How the AI agents work

You might wonder why the system uses multiple AI agents instead of just feeding everything into one big model. Short answer: it produces better results for less money. Here's the longer version.

The problem with a single model

If you gave one AI model a 500-page RFP and a 300-page proposal and said "compare these", two things would happen:

1. It would cost a fortune (AI pricing is based on how much text you send)
2. Important details would get lost (models lose accuracy on very long inputs)

The multi-agent approach solves both problems by splitting the work.

The team

Orchestrator agent

This is the coordinator. It reads the RFP to understand what's required, then plans what to check. It doesn't read every page of every document itself - it delegates.

When it needs to verify a specific requirement ("Does the vendor have ISO 9001?"), it asks one of the specialist agents to find the answer. Based on what comes back, it records findings, adjusts severity levels, and builds the overall assessment.

The orchestrator uses the most capable (and expensive) AI model, but only for coordination and judgment - not for bulk reading.

Document reader agents

These agents read specific sections of documents when the orchestrator asks. "Read pages 45-50 of the technical proposal" or "Read the financial offer summary." They use a faster, cheaper model because the task is straightforward: read and report back.

Search agents

Instead of reading an entire 300-page proposal to find one piece of information, the search agents use semantic search. They find the relevant passages quickly - "find all mentions of environmental compliance" returns the specific paragraphs without reading everything else.

This is both faster and cheaper than having an AI read the whole document.

Verification agent

After the orchestrator finishes its analysis, a separate agent reviews the Critical and Major findings. This agent re-reads the evidence, checks the reasoning, and confirms or challenges each finding.

Using a separate agent for verification is intentional. It's like having a colleague review your work - they bring fresh eyes and might catch errors the first reviewer made.

How a typical analysis flows

1. Orchestrator reads the RFP and builds a list of requirements to check
2. For each requirement, the orchestrator decides whether to search or read:
   • "Find all mentions of team qualifications" → search agent
   • "Read the pricing section starting on page 12" → reader agent
3. Specialists report back with the relevant text
4. Orchestrator evaluates the evidence and records findings
5. After all requirements are checked, verification agent reviews the serious findings
6. Report is compiled from all the recorded findings, scores, and evidence

Why this matters for you

You don't need to know any of this to use the platform. But it helps explain:

• Why analysis takes time - multiple agents are doing multiple passes through your documents
• Why evidence is specific - the agents quote exact passages, not summaries
• Why verification can change findings - it's a separate check, not just rubber-stamping
• Why costs vary - more documents and criteria mean more agent work
• Why results are generally reliable - multiple perspectives catch more than a single pass