Howdy, picnickers!

Sometimes, the posts I curate for the Product Picnic have me reaching for theory, to connect the disparate parts of the discourse into a coherent narrative. Sadly, today’s issue is going to be much more tangible.

In a move that could only be described as Grinch-like, Amazon has abruptly laid off 14,000 people by SMS ahead of the holiday season (and its earnings report, which they needed to find some way to juice because they spent too much money on nVidia chips).

Andy Jassy, like other Magnificent 7 CEOs, has taken a step back from beating the AI drum and isn’t crediting AI for the layoff. Rather, it is a continuation of last year’s equally massive layoff aimed at hollowing out the middle of the corporate ladder.

Despite Jassy’s denial, the layoff announcement immediately got podcasters baselessly speculating that Amazon has finally cracked the code, and actually gotten AI to replace some jobs like hype artist Dario Amodei keeps promising will happen “by the end of the year” every year since 2023. We are once again asked to believe that tech is on the precipice of inevitable transformation and the next release will change everything, just like the last one didn’t.

The technology is not the product; as Scott Jenson explains, the hype is the product. Even mainstream media like CNN has gotten tired of this bait-and-switch, and started picking up what critics have been putting down for years:

The article is a good overview of the actual state of AI adoption: a lot of demos, a lot of promises, not much in the way of realized gains — because the magic does not survive contact with reality. Not only have the 500x productivity gains not materialized, but (at least in our field) the very things AI tools do today are the wrong things for anyone to be doing in the first place, and using the tools to accelerate those tasks just makes us more wrong.

Instead, employers are using AI as an excuse for short term cost-shedding exercises. Depending on your degree of cynicism, you might consider companies hiring back the same people they laid off “in anticipation” of AI efficiency to be an admission of error, or the plan from the very beginning. Certainly, a lot of job postings going up today read like “we did a big oopsie and need someone who knows what they are doing to fix it.”

Some companies are only at the beginning of the FAFO curve; others, like Klarna and Duolingo, have already reached the end and rapidly backpedaled. Accenture is somewhere in the middle: after failing to “re-skill” their own workforce to be productive with AI they simply fired 11,000 people and now want to sell this process that doesn’t work to enterprises. Again, the product is the hype, rather than observable and tangible benefits.

As for Amazon itself, Jassy’s explanation that culture, rather than finances are behind the latest round of layoffs certainly sounds like yet another industry titan tacitly admitting that going all-in on AI will not pay off.

The AI focus may be gone, but the impossible dream of magically improving productivity remains. This time, the silver bullet isn’t tooling, but slashing the decision-making process. We are asked to believe that this is a culture problem, and can be fixed by mass layoffs.

But trying to fix your culture with mass layoffs is like that old joke, fighting a war for peace. Layoffs damage your culture irreparably, and cause your high performers to head for greener pastures.

And using layoffs to fix the specific thing Jassy is worried about — improving the speed of making decisions by eliminating process — is especially counter-productive. I’ve written more than enough about how good process actually makes us go faster. Hell, this is basically Good Process Makes Us Go Faster: The Blog. But this post by David Demaree hits the nail on the head for why what Jassy is doing won’t work for Amazon specifically:

But Pavel, you might object, that post is about Meta and not Amazon. Maybe Amazon’s problem is actually different.

Alright, let’s take a case study: Amazon Prime Gaming — a platform that aimed to compete with video game marketplace Steam, but that most PC gamers didn’t even know existed.

Does this sound like an execution problem? Were teams taking too long to make decisions, like Jassy claims? Or were half-assed PRFAQs getting rammed into roadmaps as quickly as it was possible to write them, without the friction of a real process to get everyone to pause and figure out if any of it was valuable?

There’s no such thing as “no process.” There’s either intentional process, or unintentional process. And actual innovation requires intentional process.

Sure, it needs to be the right type of process (bottom-up adaptive, rather than top-down predictive). But trying to do “no process” means that what you have is an unintentional process, with teams left to fend for themselves against the prescriptive whims of highly placed stakeholders. In other words, it’s the perfect way to scale the dismal failure of Prime Gaming to the rest of the company.

Managing deliverables rather than outcomes — trying to chase certainty of the design process, rather than finding certainty through the design process — will always fail to achieve real impact.

And while Amazon has consistently failed to learn this lesson, the public sector is managing a lot better. James Plunkett wrote about the long journey of shortening feedback cycles in government tech with a list of great examples and lessons. If dense government bureaucracy can do it, then a tech company can do it too. Just not through firing all the people with a clue.

— Pavel at the Product Picnic