Must-read for Executives: Why Do AI Projects Always Fail? It's Because You Haven't Done These 5 Things Right

Sam Keller didn't initially start a corporate training program. At first, he was just researching generative AI with his 12 - year - old son at his kitchen table. Later, this small research gradually evolved into the Gen AI Academy, which has now trained over 135 CEOs and their leadership teams. His mission is to help organizations smoothly enter the AI era through hands - on and experiential learning.

Original title: "Five Imperative Actions for Executives to Achieve AI Success"

"It only takes a few hours to bring about change," Sam said. "After executives customize AI according to their own positions, soon, they start thinking more critically, making wiser decisions, and becoming more creative in their work."

After participating in one of Sam's workshops, I personally experienced this change. After the event, I deeply realized that executives not only need to understand AI but also regard it as a thinking partner and use it to become better leaders.

Here are five actions that every executive should take.

1. Treat AI as a Team Member, Not Just a Tool

When a company recruits new employees, it arranges onboarding, sets goals, and assesses results. AI also requires the same management approach.

Onboarding AI means providing it with company background information, defining its scope of responsibilities (what it can and cannot do), and checking its output early on. You wouldn't let a new employee report directly to the board without supervision, and similarly, you shouldn't assign critical tasks to AI without debugging it first.

The most effective way is to create AI agents, which are dedicated assistants for different professional fields. For example, you can build them for different functions such as marketing and finance, or design agents based on customer or user profiles to conduct "stress tests" on proposals before major meetings.

A colleague of mine set her AI agent to have a rather straightforward style. When she asked the AI to design an icon for a presentation, it retorted, "Cut the crap and clarify your purpose first." Although its strong - willed personality was unexpected, the feedback it gave was precise.

A real - estate executive in our group trained a GPT with data and industry materials from CBRE. Insights that used to take days to obtain can now be generated in minutes. His reaction was, "It's amazing!"

The lesson here is that AI can deliver maximum value when leaders treat it as a colleague that needs guidance and cultivation, rather than a tool to be casually tested.

2. Make AI Literacy a "Must - Have" for Leadership

Many CEOs don't fully realize their leading role in promoting AI adoption. In fact, the active participation of leadership is the lever that can move everything.

As Steve Giondomenica, co - founder of the GenAI Academy, said, "The questions CEOs ask are all reasonable, such as: 'How important is this really? Where should it fit in the technical architecture? What are IT's responsibilities? How high is my security risk?' These questions are understandable, but they often hold CEOs back."

Therefore, cultivating AI literacy must start from the top. Personal experience can shift the narrative from "worrying about risks" to "exploring possibilities." Sam Keller shared a case: In a four - day leadership training program, a half - day AI training workshop was arranged on the first day. The participating CEOs said that this training directly changed all subsequent discussions: "This has enhanced my understanding more than any previous experience."

In a construction company, the CEO didn't just stop at training. He assigned a person in charge to specifically look for AI application opportunities. In just a few weeks, the team found that the marketing department could use AI to write proposals. Previously, this task took up over 70% of their time.

To reach a consensus, it doesn't take months; sometimes, a few hours are enough.

3. Focus on Practice to Develop AI Proficiency

When leaders integrate AI into daily work, such as morning meetings, strategic meetings, or even company celebrations, the company's acceptance of AI will accelerate significantly.

In my company, we developed some fun AI agents, including a sarcastic "teammate robot" and customized entrance music for speakers using Suno.ai. We even organized several ice - breaking activities where the team created AI superheroes to "combat" the most troublesome work problems. These fun moments not only bring joy but also encourage everyone to try and explore together.

Some teams also use AI in more important scenarios. Sam mentioned that a company used a customized GPT agent to simulate a difficult customer. This model accurately predicted the customer's possible objections and helped the team recover overdue payments within a few days.

Constantly discussing AI can eliminate the prejudice that "using AI is cheating" and redefine it as "a tool to enhance the depth and reliability of work."

4. Build an Ecosystem for AI Applications

Gartner predicts that by 2028, over 95% of enterprises will use generative AI interfaces, generative AI models, or generative AI - enabled applications in their production processes.

While the industry focuses on the technology hype cycle, we should pay more attention to the infrastructure that supports AI development.

In the process of promoting AI applications, the following key questions need to be clarified:

• What data can be safely input into AI?

• What tools can new employees use?

• Where should AI fit in the workflow?

Even a rough guiding framework can set boundaries for AI applications in work and ensure the consistency of workflows. Without this foundation, all efforts will remain at the scattered pilot stage and cannot be scaled up.

5. Set a Rule Framework and Then Let Go Appropriately

Sam's survey of over 135 executives shows that their concerns about AI are highly consistent: data security, accuracy, and credibility. The most frequently asked questions by executives include:

• Where should I start? - Start from the top and lead by example.

• Can AI be trusted? - Yes, on the premise of having governance mechanisms and transparency.

• How to deal with the team's resistance? - Once people see the direct value of AI, the resistance will fade.

• How to achieve scale - up? - Appoint a person in charge, record application cases, and invest in infrastructure.

In reality, we can't quickly foresee all problems and design perfect solutions. This isn't a matter of "solving all problems at once and then implementing." The best thing we can do is to set a clear rule framework while leaving room for improvement where "today's wisdom surpasses yesterday's."

As Sam said, "The trend is very clear. The real challenge is, how can we get everyone to keep up?"

The Compound Effect of AI

Think of AI readiness as a compound - interest investment: the earlier you invest, the faster the returns accumulate.

Sam pointed out, "The so - called interest rate is extremely high. If some companies start taking action a few months earlier and keep learning, other companies may never catch up."

AI transformation isn't about chasing the coolest tools but about the combination of leadership, learning ability, and development potential. Leaders who start investing now can not only save time but also build a competitive advantage of "daily compounding."

This article is translated from:

https://www.forbes.com/sites/triciaemerson/2025/09/23/5-actions-every-exec-should-take-for-ai-success/

This article is from the WeChat official account "Forbes" (ID: forbes_china). Author: Tricia Emerson, Translator: Björn & Rach, Proofreader: Lemin. Published by 36Kr with permission.