From Unusable to Usable in 3 Hours: How Page One and Becker & Poliakoff Turned a Flawed Production into Deep Insight with Custom Analyses

How did they do it?

• Turned an incomplete third-party production of roughly 18,000 documents into review-ready data using custom analyses in aiR Review
• Split the population into emails and loose documents, then wrote purpose-built prompts to extract the fields and summaries each group needed
• Tested and refined every prompt on a subset of documents before deploying across the full set, completing the run in about three hours from start to finish

Back to the Basics

When Becker & Poliakoff received what should have been a standard third-party production, they faced a surprisingly common conundrum: the documents were all there, but the basic structure around them was not. The production lacked the metadata and load-file information a legal team expects to work from – who sent an email, who received it, when, and so on. Even though that information existed inside the documents, it wasn't in a form anyone could filter, sort, or review efficiently.

The team had a few options:

1. Go back to the third party and request a corrected response, which would delay case management;
2. Assemble a team and manually code, compiling the missing information. This would have involved standing up a support staff and countless hours of document review and objective coding; or
3. Outsource the effort through a traditional, labor-intensive manual review, ultimately adding further cost and time.

None of the options were particularly appealing. All involved significant delay and/or additional cost.

What Becker needed was an efficient to make the response usable without sacrificing confidence in the results. Together with their partner Page One, they opted for a different route – a custom AI workflow with Relativity aiR.

Custom Analyses for a Common Challenge

A no-code, natural-language AI feature, custom analyses lets legal teams create matter-specific extractions, classifications, and questions beyond standard responsiveness or privilege checks. That flexibility is exactly what Becker needed.

Page One and Becker got to work right away, designing a workflow to extract key metadata and create useful document-level summaries directly from the production itself. To start, they divided the approximately 18,000-document population into two groups:

• Emails: Email From, Email Recipients, Email Subject, Email Date/Time Sent, and a summary.
• Attachments and loose documents: Names, first document date, document title, document type, and a summary.

Becker’s legal team defined the information they needed from the response, and Page One turned those requirements into carefully crafted prompts for each group. The result recreated much of the practical value of the missing metadata and load-file fields, while adding narrative insight drawn from the content of the documents.

"Page One worked closely with our team to understand the challenges with the third-party response and developed a practical workflow using aiR Review," says Tracey McAnally, director of litigation and e-discovery at Becker & Poliakoff. "By testing the approach together and then applying it consistently across the document population, we were able to get useful, structured information quickly and efficiently, and we were very impressed with the results."

The Validation: Test First, Then Scale

The team knew custom analyses would help them go fast, but in legal, speed is rarely the primary concern. The question was: could they trust the output?

To ease their client’s hesitation, Page One created a careful and controlled workflow where they separated the document population by document type, wrote tailored prompts for each group, and then tested those prompts against a subset of documents. The Page One team then reviewed the results and refined the prompts with Becker’s team before expanding the workflow to the full population.

That sequence of test, refine, scale gave the legal team a chance to check the work before expanding on it – and it meant that every document in each group was analyzed against the same tested criteria.

The approach also offered something manual coding couldn’t: machine-level consistency. Manual coding allows for discretion, individual judgment calls, and, sometimes, human error. aiR Review, on the other hand, reads the extracted text of each document and applies the same instructions every time, which made the output consistent and defensible.

"This was a true collaboration between Becker and Page One, with both teams working together to identify the information we needed, test the approach, and refine it before applying it across the production," says Tracey. "The speed, efficiency, and consistency of the results exceeded our expectations and demonstrated how generative AI can solve a very practical e-discovery challenge."

A Usable Data Set in Hours, Not Weeks

Once the workflow was designed and validated, Page One processed the full population of approximately 18,000 documents in about three hours from start to finish. A flawed, nearly unusable production became organized metadata and summaries in roughly three hours.

aiR Review not only delivered speed, but it also gave Becker’s legal team the consistency and validation they needed to have confidence in the result. Additionally, treating emails and loose documents differently meant each group yielded the fields most useful to attorneys, effectively doubling the types of insight captured from the same production. And with key fields in view (sender, recipients, dates, subjects), the team was set up for next steps like assessing case timelines and analyzing communications.

"Working together, Becker and Page One were able to take a production that presented significant challenges and develop a customized approach using aiR Review that addressed exactly what our legal team needed," says Tracey. "Seeing approximately 18,000 documents handled consistently through a tested workflow in just a few hours was an impressive example of what thoughtful collaboration and the right technology can accomplish."

Turning a Workaround into a Future Workflow

For Page One, the project expanded what the team can offer when a production or response arrives incomplete. A job that might have meant a delay for re-production, a separate outsourcing engagement, or substantial manual effort can now be handled with a targeted custom analyses workflow, and priced and scheduled accordingly.

Page One sees the potential to take on similar projects faster and to design new, matter-specific analyses around the information attorneys actually need, rather than being limited to predefined fields or traditional processing.

"Custom analyses in aiR Review changed how we approach incomplete productions. Instead of asking for a re-production or throwing people at manual coding, we can build an analysis around exactly what the attorneys need. We’re excited to bring that to more matters."
Andrew Milauskas, Chief Operating Officer, Page One