AI Can't Persuade

AI can do a lot of things.? But it can’t persuade.? Heck, humans can barely persuade. ? As Ben Franklin said, “People are best convinced by reasons that they themselves discover.” ? It’s all we can do to guess what reason would be most persuasive to another human and then drop breadcrumbs that lead them to it.??

Despite that, everyone is discussing the myriad ways AI can multiply sales results.? It’s a very attractive idea.? By comparison, consider an Amazon fulfillment center - If you have never toured an Amazon fulfillment center, I highly recommend it - it’s fascinating.? It quickly becomes obvious that it is humans working for machines.? They have automated the entire order-to-ship process and have plugged humans into the few remaining spots where robots and AI can’t do it as effectively.? And are methodically replacing those few remaining workflows with automation as quickly as the computer catches up.

Plugging AI into sales begs to be similar.? Automate everything, and plug humans into the hard parts.? This may work for simple, high-velocity sales cycles.? But a sales process that requires any level of persuasion, creativity, or strategic thinking - e.g. the classic B2B enterprise sales cycle - can not be automated.

AI can help - but it needs to be AI working for humans, not the reverse.? Use it to automate the rote, repeatable tasks - examples include account research, prospect identification, and email sequencing.? Ideally, this should already be moved to roles that are optimized in terms of skill and cost based on the task.? If that hasn’t been done yet, there is an opportunity to do it now and leapfrog your overall efficiency forward..

Here are 2 things AI won’t fix (and might make worse):

• AI will not solve a persuasion problem.? If the dogs aren’t eating the dogfood, no amount of automation, acceleration, or increased volume will help.
• AI will not fix disconnects between departments.? A Marketing to Sales gap will not be improved.? If there is no alignment on what is an MQL, producing more bad MQLs faster won’t solve it.? “We need more leads” is usually “we need more good leads.”? First, agree on what is good, where & how to get them, and then tackle automation.

Here are several places where it can be applied with high value for the time invested. ? These are classes of use cases, there are many variations.

• Provide frameworks and starting points for persuasion.? LLM’s are not bad at creating content that is helpful to sellers as they prepare for meetings, customize emails and decks, develop discovery questions, and create value propositions.? And they do it fast, with unlimited variations (by role, industry, pain point, etc. etc.)
• Do summarization and analysis of sales call transcripts.? Most call recorders do this during or immediately after the call.? Build on those transcripts and summaries by bringing them together into meta-summaries that find themes across a series of meetings or an entire sales cycle.
• A more difficult strategy, but worth exploring in larger organizations with a high volume of opportunity data. ? Use the pattern-identification capabilities of vector databases (an underlying LLM technology) to find commonalities across leads, opportunities, and sales calls.? Look for both the good and the bad.

I believe 10X selling is possible for enterprise sellers, but not with end-to-end automation of the sales process.? It requires decomposing the process and applying automation to specific tasks, with adequate guardrails and tracking.? Now is the time to explore ways to make sellers much more efficient and productive.