JBS Dev: Navigating the Challenges of Imperfect Data in AI Implementation

May 12, 2026 420 views

Misconceptions about the data requirements for leveraging generative AI systems have persisted among industry leaders, often leading to unnecessary hesitations in deploying these technologies. Joe Rose, the president of strategic technology provider JBS Dev, emphasizes a critical insight: perfect data isn't a prerequisite for engaging with generative AI. This assertion counters the traditional narrative propagated by many software vendors and consultants, which suggests that organizations must first establish extensive data lakes and undergo lengthy data transformation initiatives before utilizing AI tools.

Challenges with Data Quality and AI Utilization

The current state of AI tooling is significantly more advanced, particularly regarding the handling of imperfect data, according to Rose. He highlights that modern large language models (LLMs) can effectively comprehend even poorly structured prompts. This evolution invites organizations to rethink their approach: rather than postponing AI integration until achieving 'perfect' data, they should exploit existing tools to their advantage while maintaining adequate oversight. “People are… used to ‘we build it, it works, we forget about it,’” Rose points out. The reality of generative AI systems requires continuous adjustment and oversight rather than the traditional set-and-forget mentality.

Case Study: AI in Healthcare Billing Systems

Rose illustrates his point with a case from the medical industry where a client faced significant challenges in migrating to a new billing reconciliation system. The data records were far from uniform, comprising PDFs and images, with inconsistencies such as names being misplaced. However, the generative AI tools proved capable of discerning valuable insights from these disorganized records. By employing optical character recognition (OCR) and text extraction techniques, the AI generated a clearer dataset from these disparate sources. Furthermore, agentic approaches were utilized to compare customer records against insurance contracts, ensuring correct billing practices.

Iterative Improvement Through AI

A noteworthy aspect of Rose's approach is the concept of gradual automation in workflows. He advocates for starting with low levels of automation, possibly around 20%, and slowly increasing that percentage as the systems adapt and learn. The expectation is not a perfect system from the outset, but rather a scalable increase in automation and efficiency. This insight is vital for organizations hesitant to dive headfirst into generative AI integration due to fears of unreliability.

Shifts Toward Sustainable AI Implementation

The future of AI discussions seems poised to pivot from explosive advances in model capabilities to more pragmatic concerns like cost sustainability and operational efficiency. Rose predicts a growing focus on developing models that can run efficiently on standard computing devices—think laptops or smartphones—rather than necessitating massive data centers. “How do we make the cost more sustainable that we don’t have to build data centers at the rate we’re building data centers?” he asks, highlighting a pressing concern for many organizations gauging their AI investment strategies.

The DIY Approach to AI

At the upcoming AI & Big Data Expo, Rose plans to advocate for a more hands-on approach to AI integration, urging companies to consider implementing solutions in-house instead of relying heavily on SaaS vendors. He argues that with most businesses already having some cloud presence, they can utilize existing tools to start implementing agentic workloads immediately—bypassing significant costs associated with new software and training. “It’s not as hard as it sounds,” says Rose, suggesting that organizations have more capability at their fingertips than they realize.

The Path Forward

As AI technology continues to evolve, organizations within the industry must reevaluate preconceived notions about data and adaptability. The instinct may be to wait for polished data before embarking on AI partnerships, but this hesitance could stifle transformative opportunities. The real value lies in the ability to harness imperfect data and evolve systems progressively, positioning organizations to better respond to the unpredictable outputs of generative AI.

Looking ahead, the conversation surrounding AI will invariably shift toward practical implementations and sustainable practices—how can organizations leverage available resources to foster innovation while ensuring cost-effectiveness? The technological revolution isn't just about what we can achieve but also how efficiently and affordably we can get there.

Watch the full interview with Rose below: