Corporate AI Slop and the Case for Saying Less
Last Tuesday, between clinic notes and a late message from a colleague, I opened an email thread that had already started to feel tired before I reached the end. A thoughtful paragraph I had written about a workflow change in our group had been answered with a polished block of language that sounded calm, strategic, and pleasantly vacant. It had the right nouns. It had almost no meaning.
I stared at it longer than I wanted to. Then I did what doctors do when something in a chart does not fit the story, I read it again, slower. The words looked competent, but they did not behave like a human response. No one had clearly read what I wrote. No one had weighed the tradeoffs. The chain was now longer, and less alive.
The professional courtesy of brevity
I used to think verbose replies were a sign of seriousness. If someone took the time to write five polished paragraphs, I assumed they had engaged with the material. In medicine, that instinct has been trained out of me. The chart can be full and still be empty. The consult note can be elegant and still miss the patient in front of you.
That is why corporate AI slop unsettles me so much. The real problem is not that the message is generated. The deeper problem is that the message performs understanding without paying for it. It creates the appearance of attention while outsourcing the effort of attention itself. In a team environment, that can feel like a breach of professional courtesy.
I have started calling the useful opposite of that behavior earned brevity. It is the short message that comes after thought, not instead of it. The line that says, clearly, “I read this, I disagree for this reason, and I suggest this next step.” No decorative fog. No synthetic warmth. Just enough language to move the work forward.
That preference did not come from a seminar or a productivity guru. It came from years of clinical work, where a hurried sentence can be dangerous and a precise sentence can save time, trust, and sometimes skin. In the exam room, I have watched patients relax when I stop performing and start speaking plainly. In the inbox, I have seen the same thing happen. Clarity lowers the temperature.
What AI exposes about workplace reading
The current wave of corporate AI writing tools reveals a blunt truth about many organizations: a lot of our communication was already performative. The machine did not invent the habit of recycling phrases that sound strategic but do not sharpen a decision. It just made the habit faster, cheaper, and harder to ignore.
That is why the rise of generic AI language feels different in medicine than it does in a marketing meeting. Clinical teams already drown in messages. When an inbox reply arrives that seems assembled from ambient professionalism rather than actual review, it adds a second burden. First, someone has to decode the response. Then someone has to wonder whether the original message was read at all.
That suspicion matters. Trust in a team depends on the sense that attention was paid. Not perfection. Attention. If I send a colleague a nuanced question about risk, and I get a templated paragraph about alignment, the chain is now performing itself. The content may be fluent, but the relationship has gone brittle.
This is where shorter is often better. Shorter can force honesty. A tight reply has to choose. It has to identify the point of friction, the decision needed, or the next action. It leaves less room for fake confidence. It also leaves less room for the recipient to mistake style for thought.
What the evidence suggests, and what it does not
The data around AI and communication are more useful than the online argument. In Ayers et al. in JAMA Internal Medicine (2023), 195 patient questions from a public social media forum were compared with physician and chatbot responses. Licensed health care professionals preferred the chatbot response in 78.6% of comparisons, and the chatbot replies were rated higher for quality and empathy. The point was not that machines have become wiser than clinicians. The point was that length and polish can outperform human haste in some narrow settings.
But another study complicates the celebration. In Garcia et al. in JAMA Network Open (2024), AI-generated draft replies to patient inbox messages were used in only a minority of cases, with utilization around 19.4% in later summaries and about 13.3% in some clinician subgroups. That is the part people skip past. Even when the draft exists, clinicians do not automatically trust it. They review, edit, reject, or ignore it. The machine can draft. The human still owns the judgment.
A third line of evidence comes from the older problem AI keeps wandering into: interface friction. Melnick et al. in Mayo Clinic Proceedings (2020) reported a mean System Usability Scale score of 45.9 for EHR usability, a failing score, and found that each 1-point increase in usability was associated with about 3% lower odds of physician burnout. That is not a sidebar. It is a warning. Clumsy systems create the conditions for slop, because people stop editing carefully when the work is already exhausting.
One more number sits in the background of this conversation. In a workflow analysis summarized in later reporting, AI draft use was associated with a modest reduction in message turnaround time, roughly 6.76 percent. Small gains matter in a clinic inbox. They also make the wrong thing look efficient. A faster bad reply is still a bad reply.
