Synfinner

One of the most exciting advancements in modern computing is the current development iteration of Large Language Models (LLMs) and Artificial Intelligence, particularly the current iterations of LLMs. We're also beginning to see LLM service providers integrated into many new products, both physical and software. In fact, we even see some of the common trademarks of a good product: 

• community engagement
• executive hype
• news coverage
• tech reviews
• positive and negative sentiment
• product integration
• competition

With that all said, we're starting to see articles, such as that from 9to5mac, that "Most iPhone owners see little to no value in Apple Intelligence so far". 

I chose Apple as a good example for this article, given their existing product use and market share. Specifically, there are ~150 million iPhone users in the USA and a global footprint of ~1.38 billion. Even if 100,000 of US users were to have updated models, running the latest iOS, and enabled Apple Intelligence, there would be a decent sample size of users; though I suspect the number is much larger. SellCell sampled Android and iPhone users and found that:

• "73% of iPhone users and 87% of Samsung users say AI features add little to no value, highlighting widespread apathy across both brands."
• "Two-fifths of AI-supported Apple iPhone owners (41.6%) responded ‘yes’; 58.4% said they had not used any of the new Apple Intelligence features"

I understand the argument that these features are new and, in some cases, being rolled out gradually. As they become more widely available, we can hopefully expect increased positive sentiment and adoption. It’s true that there are AI tools for generating music, voices, videos, and images. These models, which enable greater creativity and expression, are undoubtedly impressive. However, we’re left questioning how many people are using them for purposes beyond generating memes, satire, or adult content. I’d argue that the number is relatively low compared to the millions of active monthly/daily users. That said, this observation is purely subjective and not based on concrete data.

Let’s return to the original intent of this post. The explosion of LLM products is truly remarkable and exciting. That said, many of these products seem to be solutions in search of problems. Nevertheless, some are indeed finding meaningful applications, such as assisting with text content proofreading and generation. While NLP has been around for decades, the advanced capabilities of LLMs—combined with their accessibility and ease of use—have brought them to a much broader audience, enabling widespread adoption and new use cases. However, there remains a significant flaw: the audience.

Any product release must consider its intended audience, or consumer. In marketing, this is referred to as target audience analysis. This process involves identifying specific groups of consumers who are most likely to be interested in a product or service. Factors such as demographics, psychographics, behaviors, and needs are all taken into account during the analysis. But what happens when a product is designed to provide access to intelligence about virtually anything? Who, then, is the intended audience? The answer: anyone who is interested. 

While I’m neither a seasoned salesperson nor a marketing expert, this approach seems like an uphill battle. To succeed, a company must create byproducts or complementary features to engage the broader audience encompassed by this undefined “anyone". This is where product integrations come into play. Examples include Apple Intelligence, Microsoft’s Copilot, Google Gemini’s AI-powered search overviews, Amazon’s AI-enhanced search, and ChatGPT. These integrations serve as accessible touchpoints that demonstrate the product’s value, helping to capture the attention of diverse users and build relevance in their daily lives.

There are some hard truths that need to be acknowledged:

• Not everyone spends their day working in Excel sheets.
• Not everyone relies on Word documents for their tasks.
• Not everyone spends their time searching for content online.
• Not everyone is in a position to use chat systems to gather data or knowledge.
• Not everyone is a programmer.
• Not everyone is striving to build the next groundbreaking innovation.
• Not everyone uses computers or mobile devices for work.

These “hard truths” may seem discouraging, but they represent critical challenges that AI and large language model (LLM) services must address to reach their full potential. The tech industry often operates under the assumption that technology is universally integrated into every aspect of life and work. This assumption creates a gap between the development of cutting-edge tools and the practical needs of diverse user groups. As someone who works in tech, I share the dream of a world where technology seamlessly enhances all facets of life. However, this vision remains far from reality for many, particularly those in industries or roles where technology plays a minimal or indirect role.

Even in industries that rely heavily on technology for day-to-day operations, cutting-edge tools and features are not always widely adopted. Instead, workers predominantly depend on well-established, foundational software like Excel, Word, Outlook, and other popular suites. These tools are deeply ingrained in workflows because they are reliable, familiar, and sufficient for the tasks at hand. While advanced features and AI-driven solutions offer exciting possibilities, they often remain underutilized due to a lack of accessibility, training, or immediate relevance to users’ needs. The challenge, therefore, is not just about creating innovative solutions but also ensuring they are practical, approachable, and meaningful to those who rely on them.

When examining data from the U.S. Bureau of Labor Statistics (BLS), it becomes clear why many business software integrations are targeted toward certain sectors. As of 2023, the industries with the largest employment numbers include:

1. Professional and business services: 22,840,100 workers
2. Healthcare and social assistance (private sector): 21,524,500 workers
3. State and local government: 19,856,600 workers
4. Leisure and hospitality: 16,592,800 workers
5. Retail trade: 15,590,100 workers

[source]

Naturally, businesses focus their technological efforts on these sectors because they represent the largest potential user bases. A key commonality among these industries is that they are data-intensive, making them ideal candidates for the capabilities of large language models (LLMs). LLMs excel at handling and manipulating data, positioning them to play a transformative role in these fields. However, much of this transformation is likely to occur behind the scenes. Over time, most LLMs will become virtually invisible to employees, seamlessly integrating into background processes. For instance, they could filter or rank data, extract relevant text for specific tasks, highlight key points for stakeholders, analyze medical imaging data, and much more.

