AI for the Impact Sector

What we build

Chatbots, voice tools, and analytics that run on your own programme data, inside your own systems.

WhatsApp-native chatbots

Beneficiary support, FAQs, service navigation, and data collection on the channel people already have open. 20+ deployments running.

Voice AI and spoken assessments

Speech interfaces in Indian languages, including assessments scored from a spoken answer, for low-literacy and low-connectivity settings.

Toll-free IVR helplines

Interactive voice response with localised content, for people calling in from a basic phone. One of these has been in continuous operation for over a decade.

Document summarisation

Field reports, partner submissions, and proposals turned into a readable brief, with the source paragraph still attached.

Report and content generation

Donor reports, case notes, and programme updates drafted from the data already in your MIS, for a person to edit rather than write from scratch.

Chat with your own data

Ask your MIS a question in plain language and get the number back with the query behind it. MCP-ready, and you can plug in your own API key.

Recommendation engines

Next-best-action for a field worker, or which households a supervisor should visit this week, ranked from your own programme data.

Natural language analytics

The analytics layer over programme data, queried in words instead of filters, for leadership and donors.

How it plugs in

Most of our systems now ship MCP-ready with chat-with-your-data built in. You bring your own API key, so the model provider is your choice and your bill.

On top of your MIS

The AI layer reads the system you already run, whether that is mGrant, mForm, or a platform someone else built.

Your key, your data

Bring your own API key. Data residency and retention follow the same DPDP rules as the rest of your stack.

Multilingual by default

Multilingual natural-language understanding, so a field officer and a programme director can each ask in their own language.

Radar, if you need it packaged

Radar is our own product built this way: due diligence and risk intelligence on partners and grantees.

20+Chatbot deployments

Voice + textInterfaces in Indian languages

10+ yearsLongest-running voice service

Talk through an AI use case →

Common questions

Does our programme data go to a model vendor?

You bring your own API key, so the model provider is your choice and your bill. The AI layer reads the system you already run rather than exporting it elsewhere.

How do we know the answer is right?

Answers come back with the query or the source paragraph attached, so a person can check the number rather than take it on trust.

Do we need a new system for this?

No. Most of our systems already ship MCP-ready, and the layer sits on top of whatever MIS you have, including one someone else built.

Answers from the data you already hold.