Walk Slowly but Surely
Unless one lives in a cave, one cannot go through a week and not hear about AI and how this is going to change the world. Every cloud and business application company is embedding AI into their platform – from Microsoft to SAP to Oracle to Salesforce/Tableau and Alteryx to name a few. 1 I also know of conversations at the Board level where CFOs and CIOs are being asked about their AI strategy and its impact on the business. This conversation is not about just using co-pilot for MS office or summarizing Teams meetings or cleaning up Python code or improving call center experience. It is more about the real business use of AI that would increase ROI, deliver better value to clients, and improve sustainable competitive advantage.
But before one starts the journey towards AI, the question is whether one even has the right equipment and capabilities to undertake this journey. And believe it or not – it begins with “data” and “data quality”. I believe that the most precious corporate mineral that requires locating, extracting and refining is “data”. I have said this before and will say it again – data is the new gold, information from that data is the new platinum and actionable insights generated from that information whether through BI or through predictive analytics or through ML/AI is the new rhodium.
The Role of Data and Data Quality
The right granular level data allows businesses to become more competitive, tap into new markets, develop better products and services and generally increases profits and ROI.
However, just like an ore, data alone is useless. It must be processed and shaped into something useful before it can add value to an enterprise. This transformation of data or “data alchemy” is often described as big data, Internet of Things, predictive analytics, machine learning and artificial intelligence. With the advent of Generative AI, we are now witnessing a new frontier in data analysis, where machines can not only analyze data, but also create new insights and solutions based on that data and serve it to users through “prompts”.
Given the buzz around ChatGPT, Open AI, generative AI and more, boards of companies are turning to their C-level leaders to lead the company through this AI minefield. In manufacturing companies, this is sometimes referred to as Industry 4.0. The companies that successfully implement their AI vision that delivers better value to clients will be richly rewarded and the rest will face challenges in their business model over the next 3-5 years. It is best to keep in mind that AI is not the answer for all problems or organizations.
So, how can modern CFO/CIOs undertake this Data to BI to AI journey?
Here are 7 axioms:
Axiom 1: Look for data everywhere
For AI to be really helpful, it needs access to data across the enterprise. So, the first step is to document/identify all sources of data on the corporate repository and application, which can then be leveraged for “ease of access” or serve it for AI modeling. The data requirements for a project-based company, for instance, requires more in-depth insights into every project, past and present and contracts and lessons learned and documents. A software company would be more interested in data that helps to analyze new versus recurring revenues. One would also require identifying sources of competitive advantage and not just scour the “web” because customized LLMs allow one to create a customized environment.
Axiom 2: Clean your data
Unfortunately, unless data is collected by systems without any human intervention, enterprises have to live with “dirty data.”
We have all heard the saying “garbage in, garbage out.” When the data was not consistently recorded, is missing or has not been maintained, it will likely need some refining before it can be trusted for use. CFOs and CIOs need to create a disciplined approach for ensuring that data is scrubbed clean and is consistent across systems. Otherwise, insights will be derived from misleading and incorrect information.
Axiom 3: Unite islands of data or let AI do it for you
Very few companies have all their data available in one master repository. Internal data often exists in multiple business systems, and external data often arrives piecemeal. This “disconnected” data means we have to deal with “islands of data.”
The good news is that we now have AI tools that can pull data from multiple systems when a prompt or a query is given to the model once built, either within the designated corporate walls or a blend of internal and external sources. Advanced analytics and generative AI tools can also access data lakes to further enhance the insights derived from these disparate data sources.
Axiom 4: Upgrade how you visualize and serve information
Data alone or a response to a prompt is necessary but not sufficient. Lager (in terra bytes) data sets may require improved visualization to provide actionable insights. It is the CFO’s/ CIO’s role to invest in data visualization or business intelligence tools that allow for the intuitive display of historical information and also allow for a sophisticated analysis that will help drive action for the future, either through real-time predictive analytics or ML/AI tools such as Azure Synapse and Analytics. The integration of Generative AI can also provide dynamic visualization capabilities that adapt to user interactions, offering a more personalized data experience.
Axiom 5: Drive accountability for AI usage
Like any other investment, CFOs need to monitor the usage of this “data-to-insights thru AI lens” investment by key decision-makers. I typically recommend that companies start using data within corporate walls before deploying AI randomly. Moreover, if the Company’s BI journey has really not matured enough, then the AI journey is going to lead to low ROI.
Axiom 6: Identify a use case or two
The biggest issue today is to identify a use case which the C-suite believes can lead to higher ROI or satisfy an immediate business prerogative. These use case can range from revenue enablement, market segmentation, trade promotion optimization, supply chain optimization, cross/upsell modeling, predictive maintenance, price elasticity for dynamic pricing, etc. Each company would have its unique business use case. Then choose the right AI tool/platform to “model” the problem. Many open source no/low cost platforms now exist to do some “experimentation” with full understanding that data, even if you torture it, may not confess.2
Axiom 7: Ensure security and data access governance at every turn
The main challenge companies are going to face is “security,” meaning who can see what when prompted to the LLM model. For example, when someone prompts “I want to know the total business we have done with client ABC in the last 5 years along with customer comments,” should any person be allowed to see this data?
CIOs should work with all leaders to weave data security policies and practices into every fabric of the enterprise, including employment agreements, IT SLAs and privacy policies. It is vital that data security compliance and data access regulations be put into place and continually monitored.
Companies have spent considerable amount of resources to become Data-driven and to be at the forefront of understanding the business model and performance/value drivers of the enterprise. With the integration of Generative AI, they can further enhance their ability to glean insights from complex datasets and qualitative information across the business applications, transforming how quickly decisions can be made and to gain competitive advantage.
This journey form Data to BI to AI is not going to be easy, and is not for every company. It is also not cheap – from paying for the Software and having the right resources. Already, some surveys are suggesting that ROIs on internal AI projects are less than anticipated. The use cases are not identified correctly. Also, the expectation of ROI from AI are too high. It is also going to require investment in training and choosing the right tool kits. For many, this journey is going to be stop and go, and even painful. But AI is here to stay, and one has to start walking before running and be ready to stumble a few times.
Recent Posts
26 Tableau Features to Know
📝Neha Mishra (Consultant - DAS) Whether you're a novice data analyst exploring the possibilities of Tableau or a leader with years of experience using VizQL to gain advanced insights—this is your list of key Tableau features you should know, from...
How to Integrate IoT Data with Tableau for Actionable Insights
The way businesses work today has changed dramatically and with the Internet of Things (IoT), businesses can easily collect, store and use data from connected devices. Tableau, when integrated with IoT data, is a powerful tool that can visualize complex...
How to Build Effective Recommender Systems with Alteryx: A Comprehensive Step-by-Step Guide.
As businesses today must rely on personalizing user experiences, recommenders reduce the legwork for retail, entertainment, and e-commerce manufacturers, and in Alteryx, the data prep and analytics capabilities make building recommenders a streamlining exercise. We have for you the step-by-step...
Leave a Comments
You must be logged in to post a comment.