The D+ draft: an AI company's case against AI writing

Sep 21

An AI company just told its staff to use less AI. Spellbook's CEO wants the "D+ version" of your work, and customer emails written by hand.

Scott Stevenson, co-founder and CEO of Spellbook, sent a memo to his 200-plus employees updating the company's AI policy, and Business Insider published the details. His complaint: "AI nonsense" keeps landing on his desk, arriving far more polished than the thinking behind it. So he asked for the opposite. Send him the rough version. The D+ version.

Jenni and I spent a good chunk of Episode 75 on this, and the memo is sharper than that headline makes it sound. Stevenson's line is one every communicator should sit with: "I would rather that you send a scrappy five bullet point proposal than five pages of polished fluff." Coming from the CEO of an AI company, that is a real position, not a Luddite reflex. He is going after people using AI to hide thin thinking, and that problem is everywhere right now.

Where I pushed back is the last rule in his policy: customer communications should always be hand-written, so his 5,000 customers feel authentic human care. I like the intent. The wording falls apart the second you look at it. If you type an email, it is not hand-written. We have quietly redefined the word to mean "a human was involved," which is a strange hill to plant a flag on.

The bigger problem is the whiplash. For two years the advice was to use AI for the first draft. Now leaders are saying don't. And nobody, at any level, wants to hand their boss D+ work. Ask for a rough version and people will give you one. Ask for a D+ version and you have told them in advance that mediocre is the target. I would rather use AI the way I used it recently inside the ICology community: I wrote the email myself, then used Claude connected to Mighty Networks to find members with incomplete profiles and inject the specific detail each person was missing. The writing was mine. The personalization was the machine. That combination felt more human to the recipient, not less.

Jenni named the real issue, and it has nothing to do with first drafts. The problem is someone sending you five polished pages they can't actually talk about, because they never did the thinking underneath them. That is a performance management conversation, not an AI policy. And I will keep pushing on the "AI writing is now recognizable" claim too. We think it's recognizable. But the models learned from us, and now we're learning from them, so the language is converging. Plenty of what gets flagged as AI slop is just writing, and acting like everything before 2023 was A+ work is its own kind of fiction.

Also in this episode:

Jenni opened with The Advantage by Patrick Lencioni and the case for chasing organizational health instead of employee engagement. Her question was who owns clarity when it has to cascade down through every layer of leadership. My answer: it starts and ends with the CEO, and if clarity fails, that is on them. We also hit the old rule that people need to hear a message seven times. Jenni had a stat from an event last year putting it closer to 29 now, which is exactly when most communicators quit repeating themselves, right as employees are finally starting to listen.
We worked through Stephen Waddington's Wadds Inc. summer report and whether its findings translate from the UK to the US. For the most part, they mirror each other. Intangible assets are about 92% of S&P 500 value, which means reputation is the balance sheet now, and yet companies keep laying off the knowledge that creates it. On the purpose retreat, I was blunt: staying silent is still taking a position, and dressing it up as neutrality is cowardice. The number that stopped me was buried near the end, female practitioners in caregiving roles averaging 91-hour weeks, which is not a workload, it's a warning.
We dug into a piece on how the doing got automated but the talking didn't, built on Predictive Index data. Anthony Belluccia's "wire mother" image stuck with me: every company is building a beautiful AI that dispenses answers, but when change gets scary, nobody runs to the dashboard. They run to their manager. The survey gap says it all, with 74% of CEOs calling their managers very confident in tough conversations while only 20% of those managers would go in without a framework or help from HR. The leaders admitting it's hard don't worry me. The 20% winging it do.
Jenni's bonus fifth story came from a clip of Priya Parker on The Interview, arguing that Slack turbocharged workplace conflict, with a question about why Slack gets the blame when Yammer came first. My take: we blame the tool because it's easier than owning how we use it, the same move as saying you hate PowerPoint when what you hate is bad PowerPoints. Slack didn't create the conflict. It surfaced what was already there and made it searchable.

Watch or listen to the full episode on Frequency, with timestamps, links to every story and the full transcript.