What does Kay, Alleviate's AI assistant, do?
Kay is the AI assistant built into Alleviate. It drafts proposals, reports and explanations from the organization's own record, answers questions about the platform and about running the organization, reviews Transformation Program work against a rubric for a coach to edit, always identifies itself as an AI, and hands off to a person when asked.
What Kay works from
Kay has the context of the organization's account: its profile, programs, funding, filings, documents and the page the user is on. A question about a filing date is answered from the organization's own calendar; a request for a program summary is drafted from that program's data. Kay works inside the organization's own data and does not browse other organizations' accounts.
What Kay will not do
Kay does not present itself as a person. It does not send a proposal, a donor appeal, a report or a consequential recommendation without a person recording approval. It does not train on the organization's data. Where a question needs a lawyer, a CPA or a funder, Kay says so. A visible path to a person is always available, and handoffs are tracked.
The model behind it
Kay is built on Anthropic's Claude models as the primary provider, with OpenAI models in a bounded set of older surfaces. Model providers process the text of a request to answer it and do not use Alleviate customers' data to train their models.
The figures
- Kay always identifies itself as an AI assistant
- Human approval recorded before proposals, donor appeals, reports or consequential recommendations go out
- Primary model provider: Anthropic (Claude); OpenAI in a bounded set of older surfaces
Related answers
See it with your own organization. Starter is free, no card, and the first program takes ten minutes.
Start free