How is AI impacting productivity?

Dr. Ryan Ries here, back again after taking the Memorial Day week off. This week, I have 3 things I want to chat about:

1. GenAI Increased Productivity vs. Decreased Motivation – the current tradeoff
2. Breaking news from the monkey world
3. Anthropic’s Claude 4

Productivity vs. Motivation Tradeoff

A new Harvard Business Review study involving over 3,500 people dropped some sobering findings about GenAI in the workplace. 

While GenAI collaboration boosts immediate task performance and efficiency, it can undermine employee’s intrinsic motivation and increase feelings of boredom when they turn to tasks without AI assistance. Employees who used GenAI for one task and then switched to a non-AI task reported an 11% drop in intrinsic motivation and a 20% increase in boredom.

Not great.

Here's what really caught my attention: performance reviews created with GenAI were "significantly longer, more analytical, and demonstrated a more helpful tone," and emails were "warmer, more personable… containing more expressions of encouragement, empathy, and social connection." The quality improvements are undeniable.

GenAI often removes the most time-consuming and tedious parts of our tasks. 

But at the same time, for some roles, genAI also takes care of the most challenging and stimulating parts of a task—those that make work meaningful (i.e. content creation, idea generation, etc). 

When AI generates much of the content, employees may feel disconnected and less in control.

My Take

The immediate productivity gains with genAI are real and valuable, but we need to be thoughtful about implementation. The key is what the researchers recommend: blending AI and human input rather than full automation.

For example, instead of having AI write entire performance reviews, use it for outlines and structure, then let humans add the personalized insights and context. 

This preserves the cognitively engaging parts while leveraging AI's efficiency.

As we scale AI across organizations, we can't ignore the psychological costs. 

Monkey Business Goes Viral (Literally)

Okay, this story is wild. Not related to our typical topic, but I just found it super interesting. 

Scientists studying capuchin monkeys on Panama's Jicarón Island just published findings about young male capuchins kidnapping baby howler monkeys from another species entirely. Between January 2022 and March 2023, they documented at least 11 infant howler monkeys being carried around by juvenile capuchins for up to 9 days at a time. 

This is the really crazy part. 

This behavior started with one monkey nicknamed "Joker" and then spread to other young males. Scientists think that this behavior emerged from sheer boredom since capuchins live in a predator-free environment. 

Anthropic’s Claude 4

Now, back to our normal AI chat.

Anthropic dropped Claude Opus 4 and Claude Sonnet 4 before the Memorial Day holiday, setting new standards for coding, advanced reasoning, and AI agents. 

Claude Opus 4 is, at this point, the world's best coding model, with sustained performance on complex, long-running tasks and agent workflows.

Key highlights:

• Opus 4 leads on SWE-bench (72.5%) and Terminal-bench (43.2%), dramatically outperforming all Sonnet models and significantly expanding what AI agents can accomplish
• Extended thinking with tool use beta: Both models can use tools—like web search—during extended thinking, allowing Claude to alternate between reasoning and tool use to improve responses
• Claude Code is now generally available with new beta extensions for VS Code and JetBrains, plus a Claude Code SDK for building custom agents

What This Means for You

The coding improvements are impressive, but what excites me most is the sustained performance on long-running tasks. Companies like Rakuten validated Opus 4's capabilities with a demanding open-source refactor running independently for 7 hours with sustained performance.

This is huge for complex, multi-step processes that previously required constant human oversight. Think end-to-end data pipeline development, comprehensive code refactoring, or building entire applications from scratch.

How does this all tie together?

AI capabilities are advancing faster than our understanding of their workplace impact. By the time we try out one model, a newer, faster, stronger model emerges.

While this is all incredibly exciting, the HBR study reminds me that we need to take time to slow down. To preserve human engagement and creativity.

And, I guess here’s how the monkey story can tie it all together.

When basic needs are met and cognitive challenges disappear, both primates and humans can exhibit unpredictable behaviors. 

Anyways, let me know what you think about all of this and your thoughts on Claude 4. My team is very excited that it’s already available within Bedrock.

Until next time,
Ryan

Now, time for our AI-generated image and the prompt I used to create it.

Generate an image of a capuchin monkey using a computer. The monkey looks happy and is giving a thumbs-up. It is using the Anthropic Claude 4 model to build a chatbot. The monkey is wearing an AWS t-shirt.