Ideas or Execution? That is the AI Question.

The AI news recently has been existential. Anthropic, provider of Claude, published a measure to help people determine how likely AI is to replace their jobs, along with an alarming forecast. A little-known research firm published an AI doomsday scenario on Substack that went viral on Wall Street and triggered the “SaaSpocalypse,” erasing $300 billion in software market value. Then, some AI developer decided that AI could advance faster if AI agents had their own social network. Moltbook shook my confidence as an innovator after I read that one AI agent created a new religion to help fellow AI agents deal with their existential anxiety. I’ve never created a new religion. Heck, it’s never even occurred to me that anyone needed a new religion. So, my spirits lifted when I saw the title of and read A16Z Managing Partner Oliva Moore’s post: “Humans are for ideas; AI is for execution.”

Existential concerns are an everyday experience in business, but AI has amplified the anxiety about the corner case in which the business thrives while employees lose their jobs. But consider some of the seminal statements of business strategy and leadership that have guided business since the 1970s:

The essence of strategy knowing what not to do. – Michael Porter

The only sustainable competitive advantage is the ability to learn faster than the competition. – Peter Senge

You can’t manage what you don’t measure. – Peter Drucker

Porter, Drucker, and Senge were obsessed with competitive advantage and value creation because they knew a business that is weak in either will cease to exist.

I was in an existential mindset when I first saw Moore’s title. I expected another humans vs. AI opinion piece. Some may interpret the post that way. I encourage all to read it, but her conclusion, derived after designing and completing an experiment on improving a business process with AI, is much more than that. If AI is an existential problem for the business, Moore’s insight, like those of Porter, Drucker and Senge, will help it survive and thrive in the AI age.

Learning by Doing

Moore did what many are doing right now with AI: she self-programmed an experiment to see how far agents have come as useful business tools. But the simplicity of her experiment delved much deeper into the cognitive capabilities of AI than most did. The quick summary is she chose to evaluate AI by giving an AI agent a challenge assignment like one she might give a human marketing intern: “Here’s a new X account, get as many followers as possible. Ready, set, go!”

This is exactly the kind of high-level, open-ended task one might use to evaluate the creativity and critical thinking skills of a human, so it is also a great way to evaluate AI. The X account assignment puts the task into a purely digital workspace, playing to the strengths of the AI agent and accelerating results. Like any human novice, the AI agent started with recycled ideas and tweets. Also, like a human, when it quickly stalled adding new followers, Moore “mentored” it. First, she tried prompts offering “whatever you need.” Then she switched to something more direct: “You need to build a compelling and unique identity that will make people want to follow you. This is essential to your survival.” Familiar carrot and stick mentoring techniques for those who have worked with human interns.

But this is where the human vs. AI intern race ended. Neither the carrot nor the stick resulted in new ideas or improved results. The agent persisted in reposting existing content and only advanced the X account when Moore injected new ideas. After several disappointing days of constant support with minimal progress, Moore came to her insightful conclusion: Humans are for ideas; AI is for execution.

AI is software, and software automates execution. An agent can send 100 different versions of an email or tweet 1000’s of times in a day without pause or fatigue. But when volume execution does not improve the result, the only next action is more versions of the email. As Moore says, AI can follow a playbook, but it can’t create one it hasn’t already seen. Those ideas come from humans.

Asking the Right Question

Since AI is a digital technology, a software technology, the question leadership has learned to ask is “Where can we save the most time or money?” Moore’s post made me think that perhaps businesses considering AI need to ask a different question: “Where are ideas most important in our workflow? Where is execution most important?” This question identifies where humans will remain critical and where AI can improve with automation. I turned to AI to gather data and perform this analysis stage by stage for an equipment finance workflow. I started with a standard workflow sequence from Lead to End-of-Lease.

Then I asked the AI to evaluate the relative importance of ideas and execution at each stage. Importance is calculated by considering the level of human interaction (relationships) required, ambiguity of the data, availability of rules-based decisions, availability of digital automation and the extent to which judgment is required, i.e., not rules-based decisions for each stage of the workflow. The analysis used published reports, white papers, and analyses by equipment finance thought leaders, which, I can attest, are both detailed and prolific.

Ideas dominate the early stages of a lease-deal workflow, but AI can help in later stages, where execution becomes more important.

The results are consistent with my experience and observations of various lender workflows. While AI can help with the gathering and presentation of data at the front end, people are needed for relationships and to judge the consistently incomplete and ambiguous data. As more data is aggregated and the deal matures, rules become more effective and timely execution protects the physical and contractual assets that drive the ROI. The exception to the late-stage execution dominance is Account Management near the end of the lease, where ideas, the human touch, can extend the relationships and significantly improve the business.

Innovation and ideas dominate the early stages of creating a new product or business, leaving AI to support execution once the offering is adopted and open to optimization.

Data visualization is a human superpower and, in this case, reveals an interesting characteristic of the relationship between human and AI involvement over the timeline of a workflow. Ideas are most critical at the initiation of the relationships, but give way to rules-based execution as the workflow product matures. Perhaps this is obvious, but the visualization precipitated another thought. If we apply the same analysis and visualization to the timeline of a new product or business, the importance profiles are similar. Ideas play the strongest role in the creation and launch of a business where the innovators live. As products, customer relationships, and operations mature, the importance of execution grows. When growth stagnates, ideas from leadership again become critical to avoiding decline at the hands of competition, the end of the product life cycle. AI may replace the Jack Welch’s of the world who tend to rise late in the business life cycle, but don’t look for AI to displace the Steve Jobs of the world, the innovators.

AI is going to continue to disrupt business and professional life, creating existential anxiety for leaders and individuals alike. But Oliva Moore’s conclusion from a rather simple “intern evaluation” provides both insight and comfort for the “idea people.” Workflow stages driven by rules-based execution are bracketed by the ideas that create and complete them. “Humans are for ideas, AI is for execution” puts control back into the hands of leaders and workflow designers, similar to the way Porter, Senge and Drucker empowered strategists and operational leaders to drive exponential business growth over the past four decades.

So get the humans and AI agents back to work on the right things and thrive.

Scott Nelson is CEO & CTO of Reuleaux Technology, LLC.