How Often Clients Forget a Data Source at Intake

The sources a client forgets often sit within arm's reach of the intake notes.

A client's first account of where their data lives is rarely the whole account. Medicine has known this for a long time. In a 2024 thread among emergency physicians and nurses, one commenter observed that patients tell "at least 3 different stories" during a single visit, to the triage nurse, the bedside nurse and the physician, and that these "usually are all different." Another put it more plainly: "you will not get the whole story."

Legal intake, to my mind, behaves the same way. A client answers the questions put to them, in the order they are put, from memory. Memory is selective.

The gaps tend to settle where work was improvised. When remote work arrived abruptly in March 2020, firms built home arrangements that put business continuity ahead of security, and ABA Formal Opinion 498 later cautioned against free or consumer-grade products. By late 2021 the temporary had become the new normal. A home computer shared with family, a personal phone: neither feels like company data to the person who used it.

The honest starting point is a gap of its own. Statisticians, at least, have a name for an answer that is only partly complete. The U.S. Census Bureau, which studies missing data for a living, counts partial response as a distinct form of missing data: people who "give only part of the information they are being asked to provide."

Which data sources do clients forget to mention at intake?

No published study measures how often a client's source list is incomplete. The recurring omissions are known, though: personal email, vendor-run chat, phones, portable media and AI prompts.

Before I would accept a client's list as complete, I would put four plain questions to it:

1. Did you ever forward work documents to a personal email or cloud account?
2. Did anyone outside the company, such as a vendor, set up a workspace or chat channel for this project?
3. Which phones, tablets or portable drives touched the matter, including ones no longer in use?
4. Did you use an AI tool to draft anything, and do you still have what you typed into it?

The closest figure the market offers is a reported challenge, not a measured gap. In one vendor's 2026 survey of 101 legal professionals, 31% named working with clients to collect their own data as a firm-level challenge, the third most common answer. That figure does not show how many source lists were wrong. Survey statisticians hold themselves to a stricter bar: OMB's 2006 standards set an 80 percent response-rate threshold, below which a survey needed a detailed review of its errors, and I know of no equivalent test for an intake list.

The common assumption is that a client who knows the business knows where its data lives. The examples in that same vendor's guidance suggest otherwise: a custodian ticks the box for company email and never mentions the personal Gmail account used to forward documents, or the Slack workspace a vendor set up for the project.

The pattern is older than the cloud. Writing in the Florida Bar Journal in February 2002, Stephen M. Cohen listed PDAs, memory disks, telephone logs and cellular phones among the sources commonly overlooked, and observed that "while some types and sources of electronic data are commonly thought of, others are not." The devices have changed. The blind spot has kept its shape.

The newest entry is quieter still. In a September 2025 Florida Bar News report, Indiana University law professor Josh Kubicki warned that "most clients aren't disclosing when they've used AI," and that even those who do "often forget the more important part: what they told the AI to get the draft."

Volume has not made any of this easier. In 2024, 93% of 30 surveyed litigation support directors reported that the data they manage per dispute was increasing, and one respondent said sources and types worried them more than volume did.

The Census Bureau's missing-data program offers a test a lawyer can borrow: an answer that is logically inconsistent is "in effect missing," so when the client's account of the dispute mentions a channel the source list omits, the list is incomplete. The forgotten source sits just outside what the client thinks of as work. The omission is rarely deliberate. It is a matter of where attention falls.

Part of the hesitation is about where the data will go once it leaves the client's hands. That is one reason Relevant e-Discovery deploys inside the firm's own AWS account, under the firm's own keys, a posture few small-firm tools can match. Clients who ask whether a review platform learns from their files deserve a direct answer, and the question of whether your e-discovery AI trains on case data is worth settling before the custodian interview begins.

A source that surfaces only after the hold should have reached it is no longer an intake question. It becomes a preservation question, argued before a different audience.

When does a forgotten data source become a spoliation argument?

A forgotten source becomes a spoliation argument once the duty to preserve has attached and the data is lost. A documented chain of custody then shapes how that argument unfolds.

Spoliation, in plain terms, is the loss or alteration of evidence a party had a duty to keep. The duty does not wait for the client's memory to catch up.

One e-discovery provider's consultant, speaking on a 2014 podcast, traced that duty to both the Federal Rules of Civil Procedure and the common law, and made a point that has aged well: once data is destroyed, it may be lost to both parties. Preservation offers no second attempt.

The same speaker described the rules amendments approved in September 2014, which reserved the most extreme sanctions for willful misconduct and intent to deprive the other side of evidence. That narrowing was real. Contrary to a hopeful reading of it, an honest omission is not a free one. A party that fails to preserve may still have to recreate data from backups or other sources, slowly and at considerable expense, while questions gather quietly about counsel's effectiveness and the client's credibility.

