Document review is where litigation budgets go to die.
According to research by the RAND Institute for Civil Justice, review accounts for 73% of the total cost of producing documents during e-discovery[1]. The American Bar Association puts the figure even higher, estimating that document review alone can represent more than 80% of total litigation costs[2]. In complex commercial disputes, that can mean millions of pounds spent before a single argument is heard.
The bottleneck is not hard to understand. Legal teams must comb through emails, contracts, board minutes, internal reports, and digital communications to determine what is relevant, what is privileged, and what must be disclosed. An experienced reviewer typically processes around 50 documents per hour – roughly 36 seconds per decision. Scale that to a case involving hundreds of thousands of records and you begin to see why discovery timelines stretch into months.
Artificial intelligence is changing the arithmetic.
The problem is not just volume – it is fragmentation
Modern organisations generate data across dozens of systems: Email servers, collaboration platforms like Teams and Slack, cloud storage, legacy document management tools, and shared drives. When a dispute arises, all of this becomes potentially disclosable.
The difficulty is that relevant evidence rarely sits in one place. A contract amendment might only become significant when read alongside three months of email negotiation and an internal risk assessment stored on a different platform entirely. Traditional linear review – reading documents one at a time – is poorly suited to spotting these cross-system connections.
Worse, some records may exist only as scanned images or paper files, invisible to keyword searches unless they have been digitised with optical character recognition (OCR) and properly indexed.
How AI changes the discovery process
AI-powered review tools do not simply speed up the same manual process. They change the method.
Machine learning models trained on relevance decisions can analyse entire document collections and rank records by their likelihood of being responsive to a particular request. Rather than starting at the top of an alphabetical list, reviewers focus their expertise on the documents most likely to matter.
Beyond ranking, AI can cluster related documents together, flag unusual communication patterns, and identify conceptual similarities that keyword searches miss. A term like “the arrangement” in an email thread might never appear in a keyword list, but a trained model can recognise it as contextually relevant to a price-fixing investigation based on surrounding language.
The effect is compounding. Earlier identification of key evidence means legal teams can refine their case strategy sooner, assess settlement positions with better information, and avoid the cost of reviewing large volumes of irrelevant material.
The financial case is straightforward
If document review represents 73% of discovery costs, even a modest reduction in review time produces significant savings. But the financial impact extends beyond direct cost reduction.
Earlier insight into critical documents allows organisations to make faster decisions about whether to pursue, settle, or defend a claim. In time-sensitive disputes – regulatory investigations, for example – speed can be the difference between a cooperative early resolution and a protracted enforcement action.
AI-assisted review also reduces the risk of human error in high-volume exercises. Fatigue, inconsistency between reviewers, and the simple probability of missing a needle in a haystack of 200,000 documents are well-documented problems with manual review. Machine learning models, by contrast, apply the same criteria consistently across every record.
Why none of this works without structured data
There is a catch. AI-powered legal analytics depend entirely on the quality and accessibility of the underlying data.
If an organisation’s records are trapped in paper archives, scattered across disconnected systems, or stored as unindexed image files, the most sophisticated AI tool in the world cannot analyse them. The model needs machine-readable, searchable text with consistent metadata to deliver meaningful results.
This is the gap that many organisations underestimate. The technology is ready. The data often is not.
Preparing for AI-driven legal discovery means digitising paper records to a searchable standard, applying structured indexing and metadata through effective data capture, and creating unified information environments where documents from multiple sources can be analysed together.
Where Dajon fits
Dajon Data Management specialises in exactly this preparation work. Through secure document digitisation, intelligent indexing, and structured data preparation, Dajon transforms fragmented paper and digital archives into environments that AI tools can actually use.
For legal teams and the organisations that support them, this data foundation is what separates a theoretical AI advantage from a practical one. Faster evidence discovery, lower review costs, and stronger case preparation all start with getting the underlying records into the right shape.
The technology to revolutionise litigation document review already exists. The question for most organisations is whether their data is ready to meet it.
References