Today’s AI Fallacy: Why Most Companies Don't Actually Need Artificial Intelligence (Yet)

Introduction

Let's face it — no business is truly “AI-ready” today. Even industry giants like Google and Amazon wrestle with launching, maintaining, and supporting the AI products that we view as sitting at the "forefront" of AI exploration. (To be clear, this isn’t an indictment. AI is so new that we can’t expect anyone to be truly ready for it, especially not at the pace the technology is advancing.) The result is that most businesses remain inwardly AI-averse. 

Outwardly, however, companies are shouting to anyone who will listen about their new AI advancements. Despite these proclamations, the reality for many companies is a cautious approach to AI marked by minimal real-world application, primarily due to the complexities of AI implementation (especially for companies that don’t have an ingrained data and software culture) and the organizational change it necessitates (which always instills fear at the org level). 

Successful AI integration is recognized by industry leaders as a gradual, multi-year journey that begins with establishing a solid foundation in data management and analytics, contrary to the overly ambitious timelines that often lead to disappointing outcomes. Transformation in increments might appear slow, but look at companies that are 5 years into their 1-2 year AI transformation journey — the results are often underwhelming.

Because AI is hard. 

Does that mean that we shouldn’t pursue it? 

Absolutely not. 

Does that mean that we should educate and help companies understand what a realistic roadmap towards AI bliss looks like? 

Absolutely.

Understanding the AI Hype

Artificial intelligence has become the exemplar of cutting-edge innovation and a forward-looking business stance. The allure lies in the promise to revolutionize business operations through automation, personalized customer engagement, and predictive insights. 

Yet, the path to realizing these benefits isn’t straightforward, and it goes beyond simply “adopting” this new technology. It demands a comprehensive foundation in data that many organizations overlook or simply haven’t taken the time to build…yet. The hype cycle around AI further obscures the practicalities involved in its implementation, leading to a huge gap between leadership expectations and real-world results. 

Recognizing this distinction between the potential of AI and the foundational requirements needed to make it a reality is crucial for businesses aiming to leverage AI effectively.

The Data Dilemma

The primary obstacle in harnessing AI is the acquisition and management of data. AI and ML algorithms require massive volumes of high-quality, relevant data to train and function optimally. The efficacy of AI and ML depends heavily on the quality, quantity, and format of the data they are trained on. To deliver accurate predictions, automate tasks efficiently, or generate valuable insights, they require data that is:

• High-Quality: Data must be accurate, complete, and relevant to the problem at hand (that last part is particularly important!). That means error-free, consistent, and (often overlooked) without biases. Example: data used to train a model for facial recognition should be diverse and representative of various demographics and ethnicities to prevent biases.

• Sufficient Quantity: AI and ML models learn from data. The more data available, the better these models can understand patterns and make predictions. The quantity required varies by application but typically involves thousands to millions of data points. Insufficient data can lead to overfitting, where a model performs well on training data but poorly on unseen data that expands beyond the training parameters.

• Appropriately Formatted: Data should be structured in a way that AI algorithms can process. This often means converting raw data into a clean, organized format, such as tabular data with rows and columns for different observations and features. Achieving this level of data cleanliness can be time consuming. Example: text data may need to be tokenized and vectorized for natural language processing (NLP) applications, and images might need to be resized and normalized for computer vision tasks.

• Relevant: The data must be pertinent to the application or business objective on which your AI model is being trained. Example: if developing an AI system to forecast sales, the data should include variables known to influence sales figures, such as historical sales data, promotional activities, seasonality, and market trends.

• Accessible and Legal to Use: Organizations must have the legal rights to use the data, especially when dealing with personal or sensitive information. And that data needs to be secured in a manner that aligns to privacy laws and ethical considerations.

“Oh, no problem! We’ve been focused on big data for years now — we have tons of data!”

While this may be true, many organizations still lack the appropriate type or amount of data, or they struggle with converting existing data into a format that is actionable for AI purposes. This discrepancy between the available data and the data needed for effective AI applications often turns AI into a costly investment with limited near-term returns. Many organizations face challenges in meeting the criteria above for several reasons:

• Data Silos: Data might be scattered across different parts of an organization, making it difficult to aggregate and unify for AI training purposes.
  

• Legacy Systems: Older systems may produce or store data in formats that are incompatible with modern AI tools, necessitating complex (and costly) conversion efforts.
  

