isnaini azizah
A compelling example of this can be seen in the US NBA basketball team the Charlotte Hornets who successfully leveraged untapped video footage of gameplay—previously too copious to watch and too unstructured to analyze—to identify a new competition-winning recruit. However, before that data could deliver results, analysts working for the team first had to overcome the critical challenge of preparing the raw, unstructured footage for interpretation.
The Challenges of Organizing and Contextualizing Unstructured Data
Unstructured data presents inherent difficulties due to its widely varying format, quality, and reliability, requiring specialized tools like natural language processing and AI to make sense of it. Every organization’s pool of unstructured data also contains domain-specific characteristics and terminology that generic AI models may not automatically understand. A financial services firm, for example, cannot simply use a general language model for fraud detection. Instead, it needs to adapt the model to understand regulatory language, transaction patterns, industry-specific risk indicators, and unique company context like data policies. The challenge intensifies when integrating multiple data sources with varying structures and quality standards, as teams may struggle to distinguish valuable data from noise.
The sheer volume and variety of unstructured data – text documents, emails, audio files, images, videos, social media posts, sensor logs, and more – represent a colossal, largely untapped resource for businesses. Historically, the inability to efficiently process and extract meaningful information from these data silos has been a significant bottleneck. Traditional data management and analytics tools are designed for structured data, characterized by predefined schemas and relational databases. Unstructured data, by its very nature, defies such rigid organization, making it akin to trying to fit square pegs into round holes. This disconnect has led to a situation where a wealth of potential insights remains hidden, inaccessible to the decision-making processes that could benefit from them.
The complexity extends beyond mere format. The quality of unstructured data can be highly variable. A customer complaint might be eloquent and detailed, or it could be a terse, cryptic message riddled with typos and jargon. Video footage might be clear and well-lit, or it could be grainy, poorly framed, and filled with background noise. This inconsistency demands sophisticated techniques that can not only extract information but also assess its reliability and relevance.
Furthermore, the context in which unstructured data is generated is crucial for its interpretation. A medical report from a hospital will use a vastly different lexicon and structure than a legal brief from a law firm, even if both are essentially text documents. Generic AI models, trained on broad internet datasets, often lack the nuanced understanding required to decipher these domain-specific languages and customs. This necessitates a significant investment in data preparation and model fine-tuning to ensure that AI systems can accurately interpret and leverage the unique characteristics of an organization’s proprietary unstructured data.
How Computer Vision Gave the Charlotte Hornets an Edge
When the Charlotte Hornets set out to identify a new draft pick for their team, they turned to AI tools, including computer vision, to analyze raw game footage from smaller leagues, which exist outside the tiers of the game normally visible to NBA scouts and, therefore, are not as readily available for analysis. “Computer vision is a tool that has existed for some time, but I think the applicability in this age of AI is increasing rapidly,” says Jordan Cealey, senior vice president at AI company Invisible Technologies, which worked with the Charlotte Hornets on this project. “You can now take data sources that you’ve never been able to consume, and provide an analytical layer that’s never existed before.”
By deploying a variety of computer vision techniques, including object and player tracking, movement pattern analysis, and geometrically mapping points on the court, the team was able to extract kinematic data, such as the coordinates of players during movement, and generate metrics like speed and explosiveness to acceleration. This provided the team with rich, data-driven insights about individual players, helping them to identify and select a new draft whose skill and techniques filled a hole in the Charlotte Hornets’ own capabilities. The chosen athlete went on to be named the most valuable player at the 2025 NBA Summer League and helped the team win their first summer championship title.
The Hornets’ success story highlights the transformative power of applying advanced AI techniques to previously inaccessible unstructured data. Traditional scouting relied on human observation, which is inherently subjective and limited by the volume of footage that can be reviewed. By leveraging computer vision, the Hornets moved beyond mere observation to quantitative analysis. The ability to automatically track player movements, analyze their speed and agility, and map their spatial positioning on the court provided objective metrics that could be compared across a wider pool of talent than ever before. This allowed them to identify potential superstars who might have been overlooked by conventional scouting methods. The fine-tuning of the computer vision models to the specific context of basketball – understanding the rules, player dynamics, and court geometry – was paramount to their success, demonstrating that off-the-shelf AI solutions are rarely sufficient for complex, domain-specific challenges.
