Some features sit in the backlog for months. You know they matter, you know they're coming, but there's always something more urgent. Pipeline was one of those features.
Then yesterday happened.
The Call That Changed Everything
A managing partner at a VC fund sent us a spreadsheet. Sixteen columns. A handful of startups from their shortlist. Company names, websites, pitch deck links, funding stages, sectors, traction notes -- the exact format their team uses internally to track dealflow.
The ask was simple: "Can your AI evaluate these? We don't want to manually enter each one."
We'd had "Pipeline" on the backlog since Q4 2025. Priority P2. Target: Q2 2026. Nice to have.
After that call, it became P1. Target: tonight.
What We Built (In One Night)
Upload Anything
Investors can now upload CSV or Excel files directly into Athanor. The system parses the file, auto-detects column headers, and maps them to our idea schema. "Company name" becomes the title. "Current funding stage" maps to funding data. "Pitch Deck (PDF)" gets its links checked.
If your headers don't match our defaults, no problem. The mapping step lets you manually assign any column to any of 44 fields across all eight sections of our evaluation form -- from basic overview data down to TAM/SAM/SOM, CAC/LTV, cap table, and go-to-market strategy.
The system remembers your last mapping. Import from the same tool twice, and it auto-applies your previous column assignments.
Every Link Gets Checked
Deal lists are full of URLs. Company websites, pitch decks on Google Drive or SharePoint, demo links, Twitter profiles. Before evaluation, every URL in every row gets an HTTP health check. Accessible links get a green badge. Dead links, 404s, timeouts, DNS failures -- they all get a red badge with the specific error.
This matters because a pitch deck link that returns 403 shouldn't silently break the AI's analysis. Now you see the problem before you spend credits.
Two-Step Assessment
We deliberately separated "prepare" from "evaluate." When you click Prepare, Athanor creates ideas from your import data and deducts credits. But evaluations don't start until you explicitly click Start Evaluations.
This gives you a window to:
- Review the mapped data
- Add investor notes (context for the AI -- "met this team at a conference, strong technical background")
- Attach founder LinkedIn profiles for team verification
- Fix any issues from the link check
When you're ready, hit Start. Evaluations run in the background. Each takes 5-15 minutes depending on the report tier (Screening, Assessment, or Verification).
Side-by-Side Comparison
Once evaluations complete, select 2-4 deals and open the comparison view. GemScore reports displayed in columns -- potential scores, readiness scores, grade bands, strengths, risks, agent breakdowns, next steps. The highest score in each row gets highlighted. Fatal flags get red treatment.
It's the IC prep meeting, compressed into one screen.
Team Sharing
Share any batch with your team. Access follows your existing team RBAC:
- Owners and Admins can do everything -- edit items, trigger evaluations, delete batches
- Editors can update items, add notes and LinkedIn profiles
- Viewers get read-only access to the batch, items, comparison, and reports
Shared batches appear in your teammates' Pipeline page with a team badge.
Report Editing
For pipeline deals, investors can edit the AI-generated reports directly. Same inline editing that admins and partners have -- click any field, modify it, save. Full audit trail. Restore to AI original at any time. Every edit is tracked with an "investor" editor type.
Because sometimes the AI gets a detail wrong, and you need to fix it before sharing the report with your IC.
Thesis Auto-Update
Here's the quiet feature that compounds over time. After a batch completes, Athanor analyzes the deals and updates your investment thesis automatically:
- Importing CleanTech deals from Europe? "CleanTech" gets added to your focus areas, the relevant geography to your preferences
- Raising amounts in a certain range? Your check size range adjusts
- Pre-seed stage? The corresponding stage gets added to your preferred stages
It's additive only -- never removes your existing preferences. And there's an opt-out toggle in your thesis settings if you prefer manual control.
The AI Learns Your Context
Two things make the pipeline evaluation smarter than a standalone submission:
Investor Notes. You can add free-text context per deal: "Our LP knows the CEO," "Strong repeat founders," "We're skeptical about the unit economics." This gets injected into the AI evaluation as additional context, giving agents signal that a cold submission wouldn't have.
Founder LinkedIn Profiles. Add LinkedIn URLs for each founder, and the AI's Team Analyzer agent uses them for verification -- cross-referencing backgrounds, checking if the claimed experience matches public profiles, and flagging discrepancies.
What This Means for Investors
If you're a solo GP reviewing a batch of 20 companies from a syndicate lead, you no longer need to enter them one by one. Upload the spreadsheet, let the AI screen them, compare the top candidates, and have your IC prep done in under an hour.
If you're a fund running a structured pipeline in Affinity or DealCloud, export your shortlist, upload it here, get AI-powered second opinions on every deal, and export the results back with GemScore data appended.
If you're evaluating a cohort from an accelerator program, import the batch, share it with your investment team, and let everyone annotate their favorites before the partner meeting.
The Backlog Lesson
We shipped Pipeline in one session. The spec was clear because we'd been thinking about it for months. The architecture was ready because we'd built the evaluation infrastructure, the team RBAC, the report editing system. All we needed was the forcing function.
Sometimes the best product decisions come from a single customer call.
Pipeline is live now for all investor accounts. Go to Pipeline in the navigation bar, upload your first spreadsheet, and let us know what you think.
Have a custom pipeline format? We'd love to see it -- contact us and we'll make sure the column mapper handles your export.
