The AI industry is witnessing a dramatic shift in values: users are no longer captivated by “parameter-heavy beasts,” but are instead seeking “application partners” that provide real, human-centered value.
GPT-5: A Performance Champion Facing a Reputation Crisis
In early August, OpenAI unveiled ChatGPT-5, sparking huge excitement across the industry. While the company did not disclose the exact parameter size, data from the launch event and independent testing platforms confirmed a massive leap in performance. GPT-5 could generate nearly 400 lines of high-quality code in under two minutes, reduce hallucination rates by 80% compared to its predecessor, and top the neutral evaluation platform LMArena with record-breaking scores across text comprehension, coding, and vision tasks.
Armed with such impressive capabilities, OpenAI discontinued several older models, expecting users to fully embrace GPT-5. But within just two days, backlash erupted. On social media, user complaints focused on two major issues:
- Unstable performance – GPT-5’s new auto-routing system often redirected queries to lightweight variants. This resulted in inconsistent answer quality, slower responses, and frequent errors even in basic tasks, creating a gap between marketing hype and real-world experience.
- Lack of empathy – Even paying Pro users reported that GPT-5, while smart, delivered cold and overly concise answers, lacking the warmth and conversational tone of GPT-4o.
The dissatisfaction was so widespread that OpenAI was forced to bring GPT-4o back online to appease its community.
When Parameter Obsession Ignores User Needs
This “bring back GPT-4o” campaign highlights a deeper contradiction: if GPT-5 is a benchmark leader powered by trillions of computations, why did users still prefer its predecessor? The incident underscores a growing misalignment between technical ambition and user expectations.
For years, the industry has equated AI value with parameters, computing power, and benchmark scores. This performance-first mindset has created the illusion that “bigger is always better.” But user reactions to GPT-5 make it clear: AI’s value cannot be reduced to cold numbers alone. Experience, usability, and empathy matter just as much—if not more.
What Enterprises Can Learn: Building “Useful AI”
This disconnect between technical advancement and real-world usefulness is not unique to consumer users—it is even more critical for enterprises.
According to Accenture’s 2025 China Digital Transformation Index , while 46% of Chinese enterprises have scaled AI adoption, only 9% have realized significant business value.
For individuals, switching back to an older model is effortless. But for enterprises investing heavily in AI transformation, a failed deployment can mean wasted capital, lost time, and weakened morale. The pressing question is: How can businesses choose AI that truly drives value?
The answer lies in mindset. “How you think” is often more important than “what you do.” Enterprises must first avoid common pitfalls, then focus on execution strategies that align AI with business needs.
Avoiding Pitfalls: Aligning with Real Demands
IDC’s Enterprise AI Transformation Playbook outlines three common mistakes that organizations should avoid:
- Mistake 1: Equating parameters with outcomes
Bigger models don’t always mean better results. Businesses should evaluate whether AI capabilities match their real-world scenarios. - Mistake 2: Treating launch as the finish line
AI requires continuous optimization. Enterprises must budget for long-term data refinement and model updates to ensure lasting impact. - Mistake 3: Confusing “closed and conservative” with “safe and reliable”
Locking data away for fear of risk diminishes its value. True security comes from choosing a trusted, end-to-end solution that enables secure data flow while safeguarding privacy and compliance.
Focusing on Execution: Turning AI into Growth
Once the fog of misconceptions is cleared, enterprises can unlock AI’s real potential. Accenture highlights four key priorities:
- Innovate to Break Boundaries
Integrate AI systematically into business models, strategies, and workflows to accelerate new products and services, gaining a competitive edge. - Build an AI-Driven Digital Core
Upgrade legacy IT into a system capable of sensing, decision-making, and self-evolution. With flexible architectures, enterprises can localize deployments across regions, regulations, and industries. - Enhance Adaptive Resilience
With modern data governance and AI monitoring, enterprises can optimize costs, improve responsiveness, and strengthen supply chain agility to mitigate risks. - Reshape Talent and Organizations
AI adoption requires not just tools, but new ways of working. Enterprises should enable human-AI collaboration, foster flexible talent models, and build trust in intelligent systems.
Conclusion: From Technical Obsession to Value Rationality
The GPT-5 controversy is ultimately a wake-up call: AI’s worth lies not in its raw intelligence, but in its ability to create real value. For enterprises, avoiding the trap of performance worship and grounding AI deployment in business context is key.
Technology is like a seed—only when planted in the soil of real-world applications can it grow into fruits of business value.
Note: This article is translated from https://www.sangfor.com.cn/news/19922caa95194ed4aa13f3b120890712
References:
- Accenture, 2025 China Digital Transformation Index https://www.accenture.com/content/dam/accenture/final/accenture-com/document-4/FY25-Accenture-China-Digital-Transformation-Index-Full-Report-Chinese-V2.pdf
- IDC, Enterprise AI Transformation Playbook https://www.volcengine.com/docs/6624/1393082
- Accenture, Why Only 9% of Chinese Enterprises Unlock Real AI Value https://mp.weixin.qq.com/s/0AK_1b0LN__zc1ELU6hyDA
