AI Everywhere
The meteoric rise of ChatGPT and the unprecedented advancements in artificial intelligence (AI) have become a clarion call for enterprises worldwide to embrace this transformative technology. However, despite the growing buzz surrounding AI, many organizations find themselves uncertain about where to begin their AI journey. The good news is that it’s still early days, and the potential for AI adoption is vast.
In the 2023 Gartner CIO and Technology Executive Survey, 32% of respondents said their organizations had deployed AI and machine learning, with an additional 17% indicating they would follow suit within the next 12 months. So, about one third of the survey respondents already deployed AI and machine learning in 2023 while that percentage will grow to 49% by mid-2024. Most respondents will still not have deployed AI and machine learning by mid-2024. The bad news for those who are hesitant is that the early AI adopters are seeing substantial benefits from their use of AI. From personalized shopping to fraud detection, AI is making a difference.
First things First, Data
Data is the fuel that propels the AI rocket. You must have a reasonably good handle on your corporate data before undertaking any AI efforts. You don’t need perfect data management to get started with AI, but you do need good data management. If your corporate data is unreliable, fractured, inconsistent, or otherwise significantly flawed – start with your data. Build a data transformation strategy, implement a modern data management infrastructure, and start migrating your data to the new architecture. See more about Infinitive’s data transformation approach here.
Five Steps to Implementing an AI Program
Define High Priority AI Use Cases
Assemble a cross-functional team to start defining those AI use cases that are a) impactful to your business, b) generally solvable with AI, and c) executable based on the corporate data you have. The determination of business value should come from your strategic plans with quantified benefits “blessed” by finance. The question of “AI practicality” will be answered through industry experience. Infinitive can help with this…. see the AI Use Cases page on our website.
Whether the corporate data needed for any AI use case is available and accessible should be apparent from the data transformation efforts you have underway (see: “First things first, data”). Pro tip #1: Data quality is more important than data quantity. AI can provide useful results with a reasonable amount of high-quality data. Rich data is key. AI will not provide useful results with any amount of low-quality data. Pro tip #2: If the corporate data needed to implement the desired AI use case is low quality, unreliable, etc. – move on to the next AI use case.
Assemble a Team with the Talent Required to Design the High Priority Use Cases
For example, if “pricing, promotions and markdown automation” is a high priority AI use case – ask who has the expertise within your organization to understand that topic in detail. Subject matter experts from areas like supply chain, finance, merchandising and IT will need to be assembled to design the AI process associated with “pricing, promotions and markdown automation.” Technology experts are needed to pick the most appropriate AI technology to apply (machine learning, rule-based systems, etc.). This design gets complicated quickly, especially the critical effort to build the results of the AI analysis into the enterprise’s business processes. Pro tip: don’t try to concurrently undertake too many AI use cases at first.
Build the Data
Get the well managed corporate data necessary to execute the AI models for the selected use case. Build out the data pipelines required to move the relevant data from wherever it is stored to the AI analysis platform. Pro tip: make sure that the data pipelines are well designed, well monitored and fault tolerant.
Train the Original Model
Prepare the teaching and testing data. Train the models with the teaching and testing data. Verify the quality of the AI-based results. Install, deploy, and test the entire functional solution (including the model). Pro tip: Perfect is the enemy of good. Once the model generates better results than the (non-AI) status quo, start a limited deployment.
Deploy, Scale, Assess and Optimize
Start by deploying on a limited basis. Measure the ultimate business results. Tweak the model. Once stable (and providing benefits above a threshold), deploy more broadly. Constantly refine the algorithms/model.
Please note that this essay provides only a simplified look at implementing AI-driven solutions. For brevity’s sake, major areas of effort such as bias testing and model governance have been left out.
Infinitive can help with all aspects of AI project implementation – from data strategies necessary to drive the data transformation required to begin the AI journey to the engineering of data pipelines to the analysis and development of the AI models that drive the business results.
With a clear understanding of the essential steps involved and the right support, organizations can confidently embark on their AI journey. By leveraging AI technologies effectively, your organization can revolutionize your operations, enhance customer experiences, and gain a competitive edge in today’s digital landscape. Embracing AI is no longer a luxury; it is an imperative for enterprises aiming to thrive in the age of AI.
