When a closing condition slips or a post-closing covenant gets missed, the after-action review almost never finds carelessness. It finds structure. The deadline was in writing. Everyone involved was competent. But the obligation lived in a document nobody reopened, on a timeline nobody owned, in a spreadsheet nobody else could read. The failure was architectural, which is the good news, because architecture can be fixed.
Transactional practice produces obligations at industrial scale. A midsize acquisition generates hundreds of them: conditions to closing, consents to chase, notices to deliver, covenants that run for years after the wire clears. Every one is written down somewhere. The problem has never been documentation. It is that documentation and attention live in different places, and deal teams disband while obligations do not.
Why obligations slip
Blame rarely survives contact with the facts. Four structural realities do most of the damage.
- Obligations are scattered. The definitive agreement holds some. The rest hide in disclosure schedules, side letters, third-party consents, the escrow agreement, and the transition services agreement. No single document contains the deal's full obligation set, so no single reading surfaces it.
- They mutate across drafts. The notice period that was ten business days in the third draft is five in the executed version. Anyone working from memory of the negotiation is working from the wrong document.
- Ownership is ambiguous. Some obligations belong to the firm, some to the client's operations team, some to the counterparty. The boundary is rarely written down, and an obligation that belongs to everyone belongs to no one.
- The tracker is a spreadsheet. The closing checklist lives in a workbook one associate built and one associate understands. Version four was emailed, version five lives on someone's desktop, and neither carries a link to the language that created the deadline. When that associate rotates to the next deal, the working memory of the deal's obligations goes too.
What extraction actually produces
The mechanical fix is extraction with provenance. A system reads the executed set, all of it, and produces a register in which every obligation carries its obligor, its obligee, its trigger, its deadline, its notice requirements, and its cure period, each linked to the clause it came from. The link is the point. A register entry that cannot show its source clause is an assertion. One that can is a fact an attorney verifies in one click.
Scope discipline matters here. The extraction runs on the executed documents, amendments included, not on whichever draft someone had open. A register keyed to the wrong version is worse than none, because it gets trusted.
Extraction also has to be honest about ambiguity. Commercially reasonable efforts, deadlines measured from events that have not occurred yet, obligations conditioned on facts outside the documents: these get flagged for attorney interpretation, not silently normalized into a date. The register records the reading an attorney chose, and who chose it.
Built properly, the register is not a standalone artifact. It is a view of the underlying graph of the deal: documents, parties, clauses, dates, and the relationships among them. That distinction earns its keep the first time an amendment lands. In a spreadsheet world, someone re-reads the amendment and hand-edits rows. In a graph world, the amendment is ingested, the affected obligations update, and the delta is visible: which deadlines moved, which notice requirements changed, which conditions fell away. On Reframe, obligation and risk trackers are a standard Harness build on the Context Graph, which is exactly this pattern: extraction into structure, register as a live view.
Signing to closing
The pre-closing window is where tracking earns trust, because deadlines are dense, both sides are watching, and the cost of a miss is immediate and visible.
Conditions precedent get individual owners and live status, not a list refreshed before the weekly call. Regulatory milestones, HSR and its equivalents in other jurisdictions, carry filing dates, waiting periods, and second-request contingencies. Third-party consents are chased with the notice provisions attached, so whoever sends the request knows the required form, addressee, and delivery method without reopening the underlying contract. The associate running the checklist stops being a human index and becomes a manager of exceptions.
The client sees the same register, filtered to what it owns. Half of what closes a deal is operational work on the client side, and a shared view with explicit ownership beats the weekly status email it replaces. The all-hands call gets shorter too, because status lives in the register instead of in twenty minutes of roll call.
After closing, the long tail
Closing is when firm attention historically falls away, and it is when the register matters most, because the remaining deadlines stretch years past anyone's active memory of the deal.
- Survival periods. When each category of representation expires, and the notice a claim must satisfy before it does.
- Earnout covenants. Measurement periods, the operating covenants that protect the earnout, and the information rights the seller can enforce while it runs.
- Escrow releases. Release dates, release conditions, and the notice mechanics that actually move the money.
- Restrictive covenants. Non-compete and non-solicit expirations, which shape the client's hiring and strategy long after the deal team disbands.
- Transition services. Service end dates, extension windows, and the termination notices that prevent an unwanted renewal.
None of these dates are hard to compute at closing. All of them are easy to lose eighteen months later, when the deal team has dispersed and the client assumes the firm is watching. The register is the difference between a well-timed follow-on engagement and an awkward phone call.
Who watches the watcher
A register without owners is a prettier spreadsheet. Every deadline needs a responsible attorney, an escalation path if it approaches unacknowledged, and alerts that arrive with the source clause attached, so the recipient can act without a research detour. The audit trail matters just as much: who was notified, when, and what happened next. If an obligation is ever missed despite the system, the firm should be able to show exactly where the chain broke. That record is also what clients, and increasingly malpractice carriers, ask about when they evaluate how a firm supervises automated systems.
Alerting is where automation meets governance. Reminders are safe to automate. Anything that acts outward, sending a notice, filing a document, communicating with a counterparty, belongs behind human approval, a boundary we examine in our piece on governing agentic workflows.
The register as market data
The compounding effect arrives quietly. One deal's register is operational hygiene. Fifty deals' registers are market data: the survival periods the firm typically accepts by deal size and sector, the escrow percentages that actually get negotiated, the consent thresholds that keep recurring. The next negotiation starts from evidence instead of anecdote, partners quote those numbers in pitches, and clients notice the difference between a firm that remembers its deals and one that re-learns them. The diligence phase of the next deal, where the register is born, sharpens too; we walk through that sequence in the diligence checklist.
Adoption does not require a heroic backfill. Start with live deals at signing, when the executed set is fresh and the pre-closing window makes the value obvious. Backfill selectively: deals with active earnouts, surviving indemnities, or running transition services agreements. That is how Reframe deployments typically begin, one live deal at signing, then expansion once the rhythm proves itself. The archive can wait. The deadlines cannot.
Closing conditions do not slip because lawyers are careless. They slip because obligations outnumber and outlive the attention that created them. Extraction with provenance, a register that is a view of the graph, owners, alerts, and an audit trail: that is the structure that holds when attention moves on, and it is the difference between remembering a deal and being able to prove what it still requires.