Courts have long asked for more than good faith. In Gates Rubber Co. v. Bando Chemical Industries, a 1996 federal decision, the court held that a party has "a duty to utilize the method which would yield the most complete and accurate results." Illinois went further, as Stephen Cohen noted in the Florida Bar Journal, by requiring counsel to review the client's computers and attach an affidavit that disclosure is complete.

Completeness is the standard. Memory is not a method.

The American Bar Association puts the human side plainly: "a lawyer cannot do their job if a client hasn't shared important details." Its guidance on client communication pictures a familiar moment, in which "the client swears they told you over the phone a few weeks ago" and the lawyer is sure they did not. Its remedy is the record, so that when something is overlooked, counsel can show "it wasn't your negligence that caused it."

The difference between a weak record and a strong one is easy to see side by side:

• Before: "Client confirmed all relevant data is in company email."
• After: "Client interviewed; asked about personal accounts, vendor workspaces, phones and departed users; client identified a personal cloud folder; hold extended to that folder the same day."

The second note does not prove the inventory was perfect. It proves the questions were asked, which is what a court will want to see when a source surfaces late.

That record is what Relevant e-Discovery is built to keep: immutable originals with content hashing, append-only audit trails and a documented chain of custody, so the process holds up if opposing counsel challenges it. Late-found sources also tend to sit in places automated tools pass over, which is why the gaps in one-click cloud collection deserve a careful look before anyone calls a collection complete.

None of this requires an enterprise budget. It requires asking in a particular order, and writing the answers down.

How can a small firm run a source inventory it can afford?

Pair a short questionnaire with custodian interviews, then test the inventory against the real collection early, where a missing source reveals itself before opposing counsel finds it.

A source inventory, here, means a written list of every system, account and device that may hold relevant data, built from questions rather than from the client's first answer. Three steps carry most of the weight:

1. Send a pre-collection questionnaire covering email, file servers, cloud storage, mobile devices, messaging apps and any third-party platforms the matter touched.
2. Follow it with custodian interviews that probe for systems or habits the questionnaire did not anticipate.
3. Load what was collected for early assessment and read it for references to places that never made the list.

The first two steps come from one collection vendor's published guidance, which calls the questionnaire "a starting point, not the finish line" and argues that the pairing leaves a documented record of reasonable, systematic effort. I would add the third, because documents are candid about their surroundings in a way custodians rarely are. A thread that ends with a promise to continue in another app has just named a source.

The money is already moving in this direction. ComplexDiscovery's figures, published in 2025, show collection rising from 8% of eDiscovery task spending in 2012 to 16% in 2024, while review fell from about 73% to 64% and was projected to reach 52% by 2029. The same analysis noted that Slack, Microsoft Teams, mobile content and ephemeral messaging apps had pushed review well beyond email and documents.

An affordable tool still has to keep the defensible record, which is why Relevant pairs Bates numbering, privilege and chain of custody with cited answers and chronologies in one place. That record should begin at intake.

Conventional wisdom holds that careful early assessment belongs to large matters with large budgets. The 2025 ComplexDiscovery analysis argued the opposite, observing that generative AI could bring savings and scale especially to small and mid-sized matters where full technology-assisted review had been cost-prohibitive. A larger collection is also not automatically a slower one, a point worth reading in whether a bigger document set always means a longer review.

On cost, Relevant e-Discovery's position is that the economics should be visible on the firm's own matter: AI-assisted review runs cents per document, against the dollars per document and roughly $19K per gigabyte of manual review, and the proof it offers is a demonstration in which the buyer brings a messy collection and a hard question, then watches the platform read, code and cite their own evidence.

That is where the third step stops being theory. A forgotten workspace, a personal account or a departed user's mailbox tends to surface as a reference inside documents already in hand, provided someone reads them early enough to act.

What no platform can do is make a client remember. That part still happens across a table, one considered question at a time.

What should change at your next client intake?

Plan for a second conversation. A single intake meeting captures what the client remembers that day, and the sources it misses are the ones most likely to surface later.

Professions that live by intake already assume this. In a 2026 discussion among practicing therapists, one clinician noted that about a third of the time the treatment plan takes 2 sessions, though it usually rests on data gathered in the first. The second meeting does not replace the first. It completes it.

The record matters as much as the meeting. I would have the client read the written inventory after that second conversation and confirm it, so the file shows what was asked, what was answered and when.

I would book the follow-up before the first meeting ends, and put the home computer and the personal phone on its agenda by name. A forgotten source found in the second week is an inventory note. Found later, the question becomes whether counsel's steps were reasonable, and the dated note is the answer.