If you could click once and let Al do your disclosure ... would you? Should you?
The idea of the "perfect review" is closer than ever - but what are we risking to get there?
General enquiries: 0800 064 0204
eDiscovery: 020 7242 9601
Reprographics: 020 7405 9178
57 Carey Street, London, WC2A 2JB
Email: info@legastat.co.uk
2 min read
Jamie Gunn : Jul 2, 2025 11:05:24 AM
The idea of the "perfect review" is closer than ever - but what are we risking to get there?
“The perfect review.” It’s a phrase we’ve all used — sometimes lightly, sometimes with ambition, but rarely with genuine expectation. For years, it was more an industry ideal than a real target: something to mention on a panel or joke about while reviewing hundreds of thousands of documents. Everyone understood that true perfection wasn’t achievable. The sheer volume of data, tight deadlines, and cost pressures meant it simply wasn’t viable. So, we aimed for something else: Reasonable. Proportionate. Defensible. Something we could justify if asked. Something a court, regulator, or client would understand. A process that made sense, even if it wasn’t flawless.
But that’s starting to change.
With the introduction of generative AI and large language models into legal workflows, the notion of a perfect review is no longer just rhetorical. It’s creeping into view — quietly but quickly. And for the first time, we can ask the question seriously: “What would a perfect review actually look like?”
Is it a Fully Automated Review? AI reviews, tags, and then discloses the relevant documents without a single human ever looking at them. It’s fast, efficient, and arguably more consistent (if the tech holds up) than a team of humans.
Or is it a Relevance-Only Review? AI flags documents as relevant, and you review only those? No time spent reading the irrelevant material, just targeted, strategic review.
Or is it a more cautious Hybrid Validated Review? AI tags everything, and the legal team reviews all relevant documents. Then, the non-relevant documents are sampled to validate accuracy and ensure the process is sound.
There’s no single correct answer, but the fact that these are real options now should make us pause. Because striving for “perfect” inevitably brings us to a more critical question:
Would you ever disclose a document you haven’t looked at?
Not hypothetically. Not in a few years. Right now. Because that’s what some AI-powered review tools are already capable of, and in some cases, already doing.
But here’s the issue. Disclosure isn’t just about relevance and privilege. It’s about strategy. It’s about understanding the material, not merely filtering and categorising it. When you review the documents, you’re not just checking boxes. You’re learning your case. You’re testing timelines. You’re spotting contradictions. You’re thinking tactically.
What happens when we remove that layer of professional insight? What happens when we rely entirely on AI for the grunt work and judgment?
It’s not just a risk. It’s a shift in what disclosure even means. There’s a cultural piece here that technology can’t replicate. Solicitors, particularly in litigation, have always taken pride in knowing the documents. Not all of them, not always line-by-line, but enough to understand the shape of the evidence. Enough to give confident, tactical advice. Enough to stand in front of a judge and answer questions without flinching. AI can help us get there faster. But it shouldn’t bypass that entirely.
The question we now face isn’t whether we can reach perfection. It’s what kind of perfection we’re aiming for, and what we’re willing to give up to get there.
If you’d like to discuss how technology can support your review, or want to learn more about our managed review services, please contact us.
The idea of the "perfect review" is closer than ever - but what are we risking to get there?
For years, the battleground has looked familiar.
Reality vs hallucination Cambridge dictionary’s word of the year for 2023 is “hallucinate”. You may think that hallucination is something only humans...