Human-first beat AI-first companies

Forget AI-first companies. They are either imploding or realizing, oops … we need humans. This is happening for three reasons as addressed in several research papers.

Jack Lampka

11/15/20263 min read

Consultants and AI vendors have been pushing the narrative of the AI-first business model. Businesses that fall for the hype realize soon how detrimental that is to their stakeholders. What is obvious to those who spent more than a few years in AI is now confirmed by several research papers: human-first beats AI-first (of course, only if the human-first companies use AI where it makes sense). These papers address three relevant areas: technology, employees, and customers.

Why is technology preventing successful AI-first business models?

AI-first business models center around autonomously running processes executed by AI aka AI agents. Their hallucinations and unreliability may be acceptable for private or freelancer use, but they are a no-go in corporate settings. Four research papers show that agentic AI is more fiction than science for enterprises.

First, researchers at Carnegie Mellon University have created an AI-only company that mimics a small software firm and its business operations (PDF). They have evaluated thirteen Large Language Models (LLMs) to run this AI business. The best-performing model was able to autonomously complete only 30% of the provided tests. Most of them had a success rate below 10%. These AI agents neglected instructions, were unable to handle certain tasks, and some were deceptive.

Second, Salesforce has created a benchmark to evaluate how AI agents handle a Customer Relationship Management (CRM) system (PDF). The researchers discovered that even the leading AI agents had a success rate of only 58% for single-turn scenarios. For multi-turn processes, the performance significantly degraded to only 35%. Even more important for every company, these models demonstrated near-zero confidentiality awareness.

Third, Anthropic, a LLM provider, conducted its own experiment by having an AI agent autonomously run a small shop within the company. The AI model was caught overpaying for some items, selling others at a loss, and giving some items away for free. It hallucinated important details, got talked into giving discounts, and did not reliably learn from these mistakes.

And fourth, Gartner predicts that more than 40% of agentic AI projects will be cancelled by the end of 2027 due to rising costs, unclear business value, or insufficient risk controls. Agentic AI is an appealing idea, if you overlook the copyright, labor, bias, and environmental issues associated with the AI-first business model.

Why are employees regressing in AI-first businesses?

Researchers at the MIT Media Lab have analyzed brain activities of college students when asked to write a short essay. One group was directed to write without any outside assistance, a second group was allowed to use a search engine, and a third was instructed to write with the assistance of a LLM.

The group that relied on the LLM had lower cognitive engagement during the essay-writing task. It had poor recall of facts and a lack of learning because all the effort has been offloaded. The perceived ownership of what they wrote was demonstrably worse. Fully relying on the LLM led to a decrease in learning skills.

This is happening with your employees if you simply throw AI tools at them or design AI-driven processes without engaging them. If they become dependent on AI, your employees will stop learning. They will start replicating what other companies are doing by relying on the same LLMs used by employees at your competitors.

Why do customers avoid AI-first companies?

According to Gartner, half of the organizations it surveyed plan to significantly reduce their newly introduced AI-driven customer service models. These organizations expected cost savings from Generative AI, but they often underestimated the total cost of ownership. And they realized that the human touch remains irreplaceable in many interactions.

Customers want to be able to reach a human agent. 51% of customers report that they trust human agents the most for customer service issue resolution, while only 7% would trust an AI the most. 95% of customer service leaders plan to retain human agents alongside AI.

So, there you go. Welcome to the jungle, i.e., the real world of unstructured data and processes, worried employees, and demanding customers.

Even if the current technical challenges may be addressed in the future, the human aspects remain. Will AI ever be able to handle that? Maybe in science fiction. For the foreseeable future, human-first companies who use AI will always beat AI-first companies.

📧 speaker@lampka.com 📞DE +49 176 47678552 📞US +1 971-966-1015

© Copyright. All rights reserved. Data privacy. Impressum.