There is now a measurable probability that your collection letter was read by an AI before it was read by your debtor — pasted into a chat window with the question "how do I respond to this?" or, more often, "do I have to?"
That is not speculation. Anthropic — the company behind the Claude AI — publishes an Economic Index mapping what people actually use AI for, matched against the task catalogs of 718 occupations. In the May 2026 release, tasks characteristic of credit counselors — advising people on debts, repayment plans, and difficult money conversations — rank 14th out of 718 by usage share (anthropic.com/economic-index). Ahead of nearly every other business function. People find debt conversations stressful enough that they rehearse them with machines before having them with humans.
Anyone who has ever drafted a difficult email four times already understood this. The machine is simply the first counterparty in history that doesn't judge, doesn't remember, and doesn't call back. For a debtor carrying shame, fear, and an unopened inbox, that combination is irresistible.
What the machine tells your debtor
Here is the part worth a CFO's attention: the advice AI gives debtors is fairly predictable, and mostly good. Asked about an unpaid B2B invoice, the major models converge on the same counsel: don't ignore it, respond in writing, acknowledge the debt or dispute it specifically, propose a payment plan, get any agreement documented, and understand that silence makes everything worse.
Read that list again. It is your wish list. The machines are, at scale and for free, telling debtors to do exactly what you want them to do. The debtors who follow the advice re-engage. The problem was never the advice — it's that a debtor deep in avoidance may ask the question and still not send the email, because knowing what to do and being able to do it are different neurological events.
Writing for two readers
The practical consequence: every collection letter now has two potential readers — a stressed human, and the calm machine that human may consult. They reward different things, and for once, the same letter can serve both.
The human reader is scanning for threat. Walls of text, legal boilerplate, and bold red deadlines confirm the danger and close the tab. Short, calm, specific, with one small concrete next step — that gets read.
The machine reader is summarizing. When your debtor pastes your letter and asks "what does this mean and what should I do," the model extracts the facts: amount, date, what you're offering, what happens next. A letter with a clear amount, a clear deadline, a genuinely reasonable offer, and a plain statement of the next step gets summarized accurately — and an accurate summary of a fair letter is a summary that recommends engaging with you. A letter that is all pressure and no path summarizes as "they're threatening you; you may want legal advice." You wrote your own opposition briefing.
The letters that fail both readers are the ones most creditors send: long, defensive, escalating in bold. The letters that pass both are short, factual, fair, and specific. This was always true. The machine reader just made it checkable.
The part nobody says out loud
If people prefer rehearsing hard conversations with machines, some fraction of your debtors would rather resolve the debt through one too — no shame, no voice on the phone that remembers them, no performance of apology. The collection industry has spent decades making contact easier for creditors. The next advantage belongs to whoever makes contact easier for debtors. Watch this space.
Until then, the arithmetic stands: your silent debtor has likely already been told, by the most patient advisor on earth, to answer you. What they need is a letter safe enough to answer — and, past a certain point, a third party for whom answering carries no shame at all.
If your letters are being read by everything except a payment, send us the file for a free review. No win, no fee.