Cap Table Document Automation: Closing the Paper Gap

A law firm doing diligence on a Series B round pulled the target company's cap table and, as a matter of routine, matched every preferred stock line against the board consent and stock purchase agreement that supposedly authorized it. One investor's row showed 120,000 shares of Series Seed Preferred. The signed stock purchase agreement, Schedule A, showed 80,000. The board consent authorizing the round showed an aggregate share count that only worked if that investor held 80,000, not 120,000. Nobody had touched the cap table maliciously. Someone updated it after a verbal confirmation of a larger check, the paperwork never caught up, and the spreadsheet had been treated as correct for three closings in a row. That is not a rare, dramatic failure. It is what happens by default when a cap table is maintained as a running belief about ownership instead of a reconciled summary of signed documents.

This post is about the mechanics of that gap: what a cap table actually is under corporate law versus what people treat it as, which documents create the legal record it is supposed to summarize, why the two drift apart even at well-run companies, and what a real reconciliation process, automated or manual, has to check line by line. For the broader document processing pipeline this fits into, our legal document intelligence platform covers extraction across contracts, corporate filings, and governance documents generally. Here the focus is narrower and more mechanical: cap tables specifically, and the paper that is supposed to back every row in one.

The Cap Table Is a Model, Not a Legal Record

Most people, including a fair number of lawyers who should know better, talk about "the cap table" as if it were the company's official record of who owns what. It isn't, and the distinction is not academic. Delaware corporate law, under which the large majority of venture-backed US companies are incorporated, defines a specific legal record for stock ownership: the stock ledger. Section 224 of the Delaware General Corporation Law defines "stock ledger" as one or more records administered by or on behalf of the corporation in which the names of all stockholders of record, the address and number of shares registered to each stockholder, and all issuances and transfers of stock are recorded, and it cross-references the exact statutory obligations that ledger has to satisfy: it must be capable of producing the stockholder list required under Sections 219 and 220, it must be able to record the information specified in Sections 156, 159, 217(a), and 218, and it must record transfers as governed by Article 8 of the Delaware Uniform Commercial Code. That is a specific, enumerated statutory definition with real teeth, because Section 220 gives stockholders an enforceable right to inspect that ledger, and in a dispute a Delaware court looks to whether a document actually meets the Section 224 definition of a stock ledger, not to whether a spreadsheet labeled "cap table" looks authoritative.

A cap table, by contrast, is not defined anywhere in the DGCL. It is a modeling artifact: usually a spreadsheet or a table inside cap table software, built to answer questions the statutory stock ledger was never designed to answer quickly, like fully diluted ownership percentages, option pool sizing, or waterfall outcomes under different exit scenarios. Those are genuinely useful questions and the reason cap table tools exist. But the cap table's authority is entirely derivative. It is correct only to the extent it matches the stock ledger and the underlying instruments that created each entry, and nothing in how most companies operate forces that match to hold continuously. The stock ledger updates (or should update) the moment stock is issued or transferred, because it is a legal obligation. The cap table updates whenever whoever maintains it gets around to it, based on whatever information they were handed, which is sometimes a signed document and sometimes an email summary of one.

The Documents That Actually Create Each Cap Table Row

Every legitimate line on a cap table should trace back to a specific instrument, and each instrument has its own statutory basis and its own required content. Reconciliation is the process of confirming that trace exists and that the numbers along it agree at every step.

Document	Legal basis (Delaware)	What it authorizes or evidences	Field it should match on the cap table
Board consent or board resolution	DGCL Section 141(f) permits board action without a meeting by unanimous written consent	Authorizes the corporation to issue a specific class and number of shares, at a specific price or for specific consideration	Aggregate shares issued in the round, per class
Stock purchase agreement (with signature-backed schedule of investors)	DGCL Section 152 gives the board authority to determine the consideration for shares, which may be cash, property, or services	Sets the per-investor allocation, price per share, and closing conditions	Per-investor share count and price paid, summing to the board-authorized aggregate
Stock certificate or uncertificated share notice	DGCL Section 158 permits shares to be certificated or, if the certificate of incorporation or bylaws allow it, uncertificated	Evidences that shares were actually issued to a named holder, not merely authorized	Confirms the share issuance actually closed, as opposed to being authorized but never executed
Option grant agreement and equity incentive plan	Board approval under the plan document, typically also requiring 409A-compliant valuation for pricing	Authorizes options for a named employee at a strike price tied to fair market value	Options outstanding by holder, vesting schedule, and pool usage against the plan's reserved share count
SAFE or convertible note	Contract governed by its own terms; converts into equity per its valuation cap, discount, and MFN provisions at a triggering event	Creates a contingent right to future shares, not present equity	Pre-conversion: tracked separately from issued shares. Post-conversion: converted share count must match the conversion mechanics in the instrument
Stock ledger	DGCL Section 224 defines its required content and cross-references Sections 156, 159, 217(a), 218, 219, and 220	The corporation's definitive record of who holds what, as of when	Should equal the cap table's issued-share section exactly; any divergence means one of the two is wrong

Notice that four different documents, each governed by a different section of the DGCL, all have to agree with each other and with the cap table for a single row to be genuinely correct: the board consent has to authorize an aggregate that matches the sum of the stock purchase agreement's per-investor allocations, the stock purchase agreement's allocations have to match what actually got certificated or entered as uncertificated shares, and the stock ledger has to reflect all of it. A cap table reconciliation that only checks the cap table against one of these, commonly just the stock purchase agreement, will miss gaps that show up between the other pairs.

