Cut the Hype: Andrew Ng’s Reality on AI’s Real Business ROI

Artificial intelligence (AI) is undeniably transforming the business landscape. From self-driving cars to chatbots, from predictive analytics to personalized marketing, the applications of AI are vast and growing. However, as Andrew Ng, a leading AI researcher, educator, and co-founder of Google Brain, has emphasized in recent talks—including his insightful keynote at the CrewAI Summit—there is a growing problem in the AI industry: hype. According to Ng, much of the excitement surrounding the latest AI technologies, particularly foundation models, is overshadowing the practical, ROI-driven applications of AI that can truly deliver value for businesses. While the buzz around the potential of large, generalized models is palpable, Ng stresses that focusing too much on these models can distract businesses from the real value that AI can bring—value that comes from targeted, strategic applications that solve specific problems.

In his speech, Ng called for a more measured, pragmatic approach to AI adoption—one that doesn’t chase after the latest trends or flashy promises but instead prioritizes practical, scalable applications. The key to unlocking AI’s true potential, he argues, is in using AI strategically, applying it where it can make the most impact and deliver measurable returns.

This blog explores Ng’s perspective on the current state of AI hype, and more importantly, how businesses can separate the buzz from the reality. Drawing from Ng’s insights at the CrewAI Summit, we’ll examine why the emphasis on foundation models might be misleading, why smart use cases drive ROI, and how even smaller companies can build powerful AI systems without the need for billion-dollar budgets.

Hype Distracts from Building Effective Agentic AI Solutions

Over the past few years, foundation models—large, general-purpose AI systems such as GPT-4, PaLM, and other variants—have taken center stage in the conversation around artificial intelligence. These models, often pre-trained on vast datasets and fine-tuned for specific tasks, have achieved remarkable feats in areas like natural language understanding, image generation, and decision-making. From generating coherent paragraphs to creating art, foundation models have captured the imagination of tech enthusiasts, investors, and even business leaders eager to jump on the AI bandwagon and adopt the latest innovations.

While these models have undeniably made impressive strides, Andrew Ng, one of the most influential voices in AI, has raised concerns over how they are marketed and adopted in business contexts. Foundation models, as they currently stand, are often presented as all-encompassing solutions that can address a vast array of challenges with minimal customization. The perception is that businesses must adopt these large-scale models to stay competitive in the fast-paced AI race. This narrative of “bigger is better” has been popularized by their dazzling feats and the ambitious promises associated with them.

However, Ng warns that the hype surrounding foundation models could lead businesses down a problematic path, one where the emphasis on model size and complexity obscures the more practical and actionable aspects of AI deployment. In other words, the race for adopting the latest and greatest AI technology can distract organizations from focusing on solutions that deliver genuine, measurable value.

The Misleading Promise of Foundation Models

The allure of foundation models comes from their apparent versatility. These models, capable of performing a wide range of tasks with little to no task-specific fine-tuning, are often marketed as silver bullets for businesses looking to implement AI. The promise is enticing: you can apply a single, large model to a variety of problems—from customer service automation to content generation—without the need for extensive retraining or adjustments.

While this promise sounds appealing, Ng cautions that it is misleading in several important ways. First, these models are not magic solutions; they require significant resources, both in terms of computational power and expertise, to implement effectively. Training these models from scratch or even fine-tuning them for specific tasks can be an expensive and time-consuming process, one that may not be necessary for all businesses.

Additionally, Ng highlights that foundation models often require ongoing maintenance and updates to ensure their continued relevance and accuracy. As these models are generalized, they are not always optimized for the specific use cases that businesses need. This results in a high level of inefficiency, as businesses must dedicate considerable resources to fine-tuning or supplementing these models with additional work, further complicating the overall deployment.

Rather than pursuing the “next big thing” in AI, Ng argues that businesses should focus on practical, tailored solutions that align closely with their goals. The hype around large foundation models should not overshadow the more grounded, effective approach of applying AI in a targeted, agentic manner.

Agentic AI: Focusing on Practical, Actionable Solutions

Ng advocates for a shift in focus toward agentic AI—AI systems designed with clear, actionable objectives that can operate independently to meet business goals. These solutions are often smaller, more focused, and far easier to deploy than sprawling, general-purpose foundation models. Agentic AI systems are designed to automate specific tasks or improve particular processes within a business, rather than attempting to address every problem under the sun.

For example, rather than using a massive model that can do everything from writing articles to answering customer queries, businesses could deploy a smaller, task-specific AI model designed for just one of those functions. This targeted approach allows the business to focus its resources more effectively, ensuring that the AI system provides concrete, measurable returns on investment without the need for extensive customization or high overhead costs.

