Running an analysis
Starting the AI review and understanding what happens during each phase.
Once your RFP and bid documents are uploaded and parsed, you're ready to run the analysis.
Before you start
Quick checklist:
- All RFP documents show Completed status (check the Upload RFP step in the sidebar)
- The bid has at least one document with Completed status
- You have enough credits (check your balance in the top bar)
Optional but recommended:
- Evaluation criteria are configured (gives you scoring and comparison data)
Starting the analysis
Click Analyze Bids in the sidebar workflow. You'll see a card for each bid showing its current status and document count.
Click Run Analysis on the bid you want to review. The analysis starts immediately and runs in the background — you can watch the agent activity panel on the right or navigate away and come back later.
The sidebar shows a spinning loader next to "Analyze Bids" while an analysis is running.
Analyzing multiple bids
If you have several bids ready, click Analyze all at the top of the Analyze Bids page. The system runs each bid's analysis one after another automatically — when one bid finishes, the next one starts. You don't need to babysit it.
If your organization runs out of credits mid-batch, the remaining bids will stop and you'll see why. Buy more credits and re-run.
Unsupported file warning
Before starting analysis, check the bid's document view. If any uploaded files have unsupported types (e.g., AutoCAD drawings, audio or video), you'll see a warning listing them. These files are kept for reference but won't be included in the AI analysis. Make sure the important content is in supported formats.
What happens during analysis
A run bar under the run header shows four steps. Each step has a short status line under its name, and the step in progress has a thin progress bar. The footer of the agent activity panel on the right repeats the current step, for example "Check requirements · 23 of 48".
Prepare
The AI sets up an isolated workspace for the run, detects the procurement domain, and builds the list of RFP requirements to check the bid against. If you've configured evaluation criteria or lots, it pulls those in too.
What you'll see: The status line names the current sub-step: "Preparing environment", "Detecting procurement domain", "Extracting RFP requirements", "Syncing RFP requirements", and finally "Starting analysis".
Check requirements
Each requirement, evaluation criterion, and lot is checked against the proposal by its own focused agent. "Does the vendor have ISO 9001?", "How many years of experience do they claim?", and so on. Many of these checks run in parallel.
What you'll see: A counter such as "23 of 48" that climbs quickly. When there is nothing to check, the step is marked as skipped.
Review bid
This is the heavy lifting. The lead analyst compares what the RFP asks for against what the proposal says. It reads specific sections, searches across documents for evidence, and raises findings when it spots gaps, contradictions, or missing information.
Findings from the requirement checks are combined with the lead analyst's own findings. Each finding gets a provenance badge so you can see where it came from.
What you'll see: "Looking for gaps and risks" while it works, then the number of findings. New findings appear on the results page as they are recorded.
Verify
A separate AI re-checks the most important results. Think of it as a second opinion. It re-reads the evidence for every Critical and Major finding, for requirement rows and criterion values the analysis flagged, and for a sample of requirements marked as met.
- A finding can be confirmed, downgraded, or removed as a false positive.
- A requirement row or criterion value can be confirmed or corrected in either direction. A requirement marked as not met can be upgraded to met when the evidence is there, and a met one can be downgraded when it isn't.
This step exists because AI can make mistakes, and you want the high-stakes results double-checked.
What you'll see: A counter such as "3 of 12 · verdict after this", with a verification status appearing on each item as it is checked.
When the run finishes, the run bar shows the verdict instead of the steps: the recommendation (for example Submit with improvements), tinted by how serious it is, and a one-line reason built from the counts, such as "3 of 40 requirements not met, 2 partly met · 1 critical finding". Findings verified appears when the Verify step ran. Click Run steps to see the four steps again.
If the run needs your answer, the current step reads "Needs your answer". If it fails, the step reads "Stopped" and the results page says which step it stopped on.
How long it takes
This depends on document volume and complexity:
| Scenario | Typical time |
|---|---|
| Small procurement (30-50 pages total) | 5-10 minutes |
| Medium procurement (100-200 pages) | 15-30 minutes |
| Large procurement (500+ pages, many criteria) | 30-60 minutes |
The progress indicators give you a sense of how far along things are. If you navigate away, the analysis keeps running — come back anytime to check progress.
For very large procurements, the AI may briefly pause mid-analysis to reorganize its working memory. This is normal — the progress counter will resume on its own after a few moments.
Viewing results
Once the analysis completes, the sidebar shows a green checkmark next to "Analyze Bids". Click the bid name in the sidebar's Bids section to see the full results, or use the View Results button on the Analyze Bids page.
You can export a single analysis run to Excel straight from the run page using the Download Excel button, which appears in the run header once the run has finished. It's handy when you want to share or archive one bid's results without the full comparison. See understanding your results for a detailed guide on reading the report.
Clarification requests
Sometimes the AI hits something ambiguous and needs your input. When this happens, the analysis pauses and waits — its workspace is preserved exactly as the agent left it, so it can pick up cleanly when you reply.
You'll see the question on the results page (and in the activity feed). Type your answer in the clarification panel and click Send Response. The agent resumes with your input factored in — no re-run, no lost progress.
Clarifications are rare. If you're not watching the run live, check back periodically — there's no time limit on how long a paused run can wait.
Credit cost
Each analysis uses credits. The amount depends on:
- How many pages of documents there are
- Whether you've configured criteria (more criteria = more work for the AI)
- Whether verification runs (it does by default for critical/major findings)
A typical analysis of a medium-sized procurement uses 10-30 credits. See pricing for more detail.
You can't cancel an analysis once it starts - it runs to completion. This is by design: partial results would be misleading. Make sure your documents are ready before hitting Analyze.
Re-running an analysis
You can run the analysis again on the same bid at any time — for example, after uploading updated documents or changing evaluation criteria. Each run is saved independently, and you can switch between them using the run selector dropdown on the results page. See run history for details.
When things go wrong
If an analysis seems stuck (no progress for a long time):
- Check if there's a clarification request waiting for you
- Refresh the page - the analysis runs server-side, so your browser just needs to reconnect
- For very large procurements, brief pauses are normal - the AI reorganizes its working memory and continues automatically
- If it's been more than 2 hours with no progress, contact support
See troubleshooting for more common issues.