Ignore that Goldman Sachs report questioning AI’s value. It doesn’t impact what’s happening in your workplace today.

Listen to a podcast version of this article ??

Last month’s issue of the Goldman Sachs “Top of Mind” newsletter focused on Generative AI (Gen AI), suggesting that the promises of Gen AI transformation are overhyped, and the $1 trillion being spent on capex (chips, datacenters, power grid, etc.) has yet to show value and might not ever pay off. (A PDF of the newsletter, titled “Gen AI: too much spend, too little benefit?”, is available here.)

The newsletter has been the subject of many articles and podcasts over the past few weeks. In fact, three separate people sent it to me, asking whether it affects my views on workplace AI, and/or whether I need to reconsider the investment I’m making in learning about this space.

Reactions to this GS newsletter vary based on perspective:

• AI Pessimists: “This is more proof that Gen AI is just fancy autocomplete which spits out unoriginal recycled garbage stolen from creators. AI is an overhyped bubble.”
• AI Optimists: “Economists look backwards. Gen AI is in its infancy, and it’s premature to expect a killer app yet. Its development is on course to change the world.”
• Me: “This report is about the macroeconomic impact of Gen AI, which is an academic exercise by an investment bank. It has nothing to do with how your workplace is being impacted by your employees finding and using AI tools on their own today.”

It doesn’t matter what this report says or who's right

As I alluded to above, if you’re a workplace manager or have responsibility for how technology is used in your company, the opinions of some investment bankers about the ROI of AI investments over the next decade have absolutely nothing to do with your company’s daily operations or how your employees are using the AI tools they’ve discovered.

While the impact of AI might not yet be reflected in macroeconomic data, it’s definitely real and measurable if you talk to individual employees. The more people I talk to, the more I get the sense that there’s a good sized “underground” use of AI tools by employees at companies today.

I’m not talking about the promise of the “citizen developer” or the “AI-powered assistant” that many experts believe will come into the workplace over the next few years. I’m talking about something way more basic: employees copying and pasting things into ChatGPT, Perplexity, Claude, etc. for all sorts of run-of-the-mill office worker tasks, such as:

• Writing emails (often with copied and pasted email histories to get the context and tone)
• Generating first-drafts of documents, or outlines for documents.
• Getting help starting projects, reports, papers, assignments. (Basically for anything that gets assigned to them, start by throwing a ton of questions at the AI to get a starting point)
• Project planning: steps, timelines, risks.
• Creating presentations (or outlines of presentations) from documents or folders of docs.
• Generating SWOT analyses, outlines for strategic plans
• Prioritizing to-do lists, task lists, summarizing emails
• Summarizing meeting notes (from transcripts), generating action items
• Researching competitors, creating competitive analysis reports, learning about product features, functions, and limitations
• Generating creative ideas (marking campaigns, product features)
• Summarizing long reports or research papers into key takeaways (either for them or for others)
• Creating visualizations of data. Finding “interesting things” in the data.
• Writing and polishing performance reviews, self-assessments, or recommendations.
• Generating interview questions for job candidates / generating sample answers to expected interview questions
• Writing social media posts (posts and blogs, articles, etc.)
• Generating product descriptions & marketing copy
• Analyzing feedback and sentiment (from customers, session reviews, etc.)

All of these tasks are getting easier and more practical due to the incremental enhancements of common AI products widely available to non-technical people. They all understand images now. You can drag-and-drop documents and files. Claude has a "projects" feature where you can maintain context around many different projects at once. Perplexity lets you conduct research and assemble reports. Week-by-week, these tools are becoming more useful, even if their core foundational models are not changing.

Most striking is that your employees don’t need an official corporate subscription to do any of this. They are using AI for these types of tasks, dozens of times a day, on their own, whether you know it or not.

It's hard to understate the potential risk to the workplace this will become over time. Do you have any idea which of your employees are doing this? To what extent? What tools are they using? What are the biases? Are they compliant? Are employees using these tools to create better quality work? Or are they created worse quality (yet passable) work in less time? What are they doing with the extra time they free up?

But this is just a small minority of employees!

Some people argue that while it’s true that some employees are doing things like this, it’s not a big deal because it’s just a minority of the employees. (Last week Ethan Mollick suggested only 10% of employees have tried Gen AI, and only 2% use it regularly.)

Even if this is “just” 10% of your employees, this is the 10% that you should care most about. Employees who are taking the initiative to figure out and learn AI tools today, on their own, with no support from the company, are the employees you want to reach out to in order to understand where AI can be used and what the future of your workplace will look like.

Even if you argue that these employees are the “lazy” ones who are trying to use AI in dishonest ways to automate away parts of their job, wouldn’t you still want to know about them and understand where, why, and how they're using AI? Companies are paying huge fees to management consultants and AI experts to study their processes and figure out what they can automate. Why do that when you can just look around and see what your employees have already figured out? (And maybe give them a raise!)

There’s simply no plausible excuse to do nothing. (I’m not saying you must embrace AI, or change how your workplace operates. I’m saying you need to understand how your employees are using AI, including the use cases, their motivations, what works and what doesn’t, what tools they’re using, etc.)

The risks of doing nothing will continue to grow. The underground employee rumor mill will circulate the best practices and tools that employees figure out on their own, which is great! But you should make sure you’re part of that conversation. Don’t wait until your employees self-select into the haves and have nots.

If you don’t embrace workplace AI, by ignoring it, or (even worse), trying to block it, the 10% smartest and most engaged employees quit and go work somewhere else, where they’ll become super-humans accelerated by AI. At some future point you’ll have to embrace AI use by your employees, but with a workplace full of the leftover employees who didn’t care enough to try to improve things on their own.

The AI bubble might burst, but ChatGPT will still exist

Circling back to the Goldman Sachs newsletter questioning the macroeconomic value add of AI, I want to reiterate that it doesn't matter to what’s happening on the ground in your workplace today.

Everything I’ve written about in this article is based on existing technologies and products which your employees are using today. This is not some crazy future which assumes a major economic impact or progress towards artificial superintelligence, rather, this is the dumb, boring, run-of-the-mill everyday AI your employees are using to write emails and kick of projects today.

AI technologies will continue to advance. Even if Gen AI models don't get any “smarter” and don‘t achieve AGI, they will continue to become smaller, cheaper, and faster. If only 2% of your employees have leaned into AI today, that means you have a 50X increase in employees using it while it becomes available everywhere.

This is something you need to get in front of now. Even if you don‘t spend any money or buy any products, at least start understanding what the pressures and frictions are, and how these things work. (At a minimum, put in the ten hours or AI use yourself and start to see how your employees see these tools.)

All of this is exactly what I’m exploring in the Workplace AI Strategy Guide project I announced a few weeks ago. Let's figure this out together!

The Workplace AI Strategy Guide project is at strategyguide.ai. This article is AI Influence Level 0: Human created and conceived with no AI assistance. Details at strategyguide.ai/ail.