Ng’s view is that businesses should prioritize developing or deploying AI systems that act autonomously, achieving specific business objectives efficiently. Whether that involves automating customer service workflows, enhancing data-driven decision-making, or personalizing marketing efforts, agentic AI can provide significant value with far fewer resources than a massive foundation model would require. By shifting the focus from grand, all-encompassing solutions to more practical applications, businesses can unlock the true potential of AI without getting bogged down in the complexities of foundation models.

Tailoring AI to Business Needs: Practical AI Deployment

Rather than obsessing over whether a business has access to the latest foundation model or pursuing AI tools that promise to solve everything, Ng emphasizes that businesses should build AI solutions that work within the scope of their immediate needs. This approach focuses on targeted use cases where AI can drive significant business value and deliver a tangible return on investment.

Here are a few examples of how businesses can deploy more practical, agentic AI solutions:

• Customer Service Automation: Rather than investing in a large, general-purpose model to handle every customer interaction, a business could implement a chatbot or virtual assistant tailored specifically to their customer service needs. This AI could handle frequently asked questions, resolve common issues, or assist in basic transactions, freeing up human agents to focus on more complex problems.
• Supply Chain Optimization: AI models can be deployed to optimize logistics, inventory management, and demand forecasting. Instead of a massive, all-encompassing system, a focused AI solution can analyze data specific to the business’s supply chain and make adjustments in real-time, saving costs and improving efficiency.
• Personalized Marketing: AI can be used to drive targeted marketing efforts, delivering personalized recommendations, product suggestions, or promotional offers to individual customers. A smaller, specialized model tailored to a business’s customer data can perform this task far more efficiently than a general-purpose model attempting to address a wide range of marketing needs.

By focusing on smaller, specialized use cases, businesses can reap the benefits of AI without the burden of massive investments in general-purpose models. These solutions are not only more cost-effective but also more aligned with a business’s specific goals and challenges. The focus on practical applications ensures that businesses can derive real, measurable results from their AI investments.

Smaller Firms Can Leverage AI Without Huge Budgets

One of Ng’s most important points at the CrewAI Summit is that smaller firms do not need billion-dollar budgets to harness the power of AI. The widespread availability of open-source models, cloud-based platforms, and pre-trained solutions means that even small businesses can deploy effective AI systems that are tailored to their needs, without the heavy computational costs associated with training large foundation models.

In fact, Ng argues that the democratization of AI—through more affordable tools and resources—presents an opportunity for smaller firms to compete with much larger organizations. By focusing on agentic AI solutions that target specific business needs, smaller firms can level the playing field and achieve significant results without needing the resources of a tech giant.

For example, many startups can leverage cloud-based AI services from major providers like Google Cloud, AWS, or Microsoft Azure. These platforms offer powerful AI tools that are scalable, affordable, and easy to integrate into existing systems. Instead of investing millions into training a massive AI model from scratch, smaller businesses can access ready-made solutions that address specific tasks, such as predictive analytics, customer insights, or process automation.

This accessibility has made AI much more attainable for businesses of all sizes. In Ng’s vision, AI isn’t just for large enterprises with vast resources—it’s a tool that can be harnessed by any company that knows how to apply it strategically.

Pragmatic AI Deployment Over Flashy Promises

In conclusion, Andrew Ng’s message at the CrewAI Summit was clear: businesses must cut through the hype surrounding AI and focus on pragmatic, ROI-driven solutions. While foundation models like GPT-4 and PaLM are exciting and have their place in the AI ecosystem, they should not be viewed as the only or the best path to AI success. Instead, businesses should focus on developing or deploying AI systems that are tailored to their specific needs—solutions that are smaller, more targeted, and easier to integrate into existing workflows.

By emphasizing agentic AI over massive, general-purpose models, businesses can build AI systems that deliver tangible value and measurable returns. This approach not only saves resources but also ensures that AI efforts are aligned with business goals, driving efficiency, innovation, and profitability. Moreover, the accessibility of AI tools for smaller businesses means that AI is no longer an exclusive domain of the largest firms. With the right focus and strategy, even small firms can harness the power of AI to gain a competitive edge.

Ng’s strategic framework for AI deployment prioritizes practical, focused solutions over flashy promises. By cutting through the noise and staying grounded in pragmatic AI applications, businesses can unlock the true potential of AI and drive real, sustainable growth.

ROI Comes from Smart Use Cases, Not Massive Model Training

At the heart of Andrew Ng’s argument is the fundamental idea that real ROI in AI comes not from the size or scale of the model, but from how effectively it is applied. The excitement surrounding large-scale models—such as GPT-4, PaLM, and other foundation models—often leads to a fixation on their impressive capabilities. Businesses and developers often focus on the number of parameters these models contain, how they perform on benchmark tasks, or how they can handle a wide range of inputs. While these attributes are certainly interesting, Ng cautions that they are not always reflective of real-world business value.

