Documents

6/1334 → 11/976; +0.68 percentage points of formation share.

The pattern across this batch is a steady move away from generic document automation toward domain-specific extraction built for one professional workflow at a time: Landeed on Indian property records, InQuery on medical-legal record review, Docsum on contract renewal and obligation data, Provision on construction contracts and drawings. That verticalization matters because the underlying parsing problem, reading messy PDFs, scans, and layouts, has become commoditized enough that companies like Reducto, Extend, and TableFlow can sell the raw extraction layer as infrastructure, freeing later entrants to compete on judgment within a specific document type rather than on parsing accuracy itself. There's a visible split in the batch between these two layers, the plumbing (Extend, Fileforge, TableFlow) and the applied judgment sitting on top of it (Docsum, InQuery, Landeed, Provision, Mulligan), and the fact that Mulligan already got acquired by 2026 under a new name, Insure Stack, suggests the applied layer is where value gets captured fastest. None of this would have been economically viable much earlier than 2023, when vision-capable language models finally got good enough to read arbitrary layouts, tables, and handwriting without custom OCR pipelines for every document type. At 2.51x share growth, Documents is accelerating precisely because the cost of building a narrow, defensible product on top of that shared parsing layer dropped so fast that founders coming out of PwC audits or construction estimating could ship a working vertical tool in a single batch cycle.

No comparable outside activity is connected yet.