Wedge

Start with State and Local Government procurement offices evaluating software and IT services proposals. This niche faces legally mandated compliance checklists and public transparency requirements, making the pain of manual scrubbing exceptionally acute and the need for audit trails high. Expand outward into federal procurement workflows, then cross over into highly regulated commercial sectors like banking and healthcare.

Timing

Large context windows in foundational models now process entire proposal PDFs and their technical appendices simultaneously without chunking. Advanced extraction capabilities reliably parse complex pricing tables and compliance matrices that previously broke legacy OCR systems.

Why This ICP

Enterprise IT procurement teams face the highest volume of highly technical vendor proposals with rigid security and compliance requirements. Their analysts already evaluate vendors against structured scoring rubrics, providing a deterministic framework to ground model outputs.

Size Of Prize

There are roughly 40,000 enterprise and mid-market procurement departments in the US and Europe. At an average annual spend of $15,000 for proposal evaluation labor and software per department, the addressable prize is approximately $600M.

Gap Narrative

Procurement teams spend weeks manually extracting vendor compliance matrices, pricing models, and risk flags from dense proposals. Existing software tracks the RFP process but leaves the actual reading and comparative analysis of unstructured PDFs to human analysts. This creates a bottleneck where evaluation cycles stall and vendor comparisons rely on incomplete human reading.

Defensibility

Defensibility builds through workflow lock-in and accumulated proprietary evaluation schemas. As teams build their custom scoring rubrics and compliance checklists into the system, the switching costs increase significantly. However, the core document extraction capability is fundamentally a commodity tied to foundational models, meaning the moat relies entirely on deep integration into the enterprise approval and auditing system.

Why This Thesis

An Agent-based approach aligns with the workflow because the task requires autonomous reading, structured extraction, and comparative synthesis across multiple long-form documents. Software alone forces the user to do the reading, while an Agent executes the evaluation rubric directly and outputs the synthesized comparison.

Hardest Part

Extracting nested, obfuscated pricing tables and conditional clauses from unstructured 100-page vendor PDFs without dropping critical assumptions or hallucinating pricing terms.

Min Viable Scope

Focus strictly on enterprise SaaS procurement, extracting pricing matrices, SLAs, and security compliance into a standardized scoring grid. Deliberately exclude contract redlining, negotiation workflows, and physical goods RFPs.

Cold Start Problem

Bootstrapping the taxonomy of vendor 'gotchas' and pricing structures requires seeing thousands of real proposals. Break this by offering a historical audit to one enterprise procurement team, ingesting their last three years of vendor proposals to baseline the extraction engine.

Time To First Value

Same-day; the gating step is the initial ingestion of the vendor document and mapping it against the buyer's custom evaluation rubric.

Data Moat Available

true

Technical Difficulty

Moderate

SAM

~$1-1.5B US public sector procurement segment

SOM

~$20-50M

TAM

~80k government and enterprise procurement agencies x ~$50k/yr = ~$4B

Growth Rate

~12-18%/yr, driven by rising vendor bid volumes and severe shortages of qualified contracting officers

Paid Comparable Spend

~$40k-80k/yr per agency absorbed by manual contracting officer hours and external compliance consultants

Kill Thresholds

• Human edit rate > 30% per generated compliance matrix
• Security and IT rejection rate > 50% in the sales pipeline
• Pilot-to-paid conversion < 25% after 90 days
• CAC > $20k for mid-market agencies

Leading Metrics

• Time-to-first-matrix-generated
• Percentage of AI-extracted requirements manually edited
• Number of proposals scrubbed per agency per week
• Ratio of blocked versus approved vendor data uploads

What Proves Right

Contracting officers upload vendor proposals and export compliance matrices directly into their evaluation workflow without requiring manual verification. Agencies convert from pilot to paid $50k annual contracts after measuring a 50 percent reduction in time spent per RFP evaluation. High-volume procurement teams process over 80 percent of their inbound bids through the system during peak contracting seasons.

What Proves Wrong

Procurement teams abandon the tool because they distrust the AI parsing complex regulatory requirements, forcing them to manually verify every extracted data point. Government IT and security departments block deployments due to strict data privacy policies regarding sensitive vendor IP. Contracting officers revert to manual Excel matrices within 30 days because the tool disrupts their established compliance protocols.