Aspects of AI Usage Impacting AI Adoption
By Andreas Welsch
Founder & Chief AI Strategist, Intelligence Briefing
Three papers have recently shaped my thinking on how leaders can support AI adoption. On their own, each paper delivers important insights. But when you connect them, a much clearer picture emerges.
You might have experienced this over the last few months: You receive a note, a draft, or a report, and as you start reading it, something feels off. It’s good, but not great. You read it three times and still can’t figure out what it’s actually about. Maybe you even catch some prompt fragments in the copy.
AI slop, whether on TikTok or in your inbox, is taking over what we read (and write). But a business cannot afford to create ‘good enough’ or generic information, let alone it being factually incorrect. Not using AI is not an option either. But what can leaders do to encourage AI use on their team without sacrificing quality?
Shadow AI is a Social Phenomenon
Last fall, Slack published its Workforce Analysis Index. 48% of respondents shared that they use AI at work without telling their manager. Why? Because they’re afraid of being seen as lazy, incompetent, or simply getting more work to fill the newfound free time. A study by Duke University published in May not only confirmed these findings, but also added the dimension of how peers interact and hide their AI use from each other—for the same reasons and concerns, by the way.
Here’s what that means: Shadow AI use is a real thing, and we need a real leadership shift to embrace AI fully. Leaders need to reward results rather than time or presence.
Cognitive Offloading Degrades Knowledge Retention
Researchers at MIT examined the effects of cognitive offloading, aka delegating too much to ChatGPT & Co. Their paper, “Your Brain on ChatGPT,” published in June demonstrates nicely how individuals retain little information if AI tools take over entire workflows (e.g., research essay topic, propose essay structure, and write the essay) compared to AI-augmented tasks (e.g., research essay topic and propose structure), and fully human-created work where all these steps are performed without the help of AI.
Here’s what that means: Despite the push for companies to become “AI-first,” delegating too much scope to AI can degrade your professionals’ knowledge and reduce the quality of the results.
AI Slop Influences Our Language
A study published by researchers at Max-Planck Institute for Human Development found that humans incorporate the language patterns that you notice from AI assistants into their spoken language. This includes structures like “it’s not just this; it’s also that.” or “it’s not this—but also that.” as well as certain keywords like delve, comprehend, and swift that occur more frequently in AI-generated text.
Here’s what that means: AI is shaping how we communicate. While many people adopt these structures and words, others become more self-conscious about their writing and purposefully omit them. The presence of an em-dash (—) does not automatically mean (sloppy) AI use.
Leaders across the organization play a huge role in scaling AI adoption. They can support their company and teams in three ways:
- Empower the team to use AI
Your team members are already using it anyway. Offer appropriate training to lower the barrier of entry—from basic AI literacy to advanced use of tools. Encourage your team to challenge the status quo of a workflow or process and explore where AI-enabled tools could support them. Create opportunities for dialogue and exchange by inviting a team member to present on your next team meeting how they use AI.
- Provide clear guidelines for AI use
Your organization already has an AI use policy that you can build upon. If not, think about a few simple guidelines for how you would like them to use AI without sharing the first draft out of ChatGPT & Co. verbatim. Encourage the use of AI to create drafts that a human reviews, checks, and augments. The value of context, tacit knowledge, etc. that human team members add is super important here and help keep them “in the loop” of things.
- Spot-check work products before publication
Just like with any other work product, check AI-supported results from time to time to see if it meets the quality requirements of your team and organization.
If it doesn’t, provide specific, actionable feedback to the team about what to look for the next time they create similar information. High-performing teams only reach that level if the expectations are clear and how to exceed them.
At the end of the day, your team members should remain accountable for their work, whether or not they use AI in the process. Leaders can support their teams’ AI use by empowering professionals, providing team guidelines for AI, and providing timely and actionable feedback on what high-quality work products look like.
Side Note: Andreas Welsch is an internationally recognized top 10 Agentic AI leader. With over two decades of global leadership experience—most recently as Vice President at SAP—Andreas has helped Fortune 500 companies cut through AI hype to deliver measurable business outcomes. NWIRC is offering a 4-week virtual training led by Andreas, ‘AI Explorer’, to raise AI literacy in manufacturing. The series will start on November 6 and this opportunity is limited to only 20 people. Find info at nwirc.org/events.
Original Source: https://intelligencebriefing.substack.com/p/ai-slop-how-to-build-a-high-performing-ai-ready-team