• Lack of Data Strategy: Organizations without a clear data management strategy may not collect data systematically, leading to gaps in data or collection of low-quality, irrelevant data.
  

• Privacy and Regulatory Compliance: Ensuring data is used in compliance with regulations (e.g., GDPR in Europe, CCPA in California) can add complexity to data collection and usage, particularly when dealing with personal data.

Addressing these challenges requires a concerted effort to improve data governance, invest in data infrastructure, and perhaps most critically, adopt a mindset that views data as a strategic asset essential for AI efforts. This reality makes the data dilemma a critical first step to consider for companies looking to benefit from AI. 

Misaligned Expectations and Reality

A significant challenge in AI adoption is the mismatch between the expectations of what AI can do — largely based on the massive hype around the technology (as described above) — vs. the realistic capabilities of AI given the challenges outlined. AI is not a magic potion to cure all business issues. But it is a powerful tool that can deliver transformative benefits —  when applied correctly and in suitable contexts. 

Understanding realistic use cases and limitations of AI within your organization is the first step in managing expectations and achieving tangible outcomes. The alternative is outsized costs with minimal near-term returns driven by overly ambitious AI projects. 

In reality, many operational challenges can be addressed faster and more cheaply by utilizing simpler solutions that align more closely with the actual needs and capabilities of your organization. Some solutions to consider while you’re focused on readying your data for future AI efforts include:

• Business Process Optimization: Before turning to AI, companies should look to optimize existing processes. In our experience, the inefficiencies in operations can be significantly reduced by reevaluating and streamlining workflows, removing bottlenecks, and improving communication channels within the organization. This approach requires minimal technological investment compared to AI and is a precursor to the steps below.
  

• Software Supported Automation: Many of the “AI tasks” that customers ask us to help architect for them can actually be automated through software in conjunction with careful planning around the potential use cases they need to support. In fact, our proprietary Mission Control platform is designed specifically to enable these types of automations, communications, and task management. Examples: data normalization, report generation, data analysis, etc.
  

• Rule-Based Systems: For decision-making processes and communications, rule-based systems can be employed to automate responses or actions based on a predefined set of rules. These systems are particularly useful in areas where outcomes and inputs are well-understood and consistent, offering a level of predictability without the need for complex AI models.

• Data Analytics and Business Intelligence (BI) Tools: Advanced data analytics and BI tools, including the above-mentioned Mission Control, can provide deep insights and predictive capabilities using statistical methods rather than AI. Examples: trend analysis and forecasting based on historical data.

• Collaboration and Project Management Tools: Implementing or enhancing the use of collaboration and project management tools can significantly improve efficiency and communication in an organization. Examples: managing tasks, timelines, and resources more effectively.

• Customer Relationship Management (CRM) Systems: For businesses looking to improve customer engagement and sales processes, today’s CRM systems offer a wide range of workflows such as sales automation, customer service, and marketing campaigns that do not require AI to be effective.

By focusing on these simpler solutions, organizations can achieve significant improvements in operational efficiency, customer satisfaction, and decision-making processes without the heavy investment and complexities associated with AI. This approach not only ensures a more manageable implementation but also enables businesses to build a strong foundation upon which AI could later be integrated more effectively.

Delivering on AI Today and Tomorrow

Embarking on the AI journey necessitates a solid groundwork in data management and strategic planning, and Apollo 21 can help with these efforts. Companies should prioritize establishing robust data practices, clarifying the specific challenges and opportunities within their operations, and defining clear success metrics. This process often reveals opportunities that are simpler and more direct than AI, pointing towards innovation strategies that better fit the organization's immediate needs while also keeping in mind long-term goals.

For companies that have enough mature data to provide initial value through AI, Apollo 21 has designed our Mission Control platform to be the easiest way to ingest, normalize, and gain value via AI. Mission Control was created to bridge the gap between the opportunities we see most often today — automation, task management, data management, and more — and the future of AI opportunities that come with a more mature data practice. 

Conclusion

In sum, while AI offers significant potential, it is not universally applicable or beneficial for all companies. At Apollo 21, we focus on a pragmatic, strategic approach to AI adoption. That means we often advocate for the use of simpler tools to address specific problems and opportunities before we begin deploying AI. This means recalibrating from the AI hype to concentrate on enhancing data capabilities and improving core operational efficiencies. This strategy enables organizations to establish a sustainable path toward innovation and growth today while preparing for the benefits of AI and other future technologies where they best fit the roadmap.

‍