The Legal AI 'Dead Letter' Problem: Why AI-Drafted Demand Letters Are Getting Ignored by the Other Side — and What That Signals About Trust
There's a phenomenon spreading through claims departments, defense firm inboxes, and commercial collections desks that nobody has formally named yet, but every practitioner over forty is starting to recognize. AI-drafted demand letters are getting triaged into a de facto low-priority queue — not because of...
There's a phenomenon spreading through claims departments, defense firm inboxes, and commercial collections desks that nobody has formally named yet, but every practitioner over forty is starting to recognize. AI-drafted demand letters are getting triaged into a de facto low-priority queue — not because of their legal content, but because of what recipients can feel is missing from them. Call it the dead letter problem: correspondence that arrives, gets processed, and gets functionally ignored.
This isn't about AI skepticism or technophobia. It's about signal degradation. And it has real consequences for litigation strategy.
How Recipients Are Developing Informal Detection Heuristics
Claims adjusters at major carriers — and I've spoken with several over the past six months — have developed remarkably consistent informal filters for identifying AI-generated demand letters in personal injury, employment, and property damage matters. The tells are specific: an absence of granular chronological detail, liability arguments that mirror form language rather than the actual fact pattern, damages narratives that describe injuries in clinical categories rather than in the particular texture of a real person's suffering.
One senior adjuster at a regional carrier put it plainly: "When I see a letter that talks about 'significant and ongoing pain and suffering' but can't tell me whether the client still can't pick up her grandchildren or had to give up competitive cycling, I'm not opening a serious reserve. The AI did the word count. Nobody did the work."
In commercial collections, the pattern is even starker. Collection defense attorneys are now routinely flagging demand packages where the underlying account documentation appears carefully assembled but the demand letter reads like it was generated against a generic template — because it was. In New York consumer debt matters following the Midland Funding v. Madden fallout, defense counsel have noted that AI-drafted demands often fail to engage with jurisdiction-specific statutory particulars under the FDCPA in ways that suggest no attorney actually read the file. That's a credibility problem before the first response is due.
The Employment Context Is Particularly Acute
In employment discrimination and wrongful termination matters, the degradation of demand letter credibility may be doing the most damage to the plaintiff's bar. Pre-litigation demand letters in employment cases have historically done significant work — they've forced early evaluations, prompted mediated settlements, and occasionally produced document holds that prove useful later. That utility depended entirely on the recipient believing that a real lawyer with a real theory of the case was on the other end.
AI-generated demands in employment matters tend to recite the McDonnell Douglas framework, list the protected characteristics, and describe damages in round numbers. They feel, to HR counsel and employment defense partners alike, like demand letters assembled from the constituent parts of a demand letter rather than from the constituent parts of a case. The result: lower initial reserves, slower response timelines, and — critically — reduced willingness to engage in pre-litigation settlement discussions at all.
The employment plaintiff's bar built leverage on the credibility of early advocacy. That leverage is eroding.
What This Signals About Trust and Epistemic Weight
The deeper issue here isn't technical. It's epistemic. A demand letter functions as a credibility instrument. It signals that someone with professional judgment and skin in the game has evaluated this matter and arrived at a considered position. The signature of a licensed attorney on that document has historically carried weight not just as a professional certification but as a representation that a human mind engaged with this particular set of facts.
When opposing counsel or an adjuster suspects — correctly or not — that no such engagement occurred, that weight disappears. The letter becomes noise rather than signal. And noise gets filtered.
This matters for litigation strategy in a specific and underappreciated way: early-stage communications shape the settlement trajectory of cases. A demand letter that establishes credibility, specificity, and evident case preparation creates anchoring effects that persist through mediation. A demand letter that reads like it was generated in forty-five seconds creates an opposite anchoring effect — one that signals the sender may not be prepared to litigate seriously, and that lowball offers are worth testing.
How Firms Are Adapting
The more sophisticated plaintiff and claimant-side practices are already responding. The emerging workflow isn't "don't use AI" — it's "use AI for structure and research, but insist on human-generated specificity as the final layer." That means the attorney or a senior paralegal is responsible for injecting the granular factual detail that no AI can generate: the specific conversation on the specific date, the exact phrase the supervisor used, the medical notation that connects the mechanism of injury to the complained-of limitation.
Insurance defense firms are adapting differently. Several are now running incoming demand letters through informal review protocols specifically designed to assess the level of case preparation the letter suggests — essentially trying to gauge how seriously they need to take a matter based on whether the demand reads like someone actually prepared for litigation.
At least two AmLaw 200 firms have reportedly updated their litigation intake standards to flag matters where the plaintiff's pre-litigation correspondence suggests minimal engagement with the file, treating this as a variable in early case valuation.
The Bottom Line
The legal profession spent decades building the epistemic authority of written advocacy through demonstrated judgment, case-specific analysis, and professional accountability. AI tools can accelerate drafting, but they cannot substitute for the evidentiary texture of a lawyer who actually knows the file. Practitioners who are using AI to avoid doing the work — rather than using it to do more work better — are not saving time. They are burning credibility they cannot easily rebuild.
The dead letter problem is, at its core, a trust problem. And in law, trust is the product.