#58: Hot Takes in AI (Part III)
Communication, market risks, and the bigger picture
Written communication will become the most important skill in the workforce.
I see this playing out in real time. As AI becomes part of how people get things done, the ability to turn your goals into clear, specific prompts will become the differentiator. Those who learn to give AI detailed, structured instructions will boost their productivity, while others will keep using it more like a search engine and miss out.
OpenAI recently released a fascinating report, discussing the most common use cases for ChatGPT based on a prompt analysis. Recent trends show that about 70% of ChatGPT use is for non-work purposes, mainly writing and decision support.
What does this mean? ChatGPT serves as an advisor or an assistant rather than a replacement. It’s making people more efficient and in theory, leading to sharper decision-making.
Because we now have access to technology that can write for us (or help us think through how to write sharper), we need to up our game as writers.
I see written communication splitting into two key audiences: AI and humans. Writing for AI will require rational and procedural thinking, consistently breaking problems down into step-by-step instructions. This will require patient back and forth, providing context at each stage and understanding which phrases are more helpful to AI than others.
Who is well positioned for this shift in writing? Teachers and doctors are solid examples of people who are familiar with context engineering if they didn’t realize it. These professions require giving very intricate instructions to their student or patient on how to handle various situations. It’s not only software engineers who can take advantage of AI to make themselves more productive. It’s anyone who is used to or is willing to invest in strong procedural writing.
On the other hand, writing for humans will emphasize storytelling skills. Because AI can produce a near infinite amount of content, those that can cut through the noise with a distinct writer’s voice will stand out. A brand is a story, and those that can communicate better stories about their product assortment and brand ethos are poised to win loyal customers. In a world where there’s near endless words, pictures, and video vying for attention, a well-crafted narrative will help deepen the emotional connection a customer has with a brand.
And this carries over well to the advertising landscape. As we’ve talked about in #54: Your Next Customer Might Be a Bot, marketers will shift their attention to brand awareness campaigns, as AI handles direct response execution. The stories we tell will drive acquisition of new human customers, while AI acquires AI agent customers.
AI has created a bubble in the US stock market, but it will keep growing for quite some time.
On a recent episode of Prof G Markets, it was noted that AI companies (a bit broad, but I’m assuming Big Tech) are operating at a valuation that implies they “will be able to cut costs or grow companies’ revenue through the use of AI by $1 trillion in the next 24 to 36 months.”
As we discussed in #53: Why GenAI Isn’t Supercharging Businesses, most enterprises haven’t seen step-function revenue gains courtesy of AI. And although thinking AI will magically transform your business is foolish, we will need to eventually see some top-line momentum to believe that the $1T figure is within reach. Or we should expect valuations to plummet.
With that said, AI’s potential to save companies money and subsequently boost profitability seems to be more likely in play. As companies reduce hiring plans and their overall headcount because of efficiencies driven by AI, the downstream effects on the economy are a bit frightening.
Let’s say that the $1T expense savings is achieved. That means the economy will lose millions of jobs (depending on what average salary AI may eliminate), this could translate to tens of millions of jobs. We don’t need to get remotely close for this impact to throw the US economy into a recession.
It’s a catch-22 situation. Achieving $1T in an AI benefit on the expense side validates AI’s lore but at the same time destroys the employment picture.
We’re clearly in an AI bubble. But as an investor (not financial advice), I think that’s okay for now. Because people have adopted AI more so as an advisor instead of for workflow automation (OpenAI report cites only 5% of prompts are related to coding and data analysis execution as of July 2025), I imagine investors will be patient for agentic workflows to make their way to our professional and personal lives (think the development of personalized AI assistants and browsers). With that said, OpenAI’s low percentage of workflow automation prompts may be the result of other models gaining steam (thinking of Claude).
I see investors falling into two camps. There are the traditional investors, the venture capital crew, that is pouring money at record-setting clips into anything with an “.ai” domain. Then there’s the corporate executives who are signing up for enterprise AI solutions mostly out of fear of falling behind (no one wants to tell their board of directors that they don’t have an “AI strategy”).
I see the unraveling of the AI bubble occurring when one (or both) of the following factors occur. First off, when venture capitalists need their portfolio companies to transition to profitability mode. I saw it happen in early 2022 on the back of sky-high valuations set in 2020 and 2021 thanks to the zero-interest rate policy (ZIRP) era. Eventually, these AI companies need to either be self-sustainable (read: profitable and cash flow positive) or have enough momentum to entice a new investor or enter public markets. I expect a major wash-out as venture capitalists shift from hyper growth mode to sustainable growth trajectory and begin to mark down their portfolios. However, this will take a few years and may not happen until 2027 or 2028 at the earliest, after factoring in typical VC investment horizons of 5-7 years. However, I’m speculating.
What may happen sooner is the event that corporate executives throw in the towel because they just can’t figure AI out. This may be the nature of the culture of their company or the lack of vertical applications for their niche industry, however it’s unlikely that everyone is going to see a fantastic return on their AI investment. Digital transformation is difficult, especially for large organizations.
When a corporate executive for a major public company highlights that they’re pulling back on their AI budget, look out. That means lower sales for the chip companies and for Big Tech in general. These sky-high AI valuations are assuming blistering growth and if anything comes out suggesting otherwise, the house of cards will fall.
But again, I believe we are still in the early innings of AI based on use case adoptions. By the end of 2026 or early 2027, expect to see some corporates waving the white flag and pulling back.
All that matters is the bigger picture.
AI threw you a lifeline. It can take over the grunt work so you can focus on strategy. So, take advantage of it and solve bigger picture problems. You’re not needed in the rabbit hole anymore so use the lifeline to climb out of the low-level data crunching and conquer the strategy work.
Perhaps you are a data analyst, and you’re used to writing SQL and Python all day to pick apart various data sets. Now that AI can help you form most of your queries, what you will need to focus more on is the business context behind what you’re solving for. Designing experiments, validating the query you wrote answers the experiment you set out to run, and communicating your results with your colleagues will become your day-to-day. You become more of a conductor than an operator, orchestrating various machines to output a tailored response.y.
Find ways to get out of the weeds and design your workflow so that you can focus on the bigger picture problems to increase your company’s revenue. That way, the $1T implied AI savings will come partially in the way of revenue growth and employment won’t falter too much.