6,200 deals reveal where transaction work really happens

Where deal work actually concentrates

Legatics analyzed 6,200 completed transactions across law firms in the UK, Europe, US, and Asia and found a consistent pattern: the bulk of transactional work occurs in the weeks and months between engagement and signing, not at closing. Establishing workstreams, tracking conditions precedent, coordinating across counsel teams, and managing document states dominate pre-execution time. Closing day is the moment all of that work becomes visible, not the moment most of it happens.

Yet the operational attention at most firms remains concentrated at execution. Pre-execution work sits in email, spreadsheets, and personal trackers maintained by individual associates. It is invisible to partners not directly on the deal and almost always invisible to clients. This mismatch—work concentrated pre-close, attention concentrated at close—is where time, risk, and partner anxiety leak out.

The data resolved into a five-tier maturity model. Foundation-tier firms use transaction management only for high-visibility stages; pre-execution work remains in email and spreadsheets. Developing-tier firms have introduced task lists and basic document management with some centralization. Scaling-tier firms add live client access so deals become visible without asking. Established-tier firms structure every matter the same way from day one with centralized obligations; closing becomes the natural conclusion to a process that was always running. Advanced-tier firms capture and reuse knowledge across deals, automate client reporting, and accelerate new associate productivity from their first matter (company-reported).

Most firms cluster around developing or scaling. The gap in operational performance between the bottom two tiers and the top two is material enough to warrant senior partner attention.

AI cannot fix what is not structured

This data matters now because AI is entering transaction workflows at scale. The real bottleneck for legal AI is not model quality—it is what the model can see and what the model can do. A language model can summarize a document or draft a clause. It cannot reach into a process that isn't structured anywhere it can see, and it cannot act in the systems where the work actually happens.

Emerging standards like the Model Context Protocol are addressing the connectivity half of the problem. Structure is the harder half, and it is what the maturity data exposes. If the underlying transaction process isn't legible, connectivity alone doesn't make AI useful. AI applied to deals running across email, spreadsheets, and individual trackers can only do so much because there is nothing structured for it to operate on. Firms that have built structured pre-execution discipline will find that AI accelerates work that is already well-organized. Firms that haven't will find that better models do not fix an unstructured deal.

Start by measuring where you actually are

The useful question for most firms is not where they want to be in three years. It is where they actually sit now, what the deals running through the firm right now would tell an outside observer, and which one tier upward looks like a concrete six-month programme rather than a transformation pitch.

Before adopting transaction AI, audit your pre-execution infrastructure. If status requests from clients pull associates off other work because nobody can see deal progress without rebuilding it by hand, you are at foundation or developing tier. If partners discover problems late because early stages run in private inboxes, you have an AI problem downstream of a process problem. Fix the process first. The work to move up one tier is real but not heroic, and it is prerequisite work for any AI play to land.