The most consequential thing happening in evidence discovery right now isn’t a technology announcement. It’s an asymmetry.
In a growing number of disputes, one side walks into the case with modern, AI-driven discovery capability and the other side walks in with a process that hasn’t really changed in fifteen years. The first team can scope the relevant data, surface the documents that matter, and develop a clear strategic picture in days. The second team is still working through linear review, hoping nothing critical is buried in an archive nobody has properly indexed. Both teams are technically meeting their disclosure obligations. Only one of them is actually understanding their position.
The gap between those two starting points is rarely visible from the outside. Courts don’t publish league tables of how well-prepared each side was at the start of a matter. But it shows up in case outcomes, in settlement dynamics, in the speed at which one side can pivot when new facts emerge, and in the calmness with which counsel walks into a meet-and-confer.
This asymmetry is the real story of digital transformation in evidence discovery. It’s not that the tools have got better – though they have. It’s that the gap between organisations that use them well and those that don’t has become large enough to be strategically decisive.
What “evidence” actually means now
Part of what’s driven this is that the data environment underneath modern litigation has changed in ways the traditional discovery process was never designed for.
A decade ago, evidence largely meant email and documents. Today, it means email and documents plus Teams, Slack, WhatsApp, voice notes, video calls, ephemeral messages, attachments stored in personal cloud accounts, transcripts of meetings nobody knew were being recorded, edits to shared documents tracked at the keystroke level, and the chat logs sitting inside whatever niche collaboration tool a particular team adopted three years ago. More than 90% of records created today exist only in electronic form, and the variety of those forms keeps multiplying[1].
The volume problem is well known. The variety problem is the one that quietly causes more damage. A discovery process built around email and documents will not, on its own, find what it needs to find in a Teams thread, particularly if the relevant messages were edited, deleted, or sent in a side channel nobody thought to preserve. The places where the most consequential conversations happen are increasingly the places that legacy discovery processes are weakest at handling.
Manual review was never designed for any of this. Document review already accounts for more than 80% of total litigation spend[2] – around $42 billion globally per year, according to the American Bar Association. Adding modern collaboration data on top of that, without adding the tools to handle it, doesn’t make discovery harder by some marginal percentage. It makes it qualitatively different.
Where AI is actually moving the needle
AI in this space gets discussed in ways that sometimes conflate two different things: prediction and understanding.
The prediction side – ranking documents by likely relevance, clustering similar content, flagging probable privilege issues – has been around long enough that it’s no longer cutting-edge. It’s table stakes. Any modern eDiscovery platform does it, and the gains are real but bounded. Used well, this kind of capability can reduce reviewer hours by as much as 80%[3].
The understanding side is where things have changed more recently. Generative AI and the models behind it can now do things that older tools couldn’t: summarise document families in plain language, extract timelines from messy correspondence, surface contradictions between related documents, build narratives across thousands of files, and answer specific legal questions of the dataset directly. The global eDiscovery market is projected to grow from around $18.73 billion in 2025 to $46.06 billion by 2034[4], and most of that growth is being driven by AI features that didn’t exist three years ago.
For legal teams, the practical effect is that the early stages of a matter look fundamentally different. Instead of starting in the dark and gradually building a picture as review progresses, teams can now develop a working understanding of the evidence position in days rather than weeks. That has knock-on effects on every subsequent decision: settlement timing, witness selection, deposition strategy, the framing of motions, and the willingness to take a case to trial.
This is where the asymmetry gets sharp. A team with modern AI-assisted discovery isn’t just doing the same job faster. It’s making different decisions, earlier, with better information.
What AI doesn’t change
It’s worth being equally clear about what AI hasn’t replaced.
It hasn’t replaced legal judgement. It can surface a document; it can’t tell you what that document means in the context of your case strategy. It can identify a privilege concern; it can’t make the call on how to handle it. It can flag a contradiction in the timeline; it can’t decide whether to lead with that contradiction at trial or save it for cross-examination. None of the strategic work has been automated. What’s been automated is the volume problem that used to consume the time of the people doing the strategic work.
It also hasn’t changed defensibility standards. A discovery process that uses AI is still subject to the same obligations as one that doesn’t – proportionality, completeness, the duty to preserve, the duty to disclose. AI changes how those standards are met, not what they are. Courts, predictably, have started asking sharper questions about how AI-assisted reviews are validated, audited, and explained, and the answers matter.
What’s emerging is a working model where AI handles the volume and the human handles the meaning. That’s a productive division of labour, but only if both halves are actually working.
The foundation problem
Here’s the part that gets less attention than it deserves.
All of the discovery improvements above – the speed, the AI assistance, the pattern recognition, the early case assessment – depend entirely on the underlying data being in a state where modern tools can process it. And in a great many organisations, it isn’t.
Documents stored as unsearchable image files. Email archives that have been migrated three times and lost metadata each time. Historical contracts in PDFs that are really photographs. Records in legacy systems that haven’t been indexed since the system was retired. Shared drives where nobody quite remembers what’s in them. Each of these represents a blind spot. And if the relevant evidence happens to live in one of those blind spots, no amount of AI sophistication will surface it. The tool is doing exactly what it’s been asked to do – it just can’t see the data that matters.
Worse, partial visibility is often more dangerous than no visibility, because it creates false confidence. A discovery process that confidently surfaces 70% of the relevant evidence is, in some ways, more risky than one that admits up front it’s working with limited data. The 30% that’s missing is the part you’d need to know about, and you don’t know that you don’t know.
Where Dajon comes in
This is the work that tends to be invisible until it’s the only thing that matters.
Dajon Data Management helps organisations build the data foundation that modern evidence discovery actually depends on. That means digitising historical records properly, applying consistent structure to document estates, ensuring archives are searchable and indexed, bringing legacy systems into scope, and making sure that when litigation arrives – or when a regulator comes asking – the relevant data is accessible rather than hypothetically retrievable. We work most often with legal teams and the regulated organisations that support them, where the consequences of a blind spot are particularly serious.
It isn’t glamorous work, and it rarely features in the marketing for AI-driven discovery platforms. But it’s the work that determines whether those platforms deliver on their promise or just process whatever happens to be reachable.
Closing the gap
The asymmetry is widening. The organisations that have invested in their data foundation – the ones whose archives are structured, searchable, and complete – can use modern discovery tools the way they’re designed to be used. The organisations that haven’t are still doing discovery the old way, even if they’ve licensed the same software.
In high-stakes litigation, that gap is rarely the deciding factor on its own. But it shows up in everything else. In how quickly counsel can advise on settlement. In how confidently the team walks into a deposition. In whether the strategy is shaped by the evidence or shaped by what the evidence happens to surface in time.
Digital transformation hasn’t just made evidence discovery faster. It’s changed what discovery is – from a process of locating documents to a process of generating legal insight. The organisations that recognise that, and prepare for it before they need to, are the ones whose discovery process becomes a strategic asset rather than a quietly accumulating liability.
Is your evidence discovery process built for the data environment you actually operate in – or the one you operated in five years ago?
Dajon Data Management helps legal teams and organisations build the data foundation that modern evidence discovery depends on. Get in touch to find out where your current setup might be holding you back.
References