Artificial intelligence is surging across every industry, transforming how companies develop new products and how they run their business. Venture capital and corporate spin-offs have fueled thousands of new AI-focused companies. The AI industry is evolving at an unprecedented rate, with new startups forming and being acquired constantly.
The Pharma/Biotech industry has been a particularly hot sector. Fierce Pharma cited a recent report from health-tech-focused VC firm Define Ventures:
The results show a resounding urgency toward adopting AI technologies, with 70% of pharma leaders saying they view AI as an “immediate priority.” Across the top-ranking Big Pharmas specifically, that figure jumped to 85%.
Given the complexity and estimated $1B to $2.5B and 10 to 15 years to develop a new drug, it’s no wonder that companies are investing in AI with the promise of speeding up the process and reducing the cost.
More recently, there has been a rise in AI being applied to complex manufacturing and supply chain processes such as inventory management, production scheduling, predictive maintenance, and logistics. According to Rockwell Automation’s 10th annual State of Smart Manufacturing Report, 95% of manufacturers have invested in, or plan to invest in, AI/ML over the next five years. With the ever-increasing complexity of product offerings tailored to meet the unique needs of discriminating buyers, it’s no surprise manufacturers are looking to AI/ML for help.
What’s Taking So Long?
As has been true with previous technological advancements, AI is taking longer to deliver meaningful benefits than promised (or at least expected). Don’t get me wrong, AI is having an impact and AI tools seem to be improving by the day. I’m sure everyone reading this has used AI to speed up and/or improve the quality of their work.
But applying AI to complex business processes requires clearing many hurdles including data-related problems, scalability issues, human training and adoption issues, and, biggest of all, trust issues. As in, can I trust the AI generated output to be accurate? AI hallucination is a top concern of every business leader I’ve spoken to on this topic. All are interested in utilizing AI to improve their organization’s efficiency. However, most have a degree of skepticism that AI will be able to perform tasks accurately and autonomously.
Despite the concerns, companies are pouring money into AI, both by building internal capabilities and by purchasing AI products/services. The investments may not be delivering a return yet, but the size of the prize is too big for companies to sit on the sidelines. Some might argue that it doesn’t matter where you start, it’s more important that you don’t get left on the sidelines. However, making progress with AI requires a plan, just like any other major technology initiative.
But Implementation of AI is Different
AI is widely described as a like electricity or the internet that fundamentally reshaped economies. Because it can be applied to so many business functions and processes, it can be difficult for organizations to prioritize where to start. As a result, it is very common to find AI being used by individuals to automate simple tasks such as drafting minutes from a meeting or crafting a LinkedIn post. It takes more organizational alignment and fortitude to apply AI to complex problems such as optimizing production scheduling or unifying structured and unstructured data across multiple systems to build digital twins that can be leveraged to make better decisions.
The promise of AI is system-level transformations enabling autonomous, real-time decision-making that fundamentally changes how businesses operate. This takes time and a structured approach. However, you can’t implement AI like you implement a new ERP system.
Traditional enterprise software implementations center around the need to define and standardize your core business processes and structure your core business data to make them work. As complex as they are, they operate on pre-configured rules and logic. AI, on the other hand, learns from data and adapts to changing conditions. ERP implementations follow a structured methodology with extensive user training and system testing. AI implementation is more iterative with feedback loops and model training. Because AI can be applied to nearly everything it can be difficult to figure out what to prioritize and how to make meaningful progress.
Below are a few thoughts to help you get started.
- Define use cases: Assess core business processes to determine where value can be created. Ask questions such as: How might AI improve critical decision-making processes? Where are our process constraints and bottlenecks?
- Assess data readiness: AI models need access to accurate, clean data from various sources. Assess data quality across the ecosystem for accuracy, completeness, and consistency.
- Assess data literacy: AI isn’t just a technology investment—it’s a people strategy. Successfully integrating AI requires a strong change management approach to ensure employees understand how the technology works, its limitations, and the impact on their role. Data literacy is critical for success. Without it, employees are unable to check for errors or effectively use the insights generated to improve their work.
- Create an AI roadmap: Outline your AI initiatives including resource needs and investment requirements. Identify dependencies, such as data readiness, and prioritize them based on importance/value, complexity, probability of success, and investment.
Don’t Get Left on the Sideline
AI models and the applications that use them are getting better by the day. The sooner you develop an AI strategy, the better. Your customers won’t wait. Your competitors won’t wait. AI is transforming how companies operate. Yes, there is a lot of hype with AI right now. It’s not magic and there is no one-size-fits-all solution. But there are many real applications that provide real value.
Don’t get dissuaded by the complexity or implementation challenges. The potential is still immense.
October 16, 2025
Author
"*" indicates required fields