Annotation of a Basketball Match
Before data from game footage can be used, it needs to be labeled so the model can interpret it. The x and y coordinates of the individual players, seen here in bounding boxes, as well as other features in the scene, are annotated so the model can identify individuals and track their movements through time. This annotation process is a critical bridge between raw, unstructured visual data and the structured input required by AI models. Human annotators, or increasingly sophisticated AI-assisted annotation tools, meticulously label objects, players, and key events within the footage. This labeled data then serves as the "ground truth" for training AI models. For instance, bounding boxes drawn around each player at specific frames allow the computer vision model to learn to recognize players. Tracking these boxes across consecutive frames enables the model to understand player movement. Annotating specific actions, such as dribbling, shooting, or passing, further enriches the dataset, allowing the AI to identify and quantify these behaviors. The accuracy and thoroughness of this annotation process directly correlate with the performance and reliability of the resulting AI models.
Taking AI Pilot Programs into Production
From this successful example, several lessons can be learned. First, unstructured data must be prepared for AI models through intuitive forms of collection, and the right data pipelines and management records. “You can only utilize unstructured data once your structured data is consumable and ready for AI,” says Cealey. “You cannot just throw AI at a problem without doing the prep work.”
For many organizations, this might mean they need to find partners that offer the technical support to fine-tune models to the context of the business. The traditional technology consulting approach, in which an external vendor leads a digital transformation plan over a lengthy timeframe, is not fit for purpose here as AI is moving too fast and solutions need to be configured to a company’s current business reality. Forward-deployed engineers (FDEs) are an emerging partnership model better suited to the AI era. Initially popularized by Palantir, the FDE model connects product and engineering capabilities directly to the customer’s operational environment. FDEs work closely with customers on-site to understand the context behind a technology initiative before a solution is built.
“We couldn’t do what we do without our FDEs,” says Cealey. “They go out and fine-tune the models, working with our human annotation team to generate a ground truth dataset that can be used to validate or improve the performance of the model in production.” This collaborative, embedded approach ensures that AI solutions are not just technically sound but also deeply aligned with the business’s specific needs and operational realities. It bridges the gap between theoretical AI capabilities and practical, value-generating applications.
Second, data needs to be understood within its own context, which requires models to be carefully calibrated to the use case. “You can’t assume that an out-of-the-box computer vision model is going to give you better inventory management, for example, by taking that open source model and applying it to whatever your unstructured data feeds are,” says Cealey. “You need to fine-tune it so it gives you the data exports in the format you want and helps your aims. That’s where you start to see high-performative models that can then actually generate useful data insights.”
For the Hornets, Invisible used five foundation models, which the team fine-tuned to context-specific data. This included teaching the models to understand that they were “looking at” a basketball court as opposed to, say, a football field; to understand how a game of basketball works differently from any other sport the model might have knowledge of (including how many players are on each team); and to understand how to spot rules like “out of bounds.” Once fine-tuned, the models were able to capture subtle and complex visual scenarios, including highly accurate object detection, tracking, postures, and spatial mapping. This meticulous tailoring of AI models to the specific domain and objectives is what unlocks their true potential. It transforms generic AI capabilities into powerful, business-specific intelligence engines.
Lastly, while the AI technology mix available to companies changes by the day, they cannot eschew old-fashioned commercial metrics: clear goals. Without clarity on the business purpose, AI pilot programs can easily turn into open-ended, meandering research projects that prove expensive in terms of compute, data costs, and staffing. “The best engagements we have seen are when people know what they want,” Cealey observes. “The worst is when people say ‘we want AI’ but have no direction. In these situations, they are on an endless pursuit without a map.” This emphasis on strategic clarity and defined objectives is a timeless principle that remains essential for successful AI adoption. It ensures that AI initiatives are not merely technological experiments but are strategically aligned with business imperatives, driving tangible value and competitive advantage.
In conclusion, the journey from dormant unstructured data to actionable enterprise AI success is paved with meticulous preparation, contextual understanding, and strategic goal-setting. The Charlotte Hornets’ experience serves as a powerful testament to the fact that by embracing advanced AI techniques like computer vision and by working with partners who understand the nuances of data preparation and model fine-tuning, organizations can unlock the immense potential of their unstructured data, driving innovation and achieving significant business outcomes. The future of enterprise AI hinges on the ability to effectively harness this often-overlooked data reservoir, transforming complexity into clarity and insight.
