I recently had the pleasure of working with our own Trustpoint.One AI-driven review team on a high-stakes project for a Fortune 100 company facing a tight discovery deadline. What started as a daunting 200k document universe became a masterclass in how the right team structure can transform AI review from experimental to essential.
The Challenge:
Our client needed rapid turnaround on a complex commercial litigation matter. We began with 200k documents that our initial review brought down to 100k for AI analysis using Relativity aiR. The pressure was on to produce a defensible final set quickly and accurately.
The Trustpoint.One Difference
Here’s where our approach diverged from typical “let’s try AI” implementations:
Strategic Curation First
Before large-scale prompting began, Trustpoint.One applied a structured curation process, validating document quality, removing clearly irrelevant documents and file types, and segmenting data by evidentiary value. This upfront work reduced noise and ensured aiR analyzed quality evidence providing both known and unknown information.
Iterative Prompt Development
Moving beyond static prompts, our AI experts and review team took the lead, collaborating with counsel to clarify responsive criteria, address gray areas, and serve as a force multiplier throughout the Gen‑AI workflow. As opposed to trying to train law firm attorneys on the fly to be prompt engineers, it is much more effective to allow them to focus on what they do best, legal strategy, not wrangling technology. Final legal decisions remain in the hands of lead counsel; AI and review team only propose. Review managers set and validate parameters, sample results, and get sign off from lead attorneys – just like they do for manual review. That workflow consisted of:
• Translate evolving legal theories into precise, fact-grounded prompt language
• Test prompts against curated document subsets for accuracy and relevance
• Rapidly refining structure and terminology based on observed outputs
• Obtain sign-off from lead attorneys before beginning validation
• Validate results before scaling across full production
This curate-test-refine-validate cycle allowed us to converge quickly on prompts that surfaced the right evidence, not just more evidence.
The Results
Final production set: just under 20k documents
Privilege log: 3,200 entries
Timeline: Days instead of weeks
Outcome: Defensible, high-confidence results
Why This Worked
Because Trustpoint.One brought both legal-domain expertise and technical fluency in aiR, the firm avoided common pitfalls of over-prompting, misaligned issue framing, and unreliable outputs. Instead of spending weeks troubleshooting prompt logic, the legal team reached defensible answers in days—freeing firm attorneys to focus on analysis and advocacy rather than tool calibration.
The lesson? AI review succeeds when expert methodology meets advanced technology. It’s not just about having the tools—it’s about having a team that knows how to use them strategically.
Have you seen similar results when the right expertise meets AI capabilities?
Frequently Asked Questions
How can AI review help legal teams handle large document populations?
AI review can help legal teams analyze large document populations more efficiently when it is integrated into a structured workflow. In this case, Trustpoint.One reduced an initial 200,000-document universe to approximately 100,000 documents for Relativity aiR analysis before reaching a final production set of just under 20,000 documents.
Why is strategic data curation important before using AI for document review?
Strategic curation helps improve the quality of the data being analyzed. Trustpoint.One validated document quality, removed clearly irrelevant documents and file types, and segmented the data by evidentiary value before applying large-scale prompting.
How does Trustpoint.One approach prompt development for AI review?
Trustpoint.One uses an iterative curate-test-refine-validate process. AI experts and review professionals work with counsel to translate legal theories into precise prompts, test them against curated subsets, refine terminology and structure based on results, and obtain attorney sign-off before validation and broader application.
Does AI replace attorneys’ legal judgment during document review?
No. In the workflow described, lead counsel retains responsibility for final legal decisions. AI and the review team propose results, while review managers establish and validate parameters, sample outputs, and obtain attorney approval before scaling the workflow.
What were the results of this Trustpoint.One AI review project?
The project produced a final set of just under 20,000 documents, included 3,200 privilege-log entries, and was completed in days rather than weeks. The article describes the outcome as defensible, high-confidence results.