Five Ways Cap Tables Actually Drift

Drift is not usually one dramatic error. It is an accumulation of small, individually reasonable shortcuts, each of which breaks the chain between a cap table row and the document that should support it.

• Schedule amendments that never get re-executed. A stock purchase agreement's investor schedule gets redlined by email during closing, an investor's allocation changes, and the final number makes it into the cap table because someone was on that email thread. The signature page, however, was executed against the original schedule, not the amended one. Legally, the signed document still says the old number.
• Secondary transfers logged in one place and not the other. An existing stockholder sells shares to a new holder in a transaction the company approves but which gets recorded in the transfer records or the stock ledger and never propagated back into the cap table tool, or vice versa, updated in the cap table tool by whoever has access without a corresponding entry in the actual transfer ledger.
• Option exercises that update headcount but not the pool math. When an option holder exercises, the exercised shares should move from "options outstanding" to "common shares issued" on the cap table, and the option pool's available balance should update accordingly. If only one side of that entry gets made, the fully diluted totals stop tying out even though no new security was created.
• Convertible instrument conversion errors. SAFEs and notes convert according to formulas involving a valuation cap, a discount rate, and sometimes a most-favored-nation clause that can retroactively change earlier investors' terms if a later instrument has better terms. Getting this math wrong, or applying it at the wrong price per share, produces a converted share count that does not match what the instrument's own terms would produce if recalculated independently.
• Manual re-keying between systems. Even when every document is correct and every approval is properly executed, someone still has to read the board consent, the stock purchase agreement, and the option grant and type the resulting numbers into cap table software. Every manual re-keying step is an opportunity for a transposition error, a missed row, or a stale figure carried forward from a prior draft.

None of these require bad faith. They require exactly what most companies actually have: multiple people touching different documents at different times, using different tools, under time pressure to close a round, with no automated check forcing every number back to its source before it gets treated as final.

A Worked Reconciliation Gap

Here is the mechanism from the opening example, worked through with numbers so the gap is checkable rather than asserted.

A company runs a Series Seed round. The board consent, executed under DGCL Section 141(f), authorizes the issuance of 1,200,000 shares of Series Seed Preferred Stock at $1.25 per share, for aggregate proceeds of $1,500,000. The stock purchase agreement executed the same day lists three investors on Schedule A, with per-investor allocations that are required to sum to the board-authorized aggregate.

Source	Investor A	Investor B	Investor C	Total shares
Board consent, authorized aggregate	not itemized by investor	not itemized by investor	not itemized by investor	1,200,000
Stock purchase agreement, Schedule A (signed)	600,000	520,000	80,000	1,200,000
Cap table as maintained	600,000	520,000	120,000	1,240,000
Reconciliation gap	0	0	+40,000	+40,000

Investor C's allocation increased from 80,000 to 120,000 shares late in closing, based on an additional $50,000 the investor wired after an email exchange confirming a larger check. That $50,000 divided by the $1.25 per-share price is exactly 40,000 additional shares, arithmetically consistent, which is precisely why the change felt safe to enter directly into the cap table. But the schedule attached to the signed stock purchase agreement was never amended and re-executed to reflect the larger allocation, and the board consent's authorized aggregate of 1,200,000 shares was never increased to 1,240,000 to match it. The cap table total of 1,240,000 shares is 40,000 shares, or 3.33 percent of the round, higher than either the board actually authorized or the signed purchase agreement supports. Every fully diluted ownership percentage calculated off that cap table, for every stockholder in the company, is now slightly wrong, not just Investor C's line, because fully diluted percentages are computed against the total share count.

The fix is not simply editing the cap table back down to 1,200,000. The company took real money for 120,000 shares' worth of stock, so the actual defensible fix is to go back to the board and secure a proper consent authorizing the increased issuance, then have Investor C execute an amended Schedule A or a separate purchase agreement for the additional 40,000 shares, and only then does the cap table's 1,240,000 figure become supportable. Until that paperwork exists, the cap table is reporting a number no signed document actually backs, which is exactly the kind of gap a diligence review, a 409A valuation, or a future investor's own reconciliation is likely to catch, usually at a worse time than during the original closing.

What Automated Reconciliation Actually Has to Do

Solving this by hand means a paralegal or associate opening the board consent PDF, the stock purchase agreement PDF, and the cap table export side by side and manually cross-checking every row, which is exactly what happened in the diligence example that opened this post. It works, but it does not scale past a handful of rounds and it is easy to skip under deadline pressure, which is exactly when it matters most.

