Fewer Gen AI Deployments and Declining ROI: What Founders Need to Know

The latest trends in the generative AI (GenAI) space should catch the attention of every founder. Fewer AI projects are making it to deployment, and those that do are delivering diminishing returns. As excitement around AI shifts from novelty to execution, founders are discovering the high costs, complexity, and challenges involved in delivering real value. If not addressed, these challenges could reduce market valuations and investor confidence.

The Reality Check: AI’s Challenges in Execution

In the past year, businesses have faced mounting obstacles with data preparation and quality. According to Appen’s 2024 State of AI report, data bottlenecks — from sourcing and cleaning to labeling — have increased by 10% year-over-year. Moreover, the percentage of AI projects successfully making it to deployment has fallen by 8.1% since 2021, while the proportion of deployed AI projects generating meaningful ROI has dropped by 9.4%.

Founders must pay close attention to these trends. As companies venture into more sophisticated AI use cases, the need for high-quality data has grown, complicating the path to ROI.

Key Takeaway 1: Adoption of GenAI is Surging, but Data Challenges are Growing

Generative AI has grown by 17% in 2024, powered by advancements in large language models (LLMs). These models are helping companies automate tasks across R&D, IT, and manufacturing. However, the same flexibility and power that make GenAI attractive also introduce new challenges.

Generative AI outputs are often unpredictable and subjective, making it harder for companies to define success. Si Chen, Head of Strategy at Appen, emphasizes that businesses must move beyond large datasets to focus on quality — meaning diverse, labeled, and context-specific data. Custom data collection is becoming a core strategy as businesses realize web-scraped data alone won’t cut it.

“Companies are finding that just having lots of data isn’t enough. To fine-tune models, the data needs to be accurate, well-labeled, and tailored to specific use cases,” says Chen.

Key Takeaway 2: Fewer AI Deployments are Showing Meaningful ROI

Many founders are discovering that while AI promises growth, realizing that potential is another story. Projects are taking longer to implement, and success rates are declining. In fact, only 47.4% of AI projects reached deployment in 2024, compared to higher rates just three years ago. And even those that make it to deployment show lower returns, with only 47.3% delivering meaningful ROI.

This drop is partly due to businesses shifting from simpler applications, like image recognition, toward more ambitious AI use cases such as generative AI. While these advanced solutions offer greater capabilities, they require customized datasets and expert oversight to function effectively, which raises the bar for success.

“Generative AI offers incredible potential, but it’s inherently more complex and harder to implement successfully,” explains Chen.

Key Takeaway 3: Data Quality is Declining Amid Increasing Complexity

The complexity of GenAI has made it harder for companies to maintain high-quality data. In fact, data accuracy has dropped nearly 9% since 2021, further complicating AI deployment efforts. Many companies now retrain their models quarterly to keep them relevant, but the need for constant updates makes it challenging to maintain accuracy and consistency.

To overcome this, nearly 90% of companies now rely on external providers to source and manage data, underscoring the growing importance of partnerships in the AI ecosystem. Founders must think proactively about their data pipelines to avoid being caught off guard by bottlenecks.

Key Takeaway 4: Data Bottlenecks are Slowing Progress

The report highlights that bottlenecks in data sourcing, labeling, and cleaning have increased by 10% year-over-year. With AI models becoming more specialized, the process of collecting and preparing relevant data has become more time-consuming and expensive. Companies need to build robust data strategies or risk falling behind.

Some founders are addressing this challenge through strategic partnerships with data providers or by developing internal capabilities focused on long-term data accuracy, consistency, and diversity.

Key Takeaway 5: Human Expertise is Essential for AI Success

While AI technology continues to evolve, the role of human expertise remains indispensable. Appen’s report shows that 80% of organizations emphasize human-in-the-loop machine learning — a process where human input helps improve AI systems. Founders should be aware that AI alone cannot address all challenges. Human oversight is critical for mitigating bias, ensuring ethical AI, and aligning models with real-world expectations.

“Human involvement remains essential for developing high-performing, ethical, and contextually relevant AI systems,” says Chen.

For founders working with generative AI, maintaining this human-AI partnership is especially crucial. The unpredictable nature of GenAI outputs requires careful oversight to prevent errors, biases, or unintended consequences.

What Founders Should Do Next

1. Focus on High-Quality, Custom Data: As data quality becomes a critical bottleneck, founders need to shift their attention from quantity to quality. Partner with external providers if necessary, and develop internal data strategies for sustainable AI development.

2. Plan for Long-Term Deployment Challenges: The decline in deployed AI projects highlights the importance of careful planning. Build realistic timelines and budgets, and expect hurdles around data management and model training.

3. Invest in Human Expertise: Ensure that your team or partners bring the right mix of technical skills and domain-specific knowledge. AI success relies on human oversight for bias mitigation, contextual relevance, and ethical development.

4. Prepare for ROI Fluctuations: Given the challenges around AI implementation, founders should prepare for variable ROI timelines. Set clear benchmarks and adjust expectations as you move from experimentation to large-scale deployments.

Founders must confront these challenges head-on to unlock the full potential of AI in their businesses. The opportunity for generative AI remains vast, but it demands precision, patience, and strategic planning. By focusing on quality data, deploying the right expertise, and navigating complexity with foresight, founders can position themselves for sustainable growth in this rapidly evolving space.

We Love Helping Small Businesses Grow with AI Automation!

At Flowbot Forge, we are dedicated to helping small businesses thrive by providing them with AI-driven automation tools that make operations smoother and more efficient. By automating routine processes, we enable our clients to focus on what matters: strategic growth and long-term success.

If your business is looking to grow and needs help implementing automation tools to reduce manual processes, Flowbot Forge is here to guide you every step of the way.

Schedule time with one of our AI automation experts today, and get your business growing faster!