Summary of Key Points
This article exposes five common issues in the current corporate AI industry through interviews with former Smart Spectrum COO Zhang Fan and the author's own observations: using flashy demos to deceive customers, FDE (Frontline Deployment Engineers) teams that fail to live up to their name, bosses blindly investing in AI based on wishful thinking, treating AI implementation as a purely technical issue, and rushing to use AI for layoffs. The article emphasizes that the reality of corporate AI is that “building it doesn’t equal making it work effectively.” True value lies in continuously solving business problems after implementation, not just in the impressive displays during demonstrations.
Detailed Analysis
1. **Demos are impressive, but performance in real life is poor – Don’t be fooled by ‘showy AI’**
It’s too easy to create an AI demo nowadays: write some prompts and preset a few questions, and the AI can provide perfect answers, often leaving first-time viewers in awe. However, real business scenarios are not like rehearsed scripts—data may be incomplete, rules may conflict, and user issues can be unpredictable (for example, customers might ask tricky questions that weren’t anticipated in the demo).
For instance, some companies spend millions on creating a demo only to find out that the AI can’t even handle basic business processes, leading to its abandonment. The author advises that when considering AI solutions, companies should not rely solely on demos; instead, they should visit customers who have already implemented the technology to see if it’s still being used and how errors are resolved, as well as whether business metrics have improved.
2. **FDE is not a ‘panacea’ – They’re essentially just traditional implementation consultants with a new name**
Many vendors claim to have FDE teams that can diagnose and implement AI solutions. In reality, most FDEs are just traditional implementation consultants with some knowledge of AI terminology. Their role remains the same: organizing requirements, processing data, and customizing software, which does not truly address business problems.
Zhang Fan used a vivid analogy: “It’s like using more efficient engines to feed horses, not building a car itself.” This means that FDEs don’t change the underlying logic of the business; they merely use AI tools for superficial optimization. As a result, FDEs get bogged down in customization, customers fail to see value in their services, and vendors struggle with revenue collection.
3. **Bosses’ wishful thinking about AI – More harmful than the technology itself**
Vendors often boast about how AI can increase customer revenue by tenfold, leading bosses to fantasize that investing $1 million could result in a $1 billion return. This is wishful thinking, and 90% of AI projects fail due to such illusions.
The author advises bosses to ask two key questions before investing: what are the conditions for success (e.g., are there high-quality data available? Does the team have AI expertise?) and what are the costs (how much time and resources will be required)? They should also ask if the vendor can actually achieve such results; only those who can provide concrete commitments are credible.
4. **AI implementation is not a technical issue – First, figure out what business problems you want to solve**
Many bosses rush to have their CIOs develop AI solutions after being impressed by success stories. However, the real challenge in implementing AI is understanding the business and making decisions. For example, using AI for sales might involve automating simple processes or assessing business opportunities, which requires data collection, organizational coordination, and structural changes—these are much more difficult.
The author emphasizes that bosses need to get involved and clarify three critical questions: what business problems need to be solved, what goals should be achieved, and how much they are willing to invest. Focusing on products without a clear plan from the beginning sets the project on the wrong path.
5. **Don’t rush to lay off employees – AI needs human support**
Bosses often think that AI can replace all manual tasks, but AI is best suited for low-risk, standardized jobs (e.g., data entry). For complex tasks, such as handling customer inquiries or making business decisions, human expertise is still necessary.
For example, some companies lay off employees too soon after implementing AI, only to find themselves in a position where they need to hire new staff when issues arise, at a higher cost. A more prudent approach is to have AI work alongside business experts, with the experts monitoring and managing AI operations until it becomes reliable before making any major organizational changes.
Conclusion
The era of corporate AI has just begun, but success depends on practical implementation rather than impressive rhetoric. Before launching an AI project, ask yourself these five questions:
1. What specific business problem does this AI solution aim to solve?
2. How does the real business environment differ from the demo?
3. What are the prerequisites and costs for a successful implementation?
4. Does the vendor’s FDE team have the necessary capabilities? Can you interview them in advance?
5. What are your business goals, and can the vendor clearly articulate them in the contract?
These questions may not sound exciting, but they can help you avoid common pitfalls. After all, the true value of AI lies in its ability to continue to perform effectively even after the boss is no longer involved.