AI for NGOs: Automating Without Dehumanizing Your Service
How to use AI in an NGO's service without losing warmth: four criteria, the role of the team and what the flow should route to a person.
Published on By João Pedro Megid Carrilho
Automating an NGO's service without dehumanizing it is a matter of design: automation handles the repeated question, the registration and the reminder, and passes to a person everything that calls for listening. AI helps to sort and to draft, and the team reviews and replies.
The dilemma: efficiency or warmth
Small teams receive the same question dozens of times and, at the same time, handle cases that need time and care. Automation that answers the repeated question right away, even by audio, leaves the team with what needs a person. The mistake is automating what should not be automated.
Four criteria for automation that does not dehumanize
1. Tone of voice
A flow message needs to speak the language of the cause. Sales language ("take advantage", "close now") does not go with warmth. Write the messages with the program's vocabulary, and ask the people who provide the service to review them.
2. Transparency
Whoever receives the message should know they are in an automated flow and how to talk to a person. The LGPD asks, in art. 6, for specific purposes informed to the data subject. The same goes for what happens to the data.
3. Handing over
Reports of violence, emotional crisis or a legal issue call for a person. The flow recognizes these signs and transfers the conversation, with the history, so the person does not have to repeat their story. In the Longeviver course, AI classifies the messages, and whatever needs a person goes to the team (read the case).
4. Data collected with care
Ask for consent at collection, ask only what you will use and record the actions, as art. 37 of the LGPD requires. Social data calls for extra care.
The role of AI, and its limit
AI comes in as a block: transcribe an audio message, classify a question, draft a reply or a report. The team reviews what it writes. AI does not decide who is served, does not trigger a protection network and does not act without a person in the loop.
How to choose the tool
| Situation | Path |
|---|---|
| Few steps, no sensitive data | Generic tool |
| Beneficiaries on WhatsApp, personal data, reports | Platform with a team that implements alongside you |
| Selling products or courses at scale | Marketing CRM, with the necessary adaptation |
CORRE is a no-code platform, with input, blocks and output, and the team implements alongside yours. Across projects, we add up to more than 5,000 hours automated and more than 80 thousand flow executions.
See the Programs with WhatsApp page and the ICOM case.
Questions on this topic
What is the difference between a marketing tool and a platform for social programs?
Marketing tools measure leads, funnels and sales. In a social program, what is followed is the beneficiary's journey: sign-up, attendance, feedback and result. That is why the flow design and the reports are different.
Does automated service make decisions on risk cases?
No. The flow is designed to recognize signs of sensitive situations and route the case to a person on the team. The decision and the conversation stay with people.
How does the LGPD affect automated service?
The LGPD asks for a purpose informed to the data subject, the minimum necessary data and a record of operations. In practice, the person should know they are talking to an automated flow, consent is requested at collection and each action is recorded.