How Lab Nova and Ashoka mapped a whole ecosystem without 1,000 hours of manual analysis
In the Lab Nova Longevidade 2026 Mapping, with Ashoka, WhatsApp interviews, diagnoses and recommendations became a flow: over 1,000 hours automated.
Published on By João Pedro Megid Carrilho
To map hundreds of organizations without turning each diagnosis into an hour of someone's work, move the diagnosis out of the team's hands and into a flow: WhatsApp collection, data cleaning, analysis rules and email feedback. In the Lab Nova Longevidade Mapeamento 2026, this added up to more than 1,000 automated hours.
What the Lab Nova Longevidade 2026 Mapping is
It is a mapping of the social innovation ecosystem in longevity, run by Lab Nova Longevidade with Ashoka to connect hundreds of social entrepreneurs in Brazil. According to the call page, the idea is to build a "living infrastructure" (our translation) that makes organizations, solutions and resources in the New Longevity field more visible and easier to connect. The call is open until 12/31/2026.
Those who respond get three things: an exclusive page on the mapping website, an innovation diagnosis based on Ashoka's global criteria, and recommendations for connections with other actors in the ecosystem. That promise of individual feedback is what makes the project hard to run by hand.
How it was: one diagnosis per person, about an hour each
When the feedback is individual, the cost grows in a straight line with the number of participants. Reading the response, comparing it with the criteria, writing the diagnosis and finding who that organization should talk to takes about an hour by hand. That is the number we use in the project report: more than 1,000 hours of diagnoses and recommendations that stopped being manual work.
With hundreds of actors, the choice would be between generic feedback for everyone or a team dedicated only to this. The mapping only works if each participant receives something that is useful to them.
How it is now: input, blocks and output
The project follows the same design as every CORRE operation: input, no-code blocks and output.
| Stage | What happens |
|---|---|
| Input | WhatsApp interview, with the audio transcribed |
| Blocks | Data cleaning, integration with the website, external and internal diagnoses, recommendation with more than 50 rules and AI |
| Output | Email sequence and feedback to the participant |
Source: CORRE project reports.
Each block is a piece that can be swapped without redoing the rest. If the diagnosis criteria change, we change the diagnosis block, and the collection and the email carry on as they are.
Where AI comes in and where it does not
AI is one block, not the engine. It transcribes the audio and helps draft the diagnosis text, and the team reviews what it writes. The connection recommendation starts from more than 50 rules, which people write and check. When a rule and the draft disagree, the team decides.
There is a practical reason for this split: a rule can be audited, and someone who receives a recommendation can ask why it was made.
What we learned
- Start from the output. What the participant receives defines which questions are worth asking.
- Rules where there is a criterion, AI where there is text. Mixing the two makes the result hard to explain.
- Human review at the decision points. At scale, you cannot review everything, so the design has to say where review is mandatory.
- A channel the person already uses. Answering with an audio message on WhatsApp asks less than opening a long form on a computer.
In numbers
| Indicator | Value |
|---|---|
| Hours automated in diagnoses and recommendations | more than 1,000 |
| Manual time for each diagnosis and recommendation | about 1 hour |
| Recommendation rules | more than 50 |
Source: CORRE project reports.
If you run a network or a program that needs to listen to many people and give something back to each one, see the page Networks and programs that listen and the full Lab Nova Longevidade case.
Questions on this topic
How do you run an ecosystem diagnosis with hundreds of actors?
Design what each participant gets back first, and only then the questions. Collect through a channel they already use, such as WhatsApp, clean the data in a single flow and generate the diagnosis and recommendations from rules, with team review. In the Lab Nova 2026 Mapping, this added up to more than 1,000 automated hours.
How much time did the Lab Nova Longevidade 2026 Mapping automate?
More than 1,000 hours in diagnoses and recommendations, counting about 1 hour of manual work for each one. The source is the CORRE project report.
Does AI decide the connection recommendations?
No. The recommendation combines more than 50 rules and AI, and AI is one block: it transcribes audio and drafts text. The team reviews what it writes.
Who can take part in the 2026 Mapping?
According to the call page, civil society organizations, government, business, academia, media, content producers and people who contribute to rethinking aging in the country, with registration open until 12/31/2026.