infrastructure.
Fieldwork · Voices for Urban InfrastructureLive
3,434 interviews · Riyadh residents
- WhatsApp link
- QR codes
- SMS
- Social media
Age
77% Drive a car
28% Elderly or disabled at home
RemindersAI-checked incentivesWeights applied
One analytics engine, four AI products
- AI-analytics dashboardKPIs, topics and segments, with Q&A cited to the voice notes
- MCP for AI agentsYour study inside Claude, ChatGPT, Gemini or your own agents
- AI reportsFindings and recommendations, ready to share as web or PDF
- AI-processed dataTranscribed, translated and coded. Yours to keep, no lock-in
Engagement models
Two ways to work with us.
Expert-supported study
3–6weeks,end to end
Our research team designs, runs and analyses the study with you, in 3 to 6 weeks depending on scope. You get the evidence, and we do the heavy lifting.
- Design1–2 weeks
- Fieldwork1–2 weeks
- Analysis1–2 weeks
- AI-analytics dashboard with cited Q&A
- Executive report and processed data
- Knowledge-transfer workshop for your team
Enterprise Subscription
10+studiesa year
Your team runs 10+ massive studies a year on its own, with our AI assistance at every stage and our people one message away.
Example plan of 20 studies
8 done · 1 in field12 months
of platform access
- AI-assisted design, promotion and collection
- All AI products, including MCP for your agents
- Support channel and knowledge-transfer workshop
Science and trust
Born at MIT.Built for institutions.
Crowd Voice is an MIT spinoff. Based on research by Prof. Abdullah Almaatouq (MIT Sloan) and a team of international researchers.
- Data residency
- Hosted inside the country, with an in-Kingdom option.
- Compliance
- PDPL and GDPR compliant.
- Consent
- Opt-in participation. No personal identifiers in your data.
- Ownership
- Your data is yours. No lock-in.
Questions
What your procurement teamwill ask.
Security reviews, methodology, data residency: we answer them in writing, and in the demo.
How do I know the AI isn’t making things up?
Every answer is cited to the voice notes behind it. Open a citation and you get the person who said it, in their own language, with the transcript and translation beside it. Our research team reviews each study’s design and outputs.
Is the sample representative of our population?
We design recruitment for your target population across SMS, QR codes, social media and direct messages, then weight results so they reflect the population by segment. Every dashboard reports sample sizes and confidence intervals.
Where is the data stored, and who owns it?
Data is hosted in-country, including in Saudi Arabia in line with PDPL. You own your data, with no lock-in.
Why real people, and not synthetic personas?
Because institutions need to know what their people actually think, feel and propose. Every finding comes from real, consented participants, and you can trace it back to their words.
What if people don’t use WhatsApp?
WhatsApp reaches most adults in the regions we work in. For groups it misses, the same interview runs in our web and mobile app.
Which languages do you support?
10+ languages to date, including Arabic, Spanish, Portuguese and English. The AI translates responses to and from any language.
Hear them all.
We’ll design your first study with you.