Question Best Practices in the AI Era

Between cultural indoctrination and LLM interpretations, there's a lot to question.

When I consider how information is processed in the AI era, best practices often amount to recirculated information, slop, and the common wisdom of the crowd. Accepting them provides a safe answer that may weigh someone down in mediocrity.

First, consider that LLMs are probabilistic calculators of best-practice information tailored to your prompt, based on their training data. And that training data is the public web, including sources like Reddit and other social networks. Among ranking factors, recency and social cues help the search engines used by the LLM to rank the “value” of the content.

I am about to launch a new podcast, and the advice I have received about it has been interesting. Much of it is based on best practices, some from people, some from LLMs. All of this information is appreciated, yet advice and definitive lists of best practices in the AI era all strike me as worth questioning.

For many podcasters, that might be just enough to equate to success. I know in some of my work, for example, monthly accounting, etc., good enough works. But in a competitive marketplace like B2B-focused AI podcasting, best practices for length and format will make it just another podcast. And if that’s the case, why would you listen? For many, that’s failure.

Then there is the challenge of changing media dynamics. AI is not just fueling a significant amount of content; it is changing the way information is consumed, from what social network algorithms and LLMs are sourcing, to how people are choosing to receive this information. Trust in the AI era is hard to come by and with good reason. The resulting AI backlash has not been positive.

Can best practices really be trusted in this environment? I am wary.

Are Best Practices Holding You Back

Best practices are a classic example of fixed ideas, which always seems to be challenged in the face of changing circumstances. The way we have always done it or what is considered the industry’s best method, is an excuse to avoid innovation and evolution. Best practices, in actuality, can become a barrier to success.

This fixed mindset is well discussed as a major barrier to AI adoption in Now Is Gone. That being said, best practices have been subject to scrutiny for well before AI became a technology trend.

Freek Vermeulen in a Forbes article on best practices said, “…when circumstances have changed, and it has become inefficient, nobody remembers, and because everybody is now doing it, it is difficult to spot that doing it differently would in fact be better.” Vermeulen’s article goes further, saying that best practices may indeed never be best but rather the result of self-perpetuating myths, such as which movies wil be successful based on genre.

In turn, whole swaths of people can embrace these practices because everyone is doing it, and if everyone is doing it, then by adopting them we will get the same results. This is called the cargo cult thought process, where you create a landing strip on a remote island assuming cargo planes will suddenly start landing there.

As applied to business by James McElroy, cargo cult thinking is, “the confusion caused by mistaking the process, input, or symbol with the goal, output, or referent. Second, it is when part of a system is mistaken with the entirety of the system.” McElroy goes on to say best practices are laced with cargo thinking.

For example, if we write Substack posts like every political writer, I will get the same subscription revenue as they do. In actuality, I am not a consumer writer focused on the broadest of topics in the current American cultural moment. My audience is B2B, with a heavy focus on the Greater Washington/Mid-Atlantic region, with distinct information consumption habits, sales cycles, and needs. Certainly some of the best practices apply, but which ones? Which ones are irrelevant?

How to Use Best Practices: Go Beyond Social Proof

If best practices are simply accepting a group of people or an institution’s validation of a method, then we are blindly accepting social proof. In my business and personal life, generally, I don’t want to outright dismiss best practices, but I don’t want to accept them as 100% certified fact either.

I think like any piece of information, the best way to use best practices is a starting point for solutions. These practices are starting points, areas that can be used as hypotheses for additional research and testing. Consider what you would receive out of an LLM as a best practice. Would you blindly accept it, or research the AI’s sources?

Here are some questions to consider… Are they good ideas? Are there inherent biases from creators or past eras? Can these practices be proven with data? Have newer innovations surpassed these best practices? Has the market changed recently, requiring a different approach? Most importantly, if successful, do the best practices meet my needs?

As you can see, there are many factors to consider when weighing best practices. Referring back to Now Is Gone again, this is a lot of the inventory process, weighing what is good and what is not. Continuing that process, solutioning takes into account what can be used from such practices, embraces them, and incorporates them into new approaches to resolve challenges.

One could say the best practices, whether adopted or not, served a great purpose: A starting point for solutioning. In that sense, they add context.

How do you use best practices in the AI era?