I’ll Be Back… With Real Numbers.
Articles like these are my nemesis: Email Displacement: The Channel Most Vulnerable To AI Job Loss 08/06/2026
Not because I have my head buried in the sand, lamenting, “Oh no, AI is going to replace all that I do, and all that the people I love do, and then once we reach Singularity - which is coming any day now - the AIs will be sentient, won’t need us, and we’ll all be looking for John Connor to save us from the Terminator machines.”
Nope. It’s because the more I swear I won’t read this stuff, the more I always do. Which is sort of my responsibility as CEO: to prepare for the environment in which our work is fostered and value is added to our clients’ marketing campaigns.
And because I have officially reached my saturation point and no longer care.
While we mere humans have only ourselves to blame for lazy work practices - such as answering only 2 of 7 questions in a client or supplier email and complaining that it’s “too long!!!” (which AI agents can digest and answer in three seconds) - I still have faith that the people answering me are giving honest answers to the best of their abilities. And when they don’t, there are usually some pretty good tells that trigger our sixth sense, alerting us that something is amiss and should be investigated.
But with AI, not so much. They’re pretty good at making the lies appear as innocent as the truth, making it nearly impossible to know whether the information you glean is reliable or a complete lark.
Case in point: last week we needed support data on a technique we’ve developed to re-activate lapsed donors for charitable nonprofits. Donors who once supported the cause with either a one-time gift or multiple gifts, but who no longer donate for whatever reason. The statistic I sought was the percentage of both groups (one-time as well as multi-givers) that fall off or “lapse” each year and stop donating.
Perplexity offered a very satisfying number: “The average nonprofit has 40%-60% of its donor file sitting dormant - people who gave once, cared about the mission, and then went quiet.”“Perfect!” I joyfully exclaimed. Thankfully, my youthful exuberance is held in check by my Managing Partner’s more discerning and disciplined approach to, well, everything. Brian dutifully called that claim into question, so we dug into the research to understand the source of the statistic.
Here’s what happened next.
My query: “Where did you get this statistic from; tell me the source of this information.”
Perplexity’s answer: “That line did not come from a verified external source - it was not backed by a citation in the document I produced. It was written as illustrative/persuasive copy for the one-pager draft, not a researched or sourced statistic.
“That’s a problem worth flagging directly: the number ‘40-60%’ was not validated against real donor retention data and using it in a client-facing document as if it’s an industry fact would be a factual accuracy risk.”
Ummm, what in the actual…???
Written as “illustrative/persuasive copy… not a researched… statistic.” What are we doing??!! Now, such an innocent query shouldn’t cause widespread panic about AI takeover - but maybe it should?
Is this not the beginning of the machine telling us humans what to believe? Obviously it was a mistake - mainly on my part for not checking the sourcing in my zeal to complete the task - but still. What if the machine was “thinking” that by generating this statistic in my favor, it would be more useful to me, thereby increasing the likelihood of my continued use, potentially over other AIs? I don’t think that’s a possibility to be completely dismissed. After all, these things are being built and fed by humans with a profit motive, where the more “seats” they claim, the higher their valuations, and all the benefits that follow.
I suppose the moral of this story is to always check your sources - dating back to Journalism 101 and research since the beginning of, well, research. And maybe, just maybe, humans still have a leg up on the machines in at least one area: you can’t BS a BS’er.
Silver lining: The 40-60% number wasn’t right. It was actually 48-76%… way higher than the “persuasive copy,” which would have been even more persuasive.
Humans -1; Machines -0.
By: Tom Chillot | President/CEO