Picture a rep who writes a prompt that cuts account research from an hour to ten minutes, while the next rep on the team has nowhere to find it.
The survey puts numbers on a familiar habit
On October 27, Salesforce published a Pulse Survey of 250 U.S. executives at director level and above, all with decision-making authority over AI investments, fielded from September 14 to 22. Each week, 68% of them see insights from someone’s AI use go unshared, 66% see two colleagues use AI to solve the same problem without knowing it, and 59% see work stall because it sits in one person’s AI tool. Executives estimate 3.6 hours a week lost to switching between AI and other apps. Only 15% say their AI is embedded across multiple systems.
Two cautions come before I use these numbers. The survey is Salesforce’s, and Salesforce owns Slack, whose general manager is quoted in the write-up. The respondents are executives across functions and the examples in the post are mostly engineering ones, so none of it is a measurement of sales teams. I use the pattern and treat the percentages as a loose guide.
Advertisement
My position
A sales team should treat a prompt that works the way it treats a call script that works: it goes into a shared library with an owner, a date and the result it produced. The best prompts a rep has written are team assets. Right now they live in private chat histories, and they leave when the rep does.
Rob Seaman, EVP and GM of Slack, put the general case this way: “The single-player era of AI made people faster. The multiplayer era is what makes a whole company smarter, together.” His line sells a product, and the underlying point still holds for a sales floor. A rep who speeds up their own research gains an hour a day. A team that shares the method gains that hour on every rep who copies it.
Sales has a head start on the habit. Teams already keep battlecards, call recordings and objection-handling notes in a shared system because a single rep’s good call is worth little until the others can hear it. Sales leaders already run it that way, and buyers now arrive with AI answers of their own, which raises the cost of every rep improvising a different answer. Prompts, agent instructions and the checks reps run on AI output belong in the same place as the playbook.
The strongest objection
The best argument against shared AI context is exposure. A shared workspace gives more people access to what went into it, and a sales prompt can hold pricing, a customer name or notes from a deal. If the library fills with copied account details, the team has built a leak. That risk is real, and it is the reason the library needs an owner who decides what goes in.
It is also the reason I would not let the habit stay private. A private prompt with a customer’s contract terms pasted into it is already a leak, one that nobody reviews. We argued in Name the Owner of Every Sales Agent Before You Write Its Policy that an agent nobody owns is a liability. A prompt nobody can see is the same problem one step earlier. A shared library with a rule that it holds method and no customer data is easier to check than forty private chat histories.
A second objection is that sharing flattens skill, so every rep sounds the same. A good prompt is a starting point that each rep edits to fit their accounts, and the library should keep the edits too.
What I would do this month
Ask each rep to submit the one prompt they would be sorry to lose. Put them in a single folder or CRM note type that RevOps owns. For each one, record who wrote it, which task it serves, what the rep saw it produce and the date it was last checked against current product and pricing. Ban customer names and contract figures from the entries. Review the folder every quarter, and delete anything nobody has used.
The 66% figure in the Salesforce survey describes colleagues solving the same problem twice without knowing it. On a sales team, that is two reps building the same account-research routine in the same week. A folder with twenty entries ends that, and costs an afternoon to set up.