So the evidence points in two directions at once. AI can help with speed and scaffolding. It can also reward the least thoughtful habits in a workplace already too comfortable with empty fluency. Both can be true. That tension is the story.
The fraud of sounding busy
What I would not do is ask a team to send longer messages because they feel more professional. I would not treat verbosity as proof of seriousness. I would not accept a reply that merely recycles the vocabulary of alignment, bandwidth, and stakeholder value when the actual question is whether the plan can be executed on Tuesday.
In practice, corporate AI slop often arrives dressed as courtesy. That is part of why it works. It is polite enough to delay confrontation, vague enough to avoid accountability, and fluent enough to discourage a second read. In medicine, that combination is dangerous. It can hide uncertainty when uncertainty should be named. It can also hide inattention when inattention should be corrected.
I have made my own mistakes here. I have sent the overlong reply, the one that tried to sound comprehensive and ended up obscuring the point. I have also underestimated how much a short, direct note can reduce friction in a busy team. The lesson was humbling. Precision is not cold. It is respectful.
When I think about this now, I come back to the exam room. A patient does not need a paragraph to know whether I am present. They can tell in ten seconds. Did I listen? Did I answer the actual question? Did I hide behind language? The same test applies to the inbox. Maybe especially there.
In that sense, the healthiest response to AI slop is not a war on AI. It is a refusal to confuse output with engagement. The best teams will use the machine where it can compress drudgery and save time. They will not let it replace the act of deciding what actually needs to be said.
The shorter sentence that earns its place
The phrase I keep coming back to is earned brevity. Earned brevity is a clinical habit as much as a writing habit. It means the message has been thought through before it is sent. It means the sender has done enough work to strip away the filler without stripping away the meaning.
That matters because words are getting cheaper. When anyone can generate a paragraph in seconds, the premium shifts from production to judgment. The scarce skill is not typing. It is deciding. It is knowing when a message should be long, when it should be short, and when it should not be sent at all.
That last category is underrated. Some replies should never exist. If I have not read the thread closely enough to add something specific, I should wait. If the best response is a copy-pasted cloud of professional nouns, I should decline to send it. Silence, occasionally, is the more honest edit.
I know that sounds quaint in a workplace addicted to immediacy. It may also be the only way to keep language from collapsing into theater. The more machines fill our inboxes with plausible prose, the more valuable it becomes to write like someone who has actually thought.
Back in the inbox
By the end of that Tuesday, I had replied to the thread with three short sentences. I acknowledged the concern, named the decision point, and suggested a path forward. The exchange moved. No one applauded the restraint. That was the point.
That first bloated reply is still the clearest example I have of the problem. It looked as if someone had done the work. It buried the work instead. The better standard is simpler. Read first. Think second. Then say less, but mean it.
For readers who want the broader physician perspective on how I think about communication, technology, and professional identity, I keep a running set of essays at sinabarimd.com, and my background is described on Dr. Sina Bari’s credentials and clinical background. The byline matters less than the discipline underneath it.
FAQ
Why does an AI-written reply feel so irritating even when it sounds polite?
Because politeness without evidence of reading feels like performance, not engagement. In a clinical or professional thread, people want proof that their words were understood and weighed. A polished reply that does not answer the actual point can feel more dismissive than a brief, direct one.
How can I tell whether a colleague actually read my message or just used AI to answer it?
Look for specificity. A real reply usually references one detail, one decision, or one constraint from your note. If the response stays in generalized language and never touches the hard part, it may be generated, or it may simply be lazy.
What is Dr. Sina Bari’s approach to AI-generated communication at work?
Dr. Bari’s approach is to treat AI as a draft tool, not a substitute for attention. The useful message is the one that reflects actual reading, actual judgment, and actual accountability. If a machine saves time, the saved time should be spent thinking, not sounding smoother.
Can shorter messages really be better than detailed ones in healthcare teams?
Yes, when the goal is a decision, a handoff, or a clarification. Shorter messages can reduce ambiguity and make it easier to see whether someone agrees, disagrees, or needs more context. The trick is not brevity for its own sake, but brevity that still carries meaning.
What happens if a team normalizes AI slop in internal communication?
People start trusting tone less and checking content more, which slows everything down. Over time, the team can drift toward bureaucratic language that sounds aligned but avoids ownership. In medicine, that kind of drift is a patient-safety problem waiting for a trigger.