The widespread hype surrounding these massive models can create the illusion that bigger models automatically equate to better outcomes. Yet, Ng argues that this excitement is often misplaced. Businesses need to look beyond the flashy capabilities of large-scale models and focus on solving specific problems that will drive their bottom line.

The True ROI Lies in Effective Application

For businesses looking to unlock the true potential of AI, ROI doesn’t come from simply deploying the latest, most powerful AI model. It comes from the intelligent and effective application of AI to solve specific problems within the organization. The key to extracting value from AI lies in identifying the right use cases—areas where AI can make an immediate, measurable impact. This approach requires businesses to think strategically about where and how AI can fit into their existing operations, and how it can help them achieve their goals.

For instance, AI can be used to automate mundane, manual tasks like data entry, which can free up human workers to focus on more strategic, high-value activities. Similarly, AI can optimize decision-making processes, such as recommending pricing strategies based on market conditions or predicting customer behavior. AI can also be leveraged to enhance customer experiences—whether through personalized recommendations, dynamic content delivery, or tailored marketing campaigns. These are just a few examples of how AI can drive meaningful business impact.

Key Areas Where AI Can Drive ROI:

• Automation of Repetitive Tasks: Using AI to handle routine work, such as data entry or scheduling, can free up human resources for more valuable tasks.
• Enhanced Decision-Making: AI can optimize decision-making by processing large datasets and delivering insights for business strategies.
• Improved Customer Experience: By offering personalized recommendations or better customer service, AI can significantly enhance customer satisfaction and loyalty.
• Cost Reduction: AI can improve operational efficiency by identifying inefficiencies and optimizing resource allocation.

In these cases, it’s not about deploying the latest, massive model—it’s about using AI in a targeted way that addresses the unique needs of the business. Rather than sinking millions of dollars into the training of a vast general-purpose model, companies can often achieve better ROI by leveraging smaller, more specialized models that are tailored to specific tasks and optimized for the business’s particular challenges.

Focus on Small, Specialized Models

Ng emphasizes that the most effective and impactful AI solutions don’t always require the largest, most complex models. In fact, many businesses can achieve their best ROI by fine-tuning smaller, pre-trained models that are directly aligned with the specific tasks they need to accomplish. These models can often be implemented with far less investment in terms of time, resources, and computational power, allowing businesses to see faster, more cost-effective returns.

The key to success with AI is not the model’s size but its relevance and ability to meet the needs of the business. Instead of focusing on the sheer scale of a model, businesses should prioritize models that are specifically designed to solve their pain points. By leveraging pre-trained models or fine-tuning existing solutions to their own data and needs, businesses can create highly effective and efficient systems without the need for billion-dollar budgets or excessive model training.

Real-World Examples of Effective AI Deployment

Here are a few practical examples of how businesses can achieve significant ROI from smart, targeted AI applications without the need for massive model training:

1. AI-Driven Recommendation Systems

For e-commerce businesses, AI-powered recommendation systems are one of the most common and impactful applications of AI. These systems don’t need to be based on the largest foundation models. By fine-tuning smaller, more focused models on a company’s specific data—such as past customer purchases, browsing history, or search behavior—a business can create a highly personalized recommendation engine. This system can boost sales by delivering relevant product suggestions to customers, improving their shopping experience and increasing conversion rates.

2. AI-Powered Chatbots for Customer Service

AI-powered chatbots have become a cornerstone of modern customer service. These chatbots can handle a wide range of customer inquiries, from answering basic questions to assisting with order tracking. While large, general-purpose models like GPT-4 could theoretically power these chatbots, many businesses can achieve excellent results by using smaller, specialized models that focus on customer service queries. These models can be trained on specific customer interactions, allowing them to handle a higher volume of requests while maintaining a high level of accuracy. Additionally, they can help reduce operational costs by automating responses to common inquiries, thereby freeing up human agents for more complex issues.

3. Predictive Analytics for Business Decisions

In fields such as finance or supply chain management, predictive analytics powered by AI can help businesses make better decisions. Rather than relying on the latest foundation model, businesses can use smaller AI models to analyze historical data and predict future trends. For example, a retailer can use AI to predict demand for specific products, helping to optimize inventory levels and reduce waste. Similarly, AI can be applied in financial sectors to predict stock trends or customer credit risk, enabling more informed decision-making and reducing financial risks.

In each of these cases, smaller, specialized models tailored to the specific needs of the business drive significant value. They are more cost-effective and quicker to deploy, while still delivering impressive results in terms of sales, customer service, and operational efficiency.

The Smart Path to ROI

Ng’s approach to AI adoption is rooted in pragmatism. He believes that AI is not a one-size-fits-all solution, and businesses should avoid the temptation to chase after the latest trends and models simply because they are the most powerful or talked about. Instead, companies should look for practical applications that can solve specific business problems in a focused, efficient manner.