A document-automation approach to this problem has to do three things that a generic OCR pass does not do on its own. First, it has to extract structured fields, not just raw text, from each document type: from a board consent, the authorized share class, share count, and price per share; from a stock purchase agreement, the per-investor allocations on the schedule and the aggregate they sum to; from an option grant, the holder, share count, strike price, and vesting start date. Table extraction matters specifically here, because the investor schedule in a stock purchase agreement and the cap rate table in a SAFE are both tabular data embedded in prose documents, and a system that only extracts running text will miss the numbers that actually need reconciling.

Second, it has to resolve entities across documents that refer to the same thing with different labels. "Series Seed Preferred Stock" in a board consent, "Seed Preferred" in a stock purchase agreement, and "Series Seed" in cap table software all need to be recognized as the same security class before any numeric comparison is meaningful. The same problem applies to investor names, which routinely appear as an individual's name in one document and a trust or LLC entity name in another, referring to the same beneficial holder.

Third, and this is the step that actually catches gaps like the worked example, it has to perform the numeric cross-check: sum the per-investor allocations extracted from the stock purchase agreement and confirm the total matches the aggregate authorized in the board consent, then confirm the cap table's per-investor and total figures match both. Any mismatch gets flagged with the specific documents and line items involved, rather than surfacing as a vague "cap table may be inaccurate" warning. That specificity is what makes the flag actionable, a reviewer can go straight to the discrepancy instead of re-auditing the entire round.

Manual Reconciliation vs Field-Level Automated Reconciliation

Aspect	Manual reconciliation	Field-level automated reconciliation
How a gap is found	A reviewer happens to compare the right two documents, usually during diligence or an audit	Every issuance is checked against its source documents automatically at the time it is entered
Coverage across rounds	Practically limited to sampling a subset of rounds given reviewer time	Every round and every instrument type can be checked consistently
Detection timing	Often months or years after the gap was created, frequently during a financing or exit event	At or near the time the discrepancy is entered, before it compounds across later rounds
What gets flagged	Whatever the reviewer happens to notice, dependent on individual attention to detail	Every numeric mismatch between extracted fields, regardless of whether a human would have spotted it
Auditability	Reviewer's notes, if kept, are the only record of what was checked	Each cap table row can carry a link back to the specific source document and field it was extracted from

Where an Unreconciled Cap Table Becomes a Legal Problem

A cap table that is wrong by a few thousand shares can sit unnoticed for years if nobody is forced to check it against the paper. It tends to surface at exactly the moments when getting it wrong is most expensive. A 409A valuation prices options against the company's capital structure, and an inflated share count changes the per-share value the valuation produces, which then affects every option strike price set off that valuation. A stockholder exercising rights under DGCL Section 220 to inspect books and records, including the stock ledger, is entitled to the actual statutory record, and a company that can only produce a cap table export that does not match its stock ledger has a real disclosure problem, not just a bookkeeping inconvenience. In an acquisition, the buyer's counsel will typically require representations that the cap table is accurate and complete, and reconciliation gaps discovered during that diligence process routinely become last-minute closing conditions or purchase price adjustments, which is a far more expensive place to find a 40,000-share discrepancy than at the original closing. Qualified Small Business Stock treatment under Section 1202 of the Internal Revenue Code also depends on accurate records of when and how shares were originally issued, since the tax benefit is tied to the holding period and original-issuance status of specific shares, which is exactly the kind of fact a reconciled stock ledger and cap table are supposed to establish cleanly.

A Reconciliation Checklist for Law Firms and Fund Administrators

For firms handling cap table review as part of corporate work, diligence, or fund administration, a few concrete checks catch most real-world gaps without requiring a full system rebuild:

• For every issuance on the cap table, confirm a board consent exists authorizing that specific class and share count, and that the consent's aggregate matches the sum of individual allocations in the corresponding purchase agreement.
• For every convertible instrument shown as converted, recompute the conversion independently from the instrument's stated cap, discount, and trigger price, and confirm the resulting share count matches what the cap table shows.
• For every option exercise, confirm the pool's available balance decreased by the exercised amount and the common share count increased by the same amount, with no net change in the fully diluted total.
• Compare the cap table's total issued shares against the stock ledger's total, not just against a sample of individual rows. A row-by-row spot check can miss a systemic gap that a total-to-total comparison catches immediately.
• Treat any schedule amendment made after signature as unresolved until a properly executed amendment or a new instrument exists, even if the number was communicated informally and everyone agrees on what it should be.

None of this is exotic. It is the same discipline as reconciling a bank statement against a general ledger, applied to equity instead of cash. The difference is that a bank reconciliation happens monthly by default because the bank sends a statement whether anyone asks for it or not. A cap table has no equivalent forcing function, which is exactly why it drifts, and exactly why the reconciliation has to be deliberate rather than assumed. For firms handling diligence-heavy transactional work generally, our guide to due diligence data room software covers the broader document review workflow this fits into, and our contract OCR guide covers the extraction mechanics behind reading structured terms out of the underlying purchase agreements and instruments themselves. Written by Nupura Ughade.