Our thoughts on the future of digital innovation and the cloud.

 /  Craig Slack

FinOps for AI

AI is one of the fastest growing areas of interest for many organizations, but the complexity and lack of experience around AI workloads and use cases make jumping into these types of projects feel like a leap of faith.  The forecasting and budgeting can be very challenging, both for the project itself and the ongoing costs once completed, which could be enough to prevent some organizations from moving forward with AI. Those companies that have already adopted FinOps to manage their cloud costs will be ahead of the competition in their ability to track costs and value with the FinOps framework already in use.  

Here are some of the factors that make forecasting and budgeting for AI related costs difficult:

Unpredictable Cloud Costs
  • AI workloads often have variable compute demands, leading to unpredictable cloud costs. Spikes in demand can significantly increase expense.
  • Costs associated with data transfer and storage can fluctuate, especially when working with large datasets.
Complexity of AI Projects
  • The complexity and scope of AI projects can vary widely, making it difficult to estimate development costs accurately.
Evolving Technology
  • The fast pace of AI technology means that new tools and techniques are constantly emerging, which can affect cost projections.
  • Integrating AI with existing systems can be complex, often requiring unforeseen adjustments.
Data Quality and Availability
  • High-quality data is essential for effective AI, but acquiring, cleaning, and managing data can be costly and time-consuming.
Talent and Expertise
  • Hiring or training personnel with the necessary AI expertise can be expensive and competitive.
  • Continuous learning and development are required to keep up with the latest AI advancements.
ROI Uncertainty
  • The success of AI projects can vary, making it difficult to predict financial outcomes accurately.

While all those challenges can seem overwhelming, the potential benefits that AI brings are encouraging many organizations to get started on homegrown AI solutions to support their business.  How can the principles of FinOps be applied to AI projects to ensure that these early adopters are seeing the value from the investments being made in this exciting new area?

Visibility and Transparency
  • Proper cost allocation through tagging will ensure that AI-related costs are accurately allocated to the appropriate departments or projects. This helps in understanding who is consuming resources and why.
  • Regular and detailed reports utilizing real time data on AI spending helps stakeholders stay informed about cost trends and anomalies. Tracking and using real time data to make decisions can help to eliminate cost overruns for expected compute requirements.
Optimization
  • FinOps practices help optimize the use of AI resources, such as GPUs and cloud instances, ensuring that they are used efficiently and cost-effectively.
  • Identifying and implementing cost-saving measures, such as rightsizing instances and leveraging spot instances, can significantly reduce AI-related expenses.
Collaboration
  • FinOps encourages collaboration between finance, engineering, and operations teams to ensure that AI projects are aligned with business goals and budget constraints.
  • Regular communication with stakeholders helps in setting realistic expectations and making informed decisions about AI investments.
Forecasting and Planning
  • FinOps provides tools and methodologies for accurate budgeting and forecasting of AI-related costs, helping organizations plan for future expenses.
  • Conducting scenario analysis to predict the financial impact of different AI deployment strategies can help in making better investment decisions.
Governance and Compliance
  • Implementing policies to govern AI spending ensures that resources are used in compliance with organizational standards and regulations.
  • Maintaining detailed audit trails of AI-related expenditures helps in tracking and justifying costs.

With the relative newness of AI related project and workloads as well as the unknown nature of cost expectations, the life expectancy of AI within organizations will be dependant on being able to accurately show the quantifiable value these bots and other tools are bringing to users. By leveraging the FinOps framework, organizations can gain better visibility and control over their AI-related costs, ensuring that investments in AI are both strategic and cost-effective.

For more information on how the FinOps Framework can be leveraged to ensure your AI projects achieve their expected value, reach out to the Invero team!