The real ROI of AI doesn’t come from the sheer power of the model, but from how well it addresses the unique needs of a business. By focusing on solving specific problems—whether it’s improving customer satisfaction, automating repetitive tasks, or optimizing operational processes—businesses can achieve measurable improvements that have a direct impact on their bottom line. AI should be viewed not as a magic bullet but as a powerful tool that can be strategically deployed where it will make the most difference.

Ng’s message is clear: businesses need to stop chasing after the hype of massive models and instead focus on the smart use cases where AI can deliver tangible results. By doing so, they can create AI systems that drive true business value without getting bogged down by complexity or excessive investment.

Smaller Firms Can Access Cheap, Powerful Agents Without Billion-Dollar Budgets

One of the most powerful takeaways from Ng’s talk at CrewAI Summit is his insistence that smaller firms are not at a disadvantage when it comes to adopting AI. The hype surrounding massive model training and foundation models can give the impression that only well-funded organizations have the resources to harness the power of AI. However, Ng points out that smaller businesses now have access to cheap, powerful AI agents that can be leveraged for a variety of use cases.

Advances in AI, particularly in areas like transfer learning and open-source models, have made it possible for smaller companies to create powerful AI solutions without needing the vast computational resources required for training a foundation model from scratch. With tools like pre-trained models, cloud-based AI services, and accessible development platforms, even small companies can build AI systems that meet their needs, without the need for billion-dollar budgets.

In Ng’s view, this democratization of AI is one of its greatest strengths. Small firms can now compete with larger, well-funded competitors by focusing on practical applications and leveraging pre-trained models that can be fine-tuned for specific tasks. For instance, a small startup in the healthcare space can use pre-trained AI models to analyze medical data or assist in diagnostic processes without the need for massive infrastructure investments. Similarly, small businesses in retail or hospitality can use AI to improve customer service, optimize pricing, and personalize marketing campaigns without the need to develop large-scale models.

Ng’s message is clear: AI is no longer just for the largest companies with the deepest pockets. With the right approach, smaller companies can tap into the power of AI and realize meaningful returns on their investments, even without a multi-million-dollar budget. The focus should be on identifying the right use cases and deploying AI agents that are designed to solve specific problems—without getting lost in the hype surrounding model size or complexity.

Pragmatic Deployment > Flashy Promises

Andrew Ng’s insights at the CrewAI Summit offer a much-needed dose of reality for businesses looking to leverage AI. The hype surrounding massive foundation models and their transformative potential can be distracting and ultimately counterproductive. While these models certainly have their place in the AI ecosystem, Ng stresses that the real value of AI lies in its practical, ROI-driven applications. Instead of focusing on the flashiness of the latest models, businesses should prioritize pragmatic deployment that addresses specific needs and delivers measurable results.

In Ng’s strategic framework, success in AI comes from understanding the true capabilities of the technology and using it where it can have the most impact. AI productivity tools and powerful agents can drive significant returns without requiring massive model training or enormous computational resources. Smaller firms, too, can harness AI to their advantage, leveling the playing field and competing with larger organizations. The key is to stay grounded in practical use cases and focus on creating AI solutions that are tailored to the business’s specific goals.

Ultimately, as Ng cautions, the hype surrounding AI should not overshadow the pragmatic, thoughtful approach that businesses need to take. AI’s real business ROI lies not in flashy promises or the latest models, but in the smart, strategic deployment of technology that solves real-world problems. By cutting through the hype and focusing on practical applications, businesses can unlock the true potential of AI and drive long-term success.

Works Cited

CrewAI. (n.d.). CrewAI Summit. Retrieved from [Insert full URL here, e.g., https://www.crewai.com/summit] 

DeepLearning.AI. (n.d.). The Batch: Andrew Ng’s Weekly Newsletter. Retrieved from https://www.deeplearning.ai/the-batch/ 

Klover.ai. (2025, June 11). Vibe Coding – AI-Assisted Software Development. Retrieved from [Insert full URL here if available from your original search, e.g., https://www.klover.ai/vibe-coding-ai-assisted-software-development/] 

Landing.AI. (n.d.). Andrew Ng. Retrieved from https://landing.ai/andrew-ng 

Klover.ai. “AI Productivity & Laziness: Ng on Empowering Developers.” Klover.ai, https://www.klover.ai/ai-productivity-laziness-ng-on-empowering-developers/.

Klover.ai. “Why Vibe Coding Doesn’t Mean Less Work: Andrew Ng’s Reality Check.” Klover.ai, https://www.klover.ai/why-vibe-coding-doesnt-mean-less-work-andrew-ngs-reality-check/.

Klover.ai. “Andrew Ng.” Klover.ai, https://www.klover.ai/andrew-ng/.