That said, this focus does not account for the many subdivisions and jobs both within and outside these industries where there is little to no need for such technology. Additionally, while some businesses and roles are actively undergoing modernization, others face significant barriers. For some businesses and jobs, the cost of modernizing may outweigh the added benefits and efficiencies of integrating newer technologies. This aligns with the cost-benefit principle and the opportunity cost principle, which emphasize the need to weigh the potential advantages against the associated costs and the value of alternative resource allocations.

The issue, however, lies in the current narrative surrounding how LLMs are being integrated into businesses versus how they might serve consumers. In business environments, LLMs are being explored for their ability to streamline workflows, enhance data analysis, and support specialized operations. These integrations are exciting and valuable, but they cater primarily to structured and professional settings. For standard consumers, the value of LLMs remains far less clear, and the gap between what LLMs can offer and what the average consumer needs is still significant.

Additionally, it’s important to acknowledge the substantial financial investment in the AI and LLM space. With so much money at stake, companies are incentivized to push their products and integrations onto consumers—regardless of whether these tools are truly needed, finished, or widely adopted. This dynamic raises questions about the balance between genuine innovation and the pressure to force technology into consumer markets prematurely.

This distinction underscores a critical point: for now, standard consumers are unlikely to fully benefit from LLMs. As the technology evolves, it must bridge this divide by addressing everyday consumer needs in intuitive and accessible ways, moving beyond niche or enterprise applications. Yet, despite this gap, we find ourselves flooded with LLM apps and services targeting consumers. This trend is likely driven by several factors:

1. Companies are eager to pioneer and establish themselves as early leaders in the space.
2. The integration of LLMs into business environments is not progressing as quickly as many companies anticipated.
3. LLM capabilities are not yet mature enough to support widespread adoption and seamless integration across industries.

While the advancements in AI and LLM technologies are undeniably groundbreaking, their current trajectory in consumer-facing applications raises significant questions. The surge of LLM-based products targeting consumers feels premature—driven more by the race to pioneer and dominate the market than by genuine alignment with actual user needs. Businesses understandably see immense potential in LLMs and AI, particularly in data-intensive sectors such as healthcare, professional services, and retail. However, the same cannot be said for everyday consumers, whose daily lives and workflows remain largely unaffected by these innovations.

At the core of the issue lies a clear disconnect between technological capabilities and consumer realities. For AI and LLM developers to succeed, they must look beyond the thrill of being first movers and focus on creating practical, accessible, and intuitive applications that address real-world needs. This means moving past tools designed for novelty—such as generating amusing images or catering to niche interests—and working toward broader, more meaningful use cases. Whether by seamlessly integrating into enterprise workflows or delivering tangible value to everyday users, the future of LLMs will depend on closing this gap. The tech industry must also recognize that not every problem requires a high-tech solution, nor is every consumer ready—or willing—to embrace these tools.

Ultimately, the success of LLMs will hinge not only on their technical sophistication but also on their adaptability to the diverse needs of their audience. To truly thrive, these technologies must evolve from being “solutions in search of problems” into tools that empower, enhance, and enrich lives across all sectors of society. At the same time, it’s essential to acknowledge that not everyone will benefit from or adopt these advancements—a reality that should inform the design and implementation of AI-driven solutions. Companies must also prioritize clear and effective communication about what these tools can do, avoiding overly complex documentation or, worse, the absence of meaningful guidance.

Consider, for example, the integration of LLMs into social media platforms to enhance data sourcing and analysis. X (formerly Twitter) has incorporated its AI/LLM tool, Grok, enabling users to analyze data, ask questions about posts, explore potential trends, and—of course—generate humorous images. These applications highlight how LLMs can provide value when tailored to specific contexts, simplifying complex tasks and making them more user-friendly. However, this value assumes that users are both interested in social media and willing to pay for premium access—an assumption that limits the broader applicability of such tools.

Beyond social media, we see sporadic pushes for voice-enabled products. While these tools are helpful, particularly for users with visual or other impairments, their utility often fails to extend beyond simple tasks or neat tricks. Many voice assistants suffer from limited features, platform restrictions, reliance on third-party applications, and other constraints. This doesn’t make them inherently bad, but their functionality often feels underwhelming. For instance, aside from setting alarms or reading messages, they rarely achieve their full potential. That said, there are exceptions. One feature I personally appreciate is Apple Intelligence’s ability to summarize notifications directly on the lock screen—a practical application of LLMs operating transparently in the background. This underscores my earlier point: LLMs often shine in performing seamless background processes rather than acting as interactive consumer products.

It’s also important to acknowledge that some companies may push LLMs onto consumers primarily as profit-generating ventures, regardless of practical use cases. With massive investments, potentially slower-than-expected integration into businesses, and the ongoing challenge of maintaining updated models, one must ask: how much of the consumer-facing LLM market will genuinely serve users, and how much will exist solely to drive revenue?

There are undoubtedly scenarios where LLMs can provide immense benefits—my own experience is a testament to this. These tools have been invaluable for programming assistance and explaining complex concepts to diverse audiences. Yet, I recognize that many of my use cases are niche. The true impact of LLMs will depend on how thoughtfully they are developed, integrated, and applied.

I sincerely hope that, as users and consumers, we will see LLMs evolve into technologies that genuinely enrich our lives. More than that, I look forward to the day we achieve true AGI—a future where technology not only assists but profoundly enhances our understanding of the world and our ability to connect with one another.

Until next time,

Synfinner

P.S. I wanted to cover some of the privacy and bias issues within models, however, that alone is a post in itself.