Skip to content
C

Trust

AI Disclosure

Last updated: June 13, 2026

CIFIS uses AI to power Ask CIFIS and supporting features. This page explains how AI is used, what it will not do, and the guardrails behind it.

Where AI is used

  • Ask CIFIS — read-only copilot grounded in your organization's data and the CIFIS funder and opportunity catalog.
  • Intelligence Briefings — explainable summaries derived from structured records.
  • Strategic Relationship signals — graph-derived suggestions based on focus areas and beneficiary overlap.

What Ask CIFIS will NOT do

  • Draft grant proposals, narratives, LOIs, or cover letters.
  • Generate budgets or financial documents.
  • Provide legal, tax, or investment advice.
  • Make funding decisions or guarantee outcomes.
  • Answer questions outside its four allowed intents (program analysis, funding readiness, funder intelligence, gap analysis).

How Ask CIFIS is grounded

Conceptual workflow

  1. Program-grounded

    Your program profile in scope

  2. Authorized context

    Only records you can read

  3. CIFIS records

    Funder, opportunity, CAPS

  4. Grounded response

    Deterministic reasoning

  5. Citations

    Source + date on every claim

Ask CIFIS reads only records the caller is authorized to see, retrieves relevant CIFIS records, generates a grounded response and returns citations. Internal prompts and infrastructure are not exposed.

Source: Ask CIFIS pipeline — documented in methodology

Ask CIFIS only answers using data already in CIFIS — your organization profile, programs, CAPS Lite output, funder catalog, opportunities and saved pipeline. It does not browse the open web. When the underlying data is insufficient, Ask CIFIS will say so rather than speculate.

Tenant isolation in AI calls

AI prompts are constructed server-side and constrained to records the requesting user is authorized to read. Cross-tenant data is never included in a prompt.

Model providers & training

We use enterprise AI providers via the Lovable AI Gateway. We do not permit customer data to be used to train third-party foundation models. Prompts and responses may be retained briefly by providers for abuse monitoring, per their enterprise terms.

Human-in-the-loop

AI outputs are decision support, not decisions. Users are responsible for reviewing every output before acting on it. See our Disclaimer.

Return to homepage