AI is becoming default, its footprint is non-trivial and CEOs and CMOs need governance. According to PWC only one in eight (12%) of CEOs say AI has delivered on cost and revenue benefits. In today’s tech landscape, it is almost impossible to not interact with some form of AI model online or in our day to day lives, it has become the ultimate disruptor, much like the internet once was. According to TechCrunch, ChatGPT users send 2.5 billion prompts a day, which demonstrates the sheer magnitude of how frequently people are using Gen AI LLM models today. However, how many of those prompts are actually productive vs asking AI to show us what we would look like if we were a cat?
What is often swept under the rug is the devastating impact AI has on the environment. Many people forget to take into account the energy consumption AI requires to function at such a high demand, huge data centers have been built across the globe, requiring vast amounts of water to keep cool. These data centers are also often built in rural areas, away from larger cities and communities, which leads to these rural areas becoming completely overshadowed by the data centers, increasing electricity costs, destroying land, and rendering entire communities unlivable.
Tech Disruption, Resource Depletion
According to the International Energy Agency, AI data water consumption could reach between 4.2-6.6 billion cubic meters by 2027. It is also projected that by 2030, US data centers will annually contribute 24-44 million metric tons of carbon dioxide into the atmosphere, again demonstrating the sheer force of powering AI systems. Business leaders should take this into account when deploying models across their organization, and it is possible to remain sustainable while remaining cutting edge.
Treat AI usage like your budget: Allocate it, govern it and measure waste
The dominant narrative today is to use AI as much as possible, top C-Suite level executives are encouraging their product teams to roll out AI into as many branches of the business as possible, while others speak to news outlets, at events and their own blogs about how everyone needs to adopt AI now, while not fully outlining what business cases it will address. This is not an ideal strategy, nor a sustainable recommendation, companies that have scaled AI with strong foundations are pulling ahead, while those who integrate without strategy fall behind. This demonstrates a clear misalignment in how AI is being used within companies’ infrastructure.
So, what should you actually do to not only get AI right, but keep AI sustainable?
The answer to that is use case centric AI approach. AI is useful when split up and segmented into chunks against use cases that have practical application to the problems and pain points of your business. Roll out AI where it solves problems, not just because you can.
At Flock, we’ve found the fastest way is to limit AI deployment to a small set of defined use cases, then scale only when adoption and quality are proven, which we initiated with our own in house model, Starling.
Building an ecosystem that sets AI up for success – CMO Checklist
Build an ecosystem that supports positive and sustainable AI growth, having approvers, guardrails and training are vital to any AI deployment across organizations. When deploying AI across the organization or servicing a client, CMOs should consider the following actions:
Define approved use cases
A successful AI rollout should solve a business case that addresses pain points or improve an existing process the business is currently facing. If a deployment is not achieving this, it may be a waste of resources and serve to disrupt your teams more than enhance them. Ensure use cases are agreed on by key stakeholders, and that the users are looped in and aligned on what AI solutions are being proposed.
Set guardrails and approvers
An AI system is only as good as the data its fed on, so it’s important to set guardrails in place that ensure the training data is accurate, objective centric and up to date. Ensure that approvers are set in place and that output is monitored for accuracy and consistency. A good example of a guardrail for an AI deployment is setting up criteria of what a reliable data set looks like, measured against recency (is the data the latest and up to date), format (is the data clean and presented accordingly) and objective fit (is the data a good contribution to what the expected output should be)
Right-size models
The less efficient an AI tool is, the most unsustainable its use will be. By ensuring users are trained on the tool and the interface is user friendly and objective clear, the tool can serve its purpose without becoming overused or over saturated. An efficient model is built on purpose, for example deploy a smaller, narrow model for analysis objectives while reserving larger enterprise models for advanced reasoning and output purposes. Deploy prompt guides to avoid saturation and train the models on output that builds solutions rather than dragging conversation.
Track adoption and waste signals
When measuring a new AI tool, it is crucial to evaluate against at least 2 metrics: User satisfaction and resource allocation. If users are “abusing” a tool, it likely means they either aren’t fully sure how to use it, or they are not getting the output they need effectively enough. This leads to a waste of resources, both financially and environmentally. By ensuring that the intended users of the AI tool are using it for purpose and that the output is effective but not excessive, you can ensure that you are maximizing efficiency while minimizing resource drain.
Deploy training initiatives
All these strategies only work if you appropriately train the users on the tools being deployed. Run internal workshops, collect feedback, execute champions and aim for maximum transparency and collaboration in rollouts. A trained team whose voices are heard will contribute better work with AI use than those who aren’t, and educated use is sustainable and beneficial for the organization.
The final takeaway
Generative AI is an influential yet disruptive tool that is only expanding and further influencing the core of how we conduct business in the 21st century. Like with any technology, it is important to understand the broader picture and responsibly determine how we can best use such a powerful tool to enhance us, rather than replace. Business leaders have an opportunity to utilize this technology to transform their companies in a way that is both beneficial to the business, and sustainable to the environment. AI without governance becomes waste. AI with governance becomes advantage.
Determine a strategy of when to use AI, but it’s also crucial to know when not to. Ultimately, the most important part to any business will always be the people, and that should never fall short of the top priority.