Skip to main content

Documentation Index

Fetch the complete documentation index at: https://grantmaster.dev/llms.txt

Use this file to discover all available pages before exploring further.

Feature Guide: AI Assistant & Agents

Overview

GrantMaster integrates Google Gemini AI through two complementary systems: a conversational AI Assistant for interactive help and an Agent Framework for automated multi-step workflows. Both systems use a credit-based billing model.

AI Assistant

Purpose

The AI Assistant is a contextual chat interface that helps with grant writing, data analysis, compliance review, and general platform questions. Where: Click the AI chat icon in the header, or use Cmd+K > “AI Assistant”

Capabilities

FeatureDescription
Grant WritingDraft proposal narratives, budget justifications, and concept notes
Compliance AnalysisReview documents for compliance issues and extract requirements
Data AnalysisSummarize project data, spending patterns, and impact metrics
Report DraftingGenerate narrative sections for donor reports
Document Q&AAsk questions about uploaded documents using RAG (Retrieval-Augmented Generation)
Role RecommendationsSuggest organizational role changes based on activity patterns
Receipt OCRExtract data from receipt images for expense entry

Using the Assistant

  1. Open the AI chat interface
  2. Type your question or request in natural language
  3. The assistant has context awareness — it knows which page you are on and can reference your organization’s data
  4. Chat history is preserved across sessions for continuity

Example Prompts

  • “Draft a project narrative for our maternal health program targeting USAID”
  • “What are the reporting requirements for Grant #2024-0156?”
  • “Summarize our expense trends for Q1 2026 across all projects”
  • “Extract compliance deadlines from the uploaded grant agreement”
  • “Help me write a budget justification for the personnel line item”

AI Agents

Purpose

Agents are autonomous multi-step workflows that perform complex tasks without continuous user input. They combine AI reasoning with platform data and tools to complete structured work. Where: AI > Agent Dashboard

How Agents Work

  1. Select an agent from the available catalog (categorized by function)
  2. Configure inputs — provide required parameters and context
  3. Execute — the agent runs through its steps, making decisions and calling tools
  4. Review results — examine outputs, approve actions, or provide feedback
  5. Escalation — if the agent encounters a decision it cannot make, it escalates to you for human-in-the-loop input

Available Agent Categories

CategoryExamples
Grant WritingProposal narrative agent, budget builder agent
CompliancePolicy extraction agent, compliance audit agent
ReportingMonthly report agent, donor report compilation agent
Data AnalysisSpending analysis agent, impact metric summarizer
AdministrationJournal review agent, expense categorization agent

Agent Run Dashboard

Track all agent executions with:
  • Status (running, completed, failed, escalated)
  • Credit cost per run
  • Step-by-step execution log
  • Output artifacts (generated documents, analyses, recommendations)

Credit System

How Credits Work

AI features consume credits from your organization’s balance:
  • Chat messages — 1 credit per AI response
  • Document analysis — 2–5 credits depending on document size
  • Agent runs — 5–50 credits depending on complexity and steps
  • Vision (OCR) — 2 credits per image processed

Monitoring Usage

Where: Settings > Billing > AI Credits
  • View current credit balance and transaction history
  • See usage breakdown by feature and user
  • Set usage alerts at configurable thresholds (e.g., warn at 20% remaining)
  • Purchase additional credits or upgrade your subscription tier for higher allocation

Credit Reservation

When an agent starts, it reserves the estimated credits upfront. If the run completes under budget, unused credits are returned. If it fails, reserved credits are released.

Use Cases

  • Proposal Season: Use the Grant Writing agent to generate first drafts of narrative sections, then refine with the AI Assistant’s conversational interface
  • New Grant Onboarding: Upload a new grant agreement and run the Policy Extraction agent to auto-create compliance policies
  • Quarterly Reporting: Use the Report Compilation agent to pull data from multiple projects and generate a donor-ready report draft
  • Expense Review: Run the Expense Categorization agent to review and suggest corrections for bulk expense uploads
  • Document Intelligence: Ask the AI Assistant questions about your uploaded documents to quickly find specific clauses, deadlines, or requirements

Best Practices

  • Provide context in prompts — mention specific project names, grant numbers, and time periods for more relevant responses
  • Review AI outputs carefully — AI-generated content is a starting point, not a final product
  • Use agents for repetitive tasks — they are most valuable for structured work that follows a consistent pattern
  • Monitor credit usage — set alerts before your balance gets low to avoid service interruptions
  • Build a prompt library — save effective prompts as templates for your team to reuse

Limitations

  • AI responses are generated by Google Gemini and may contain inaccuracies — always verify facts and figures
  • RAG (document Q&A) works best with well-structured documents; scanned images without text layers have limited support
  • Agent executions are sequential — you can run one agent at a time per organization
  • Credits are non-refundable once consumed; failed runs release reserved (unconsumed) credits only
  • AI features require an active internet connection — they do not work with local emulators
  • The AI Assistant does not have access to external websites or real-time internet data; it only works with your organization’s uploaded documents and platform data