Scott Stevenson doesn’t want every employee’s idea to get an AI glow-up.
The cofounder and CEO of Spellbook said workers’ proposals containing “AI nonsense” routinely make it to his desk. It might be a project plan or a marketing pitch. Whatever it is, he said, it can arrive looking far more polished than the thinking behind it.
Stevenson said he would rather see the rough version first.
“I want the D+ version of your plan,” he said.
Stevenson is now asking employees at the Toronto company, which develops an AI tool for drafting and analyzing contracts, to resist the temptation to turn every idea into polished corporate prose.
His message is a notable one coming from the CEO of an AI company: Use the technology, yet don’t let it obscure your thinking.
AI slop on social media and networking sites like LinkedIn has drawn plenty of attention. Stevenson is worried about a workplace version of the problem.
Other companies are grappling with this challenge, too: The head of AI video company Synthesia warned staff earlier this year about “AI sloppification,” writing that overly wordy documents could mean that instead of one staffer taking 10 minutes to sharpen a document, “we have 10 people spending 10 additional minutes reading slop.” Business Insider previously reported on AI software startup Clay’s new companywide AI-writing policy. It focuses on four principles, including “writing is thinking” and “longer is not better.”
Stevenson is trying to tackle the problem with his own guidelines. He recently sent a memo to the company’s 200-plus employees updating the company’s AI policy and outlining how he wants them to use the technology.
Here’s an excerpt:
While we are an AI company, and believe in the usefulness of AI, we think that today it works best in highly constrained work like coding or contract review. We do not believe that models are good at unconstrained longform writing, yet. Documents end up too long while saying too little, use repetitive language that decreases trust in your thinking, and sometimes contain ungrounded nonsense.While being able to create a polished longform document with a prompt may feel powerful and professional—this impression is often only surface level.Good writing is concise and authentic. I would rather that you send a scrappy 5 bullet point proposal than 5 pages of polished fluff.Furthermore, AI writing has become increasingly recognizable.We want our customers to always feel our care when we communicate with them, and we don’t think that AI generated writing accomplishes this.Our new AI policy is as follows:You should indicate if a document or message you are sending internally was drafted with AI: this helps the reader digest it appropriately.AI should not be used for writing the main body of project proposals: we want to see your authentic reasoning. You may use AI for supplemental materials like spreadsheets and tables, but not for the core of your reasoning.
a. When in doubt, send a “D+ Version”: we often don’t want a long or polished document. Concise authentic bullet points are often enough to get a conversation going. Thinking about how to make your argument succinct is helpful for the reader.
Writing you send will be treated as your own: A lawyer is accountable if they submit an argument with hallucinated case law. We will also hold ourselves accountable for our own writing. Please review everything that you send to the team thoroughly, as it will be treated as your ownCustomer communications should always be hand-written: We want our 5,000 customers to feel authentic human care with every interaction.
Is your company rethinking its policy on AI in the workplace? How do you feel about receiving writing produced with AI? Let us know in the comments below.
