“The future is already here, it’s just not evenly distributed.”
This quote from William Gibson perfectly captures the new AI world order. Few people truly understand and embrace the power of AI. However, for those who do, the future is happening right now!
In my experience, about 70% of AI projects stall at the pilot or proof-of-concept phase which aligns with the 46–91% range found across industry studies. These AI projects never create any real business value. In my conversations with over 400 executives around the world and involvement with over 100 AI business transformation projects, I have often wondered why the failure rate is so high. What I noticed are three consistent executive blind spots. These mindset roadblocks, if addressed early, can accelerate an organization’s AI success journey, and reduce throwaway AI expenses. If ignored, they can cause projects to stall.
1. AI is a Superpower, Not Merely a Disruptive Technology
There has been a consistent dose of new disruptive technologies in our lifetime, ranging from Blockchain to Crypto to Web 3.0 to Robotics to AR/VR to IoT. However, these disruptive technologies are not the ones that “change everything”. In my definition, technology that changes everything has a double-digit percentage impact on the growth of global economies. The technologies that truly change everything have been few and far between. They include electricity, computers, the internet, and now Artificial Intelligence! AI will not only impact the global economy at the same level as the internet, but it will do so 5 times faster. Imagine the opportunities and risks this creates for your business.
Consider this real-life example of AI transformation in a mid-market manufacturing and services company based in the US Midwest. They made a strategic decision to go AI-first to solve problems and explore opportunities in their business. They choose an AI transformation platform purpose-built for mid-market to enable this decision. They have since successfully solved 8 use cases (business problems areas) with a 6x return on this AI investment in 7 months. These use cases range from human resources to financials to production forecasting to new product development. I am deeply involved with their AI success journey that broke even on the investment in 3 weeks. They expect to reduce their operating and resource expenses significantly. Yet, they have no intention of letting anyone go. Instead, they plan to leverage their team’s knowledge to scale and disrupt their industry and even acquire their competitors. With a potential 4x change in EBITDA by the middle of 2025, they are well poised to do so. These numbers are hard to believe but are going to be commonplace in the next few years (or even the next few quarters).
Consider AI as just another disruptive technology at your own peril. Harness this superpower. Embrace the future.
2. Data Myopia in an AI World
Data is not enough; the right data is even more important for AI. The saying
“there’s no data like more data” does not always apply to solving business problems using AI. It would be better to say “There’s no data like the right data.”
The phenomenon of data myopia, an over-reliance on too narrow, too broad, only internal, or incorrect datasets, poses a significant hurdle, causing 87% of AI projects to falter. If ignored, data myopia can cause high pre-processing costs, lower or partial value realization, missed opportunities, and biased and inaccurate outcomes even with the best AI solutions.
Executives who succeed understand that AI business outcomes are as dependent, if not more dependent, on the data fed to AI than the AI itself. They understand that the right data for AI is paramount. Right data or “AI-ready data” is comprehensive with internal and external data, purpose-built for the problem you want to solve, and easily understandable by AI models. They ensure their AI systems have access to AI-ready data so they can not only ensure better models for better business outcomes but also solve problems that were hidden before.
One real-life example of this happened when an automotive company had gotten control of its internal data. This mid-market company accumulated all data sitting across their ERP, CRM, and other business systems within the company for the last 10 years. This itself is not an easy task and I commend them. But when they started on their AI-success journey, they failed to drive true business value. Most of their AI solutions did not produce the desired outcomes. These AI solutions were inaccurate or hallucinating (deteriorating outcomes over time) once they got out of labs. They had to reset and reevaluate their strategy. This company had already invested a significant amount of dollars but instead of abandoning the AI initiative, they dug deeper. In a few weeks, the answer became obvious. They had excellent AI systems and talents, but the data was myopic. They had over-data, under-data, un-indexed data, missing data, a lack of external data, misaligned data and so many more issues. Simply their data was not AI-ready, and they were suffering from data myopia. To solve this problem, on the one hand, they had to truncate data and eliminate unnecessary and old biased data, and on the other hand, augment data from outside sources and even create new data by digitizing their business processes. Once they solved this, it was smooth sailing. It took a few weeks to get data AI-ready and tweak the underlying AI models, but the same AI solutions started to perform well. They have since implemented 18 AI Solutions (use cases) and improved efficiency, revenue, and innovation. The lesson here is that even powerful AI models are only as good as the data they’re fed.
Focus on the AI readiness of your data. Enjoy better business outcomes.
3. Not All AI is Created Equal: Focus on the Right One for You
In the diverse world of Artificial Intelligence, a salient truth stands out: Not all AI is created equal. As businesses increasingly turn to AI solutions, understanding the quality and specificity of these tools becomes paramount. The divergence stems from varying AI quality and its alignment with business objectives.
Knowing which algorithms, models, and tools are good for which business problem (use-case), in which industry, at what size, and with what data, leads to identifying which AI solutions align with your business objectives. In many cases, a single algorithm/model or tool may not be sufficient. You may have to stack multiple models and tools to get the outcomes you want. You must also keep up with the dynamic AI landscape where new and better AI solutions are popping up almost daily. Now combine these issues with talent availability, budget constraints, regulation, data risks, and intellectual property protection, and the task of choosing the right AI for you becomes monumental.
In over 100 AI-driven transformation projects I have been involved with, I have learned the following. Executives may not be able to find the perfect AI solution for their business needs, but they can decipher the best options for it. Executives must choose AI solutions with a deliberate intent to increase their chance of success.
In one such transformation project, a company chose to add value to their customer-facing financial dashboards using AI. They wanted reduced processing times to get reliable data, create even more insights, and simulate business scenarios with “what-if” functionality. They had no prior AI investments in any AI technology.
Their options were –
(1) Finetune GenAI LLM: LLM is a type of AI, recently gained popularity due to the success of ChatGPT. This requires tweaking a commercialized LLM AI solution (like OpenAI) for your specific needs.
(2) GenAI LLM Wrapper: Use commercialized LLM AI solution as-is by adding a few UI screens (web pages) to interact with it.
(3) Open-source GenAI LLM and
(4) Buy a Niche AI Tool
They chose option (1) Finetune GenAI LLM to get quick results and integrate with existing analytics dashboard solutions. They further added option (3) Open-source GenAI for a use case that was driven by confidential information and would allow them to monetize their IP in the future. They rejected option (2) GenAI LLM Wrapper as it was technically complicated, outcome accuracy projections were poor, and as a result not going to add value over time. They rejected option (4) Niche AI Tool as they could not find a good niche AI tool in the market for their needs and the AI
capabilities of existing analytics solutions were underwhelming.
They set up different timelines and budgets for each of these while ensuring cross- project efficiency by using the same team. They had to lean on an external consultant for option (c) for 3 months while for option (a) they trained the internal IT team.
There are frameworks you can use, or you can just apply common-sense logic to make this choice. Once you have selected the problem to solve and taken inventory of your existing AI investments, get clarity on the budget and timeline. Next, you can easily search or ask your tech team to find industry-specific AI solutions and models typically used for that specific problem. You can even go to ChatGPT and just ‘prompt’ this question.
In the realm of AI, solution discernment and their strategic alignment aren’t just recommended, but most likely are the difference between success or failure of your AI journey.
