By Alan Jacobson
According to research conducted by Alteryx, 82% of global business leaders say generative AI is significantly impacting their company goals and nearly half of board members are prioritising genAI over anything else. This is a massive tectonic shift in organisational strategy. And while many stories talk about the view that none of these very attractive genAI initiatives will work if they’re not built on a solid bedrock of AI-ready data, this isn’t the real impediment. What is the secret to making genAI drive value in an organisation? It turns out, it’s much the same as with all analytics: education for the organisation on how it all works – and finding use cases that are safe and easy to implement quickly to build muscle in this space.
So, how can organisations build their muscle and drive results?
GenAI models can be highly risky and problematic, or incredibly risk-free
Ask a genAI model to file your taxes for you, and you likely will end up in a dark locked room that you won’t be getting out of any time soon. This is not only due to data quality, but the very nature of what Large Language Models are good at and what they aren’t capable of today.
And while IT teams are focused on data quality with the report revealing that IT teams are confident with their data maturity and trustworthiness. Over half (54%) rate their data maturity as good or advanced, and 76% trust their data. On the surface, this sounds promising. But if you try executing this use case, it likely will not matter how good the data is.
Instead, if you pick a use case to use genAI to highlight what new tax codes might impact the business and automatically e-mail these to the tax strategy team for review, you might immediately have a winning use case with very little risk. In this example, you are not dependent on your internal data quality, you are asking an LLM to summarise the mountains of news articles about new tax code, which is something LLMs are quite good at, and finally, you are not exposing or requesting any sensitive data to the LLM.
There are a myriad of these easier and safer use cases that quickly allow organisations to build the muscle needed to succeed in the genAI space. But how do you identify these easier use cases? It really comes down to education. Similar to the education needed to harness the power of automation and more traditional analytics in your business, teams need to focus on upskilling and ensuring teams have the tools necessary to go on the data journey.
Unfortunately, the report shows that only 10% of businesses claim to have a ‘modern’ data stack, and nearly half (47%) are currently working on updating their infrastructure. While new employees are learning these skills in universities, they are arriving into the workplace with little beyond Excel to perform analysis, let alone to leverage genAI.
What key elements make up the modern data stack?
Organisations need to have a set of technologies that allow the storage of data in a unified location (data lake), the ability for knowledge workers to manipulate data beyond Excel (data wrangling), to automate processes (automation) and perform analysis (analytics). These tools must be accessible and easy enough for the majority of knowledge workers to leverage so that data work is not the sole domain of the IT or Data Science teams. Unfortunately, where companies have historically invested has been in tools for their technology teams, with Python and other ‘code-first’ types of tools. While these help a small number of technical experts, these tools alone will not allow a business to go on the journey of harnessing the power of analytics and genAI. And without bringing the business along, the use cases will likely continue to be suboptimal.
As organisations build out their data stack, it is important to keep an eye on ROI. Building a data lake takes significant time and resources (e.g. data engineers) and will cost significant money. And unfortunately, the act of building a data lake by itself will not deliver significant ROI. ROI will come when applications, automation and analytics are delivered. These other types of technologies will take on two forms, centralised teams building solutions and democratised teams leveraging analytics and automation. In the former case, this means investing in people to centrally build solutions. This typically again takes significant investment and tends to focus on larger problems that have good ROI but take a while to deliver. You will put your top data scientists on big problems. Democratised ways of using these technologies will take much smaller investment, as you are not building a large team of dedicated resources to build solutions, but instead upskilling the people you already have. The goal is to make these people more efficient and with time being freed up, they can then drive to higher value delivery.
Some companies get frustrated with an over investment early on data lakes with cost and returns not in-balance. Successful companies tend to drive fast returns with democratised analytics, and then re-invest a portion of these savings into their data lakes and centralised teams. They also benefit by democratising the analytics, as the business can now better articulate the priorities of what they need the data teams to deliver as well. In the end, the best data stacks are designed to deliver ROI every step of the way.
Aligning budgets with the genAI opportunity
Another challenge facing organisations is budget management. IT teams, in general, are responsible for data technology budgets, but the reality of how those budgets are allocated and adjusted tells a story that may have made genAI adoption difficult. Over half (54%) of businesses admit that budgets are not reviewed or adjusted throughout the year, even if new needs arise. Added to that, 54% say that if other priorities, projects, or spending needs arise after budgets are allocated, they cannot be adjusted. This proved to be a huge challenge last year when the pressure to adopt genAI grew exponentially.
Given how quickly genAI has moved over the last couple of years and how quickly it continues to change, encouraging cross-functional collaboration and communication and updating how IT budgets are allocated or reviewed is vital. The current rigidity among organisations will have a big impact on innovation to the data stack and creating the right foundations for successful genAI use cases.
Clearer horizons for enterprise-wide rollout
While there are many challenges to achieving a modern data stack, organisations must focus on upskilling their workforce while putting the appropriate infrastructure in place to deliver analytics across the organisation. The key is to democratise the effort and ensure the teams are engaged and able to participate in the journey, not focusing only on technology for technologists. By addressing these challenges, organisations will be able to harness the full potential of genAI, driving innovation and achieving organisational goals.
While companies are still at the early stages of seeing the full impact of genAI adoption, there is no doubt that the fundamental elements of analytic teams in the enterprise will shift, from simply building solutions to teaching the organisation and helping deliver the change management to upskill the workforce. Organisations must be prepared to drive this data literacy while navigating the age of genAI.
Alan Jacobson
