Spending With Intent: Engineering Your Return on AI

Scaling AI Catie Grasso

It’s easy to spend money on AI. The trick is getting more out of every dollar spent than your competitors and ensuring those costs don’t eclipse the value you’re getting — striking that balance between risk and reward.  

During our 2024 Everyday AI Summits in Chicago, San Francisco, and New York, Nigel Newton (Business Transformation Advisor, Dataiku), Claire Gubian (Global VP of Business Transformation, Dataiku), Kelci Miclaus (Head of AI Solutions, Health & Life Sciences, Dataiku), and Christine Andrews (Solutions Consultant, Manufacturing, Dataiku) shared actionable ways to build a sustainable AI practice measured by its Return on AI (ROAI) while, at the same time, managing risks. This blog features some of the key takeaways from those sessions on engineering your ROAI.

→ Watch the Various Sessions Here

Setting the Stage

To the world at large, 2023 was the first time anyone who wanted to could use AI. To many, it was as if the future had crash landed into the present. OpenAI and ChatGPT put AI on the agenda for every management team and board of directors across the globe. Everyone wanted (and still wants) to know how ChatGPT and other similar models can drive their business into this “future is now” scenario.

The challenge is 80% of AI projects fail. McKinsey, Gartner, AWS, IBM, and a host of others continue to report from surveys that AI, machine learning (ML), and advanced analytics projects either don’t make it into production or fail to achieve business objectives.

Yet, McKinsey estimates that Generative AI has the potential to deliver up to $4.4 trillion annually across multiple use cases. The technology is poised to increase the impact of all AI by up to 40%. It’s no surprise, then, to hear that Boston Consulting Group says a revolution in business-model innovation is coming. 

Dataiku customer Financial Services Regulatory Authority was able to go from pilot to production in just 12 weeks and deploy an LLM-powered risk assessment solution with background checks and document analysis. Moderna’s medical affairs team is saving 40 hours a month by automating the analysis of diverse medical data sources with AI.

This focus on AI being a business driver is changing how we see our ROAI.

Value Engineering in Practice

types of value gains

But it doesn’t change the principles of value engineering. Our business transformation advisors speak with customers on a daily basis who are very comfortable with expressing to their leadership how data platforms and tooling give them speed and agility to accelerate their teams’ work. Stack efficiency provides seamless project workflows, taking away the costly and time-consuming work of stringing project stages together. Control enables them to ensure secure data access and consistent coding and manipulation of data with transparency to identify and quickly correct issues. 

However, these efficiency gains rarely move the needle or attract the attention of time-stretched executive leadership teams. Simply put, in the current climate, organizations can no longer lead with platform value gains, they have to tell the business value story to leadership in a simple understandable way for them, in turn, to tell their board of directors and shareholders. That said, we are getting there: According to an AWS survey, 44% of CDOs define success as achieving business objectives.

This is AI’s time to shine and organizations have the opportunity and responsibility to tell the story in a compelling and effective way. Every workstream and project has to be told from the framework of business challenges and value generated. Tell the story of how the business shifted a team towards R&D and created a new customer experience, as a result of cutting down time spent wrangling data; and the next steps to achieve future milestones and dependencies (instead of leading with hours of productivity gains for full-time employees). 

Enablement Is the Pathway

Enablement is the pathway towards change management. Honing your value stories allows you to continue to be the internal marketer and champion of awareness. In turn, that drives understanding that scales usage of analytics and drives value gains for the business.

There’s little structural change needed in the way data teams are working. You take the same content, data and analytics projects, platforms, and dashboards and hone the usefulness in personalized ways. An example might include creating a leaderboard to show execution and usage from non-centralized data and analytics teams as a core indicator of business value enablement. 

These value stories are the 30-second elevator pitches that enable data team members to have a relatable business response to the question, “What do you do for our organization?” fostering connectivity and clarity of data-driven value gains throughout the business. Here are some examples of those stories:

  • Airline: Over $1 million in people hours saved each year from enabling customer email classification through an API.
  • Retail and CPG: 15% increase in accuracy of revenue forecasts with ML.
  • Insurance company: 3x uplift revenue in cross selling multiple products with a recommendation engine.
  • Logistics and supply chain company: Millions of dollars in revenue leakage avoided each year from an automated dashboard that highlights missing rates in pricing systems.
  • Manufacturing company: $1 million saved in material and engineering costs from automating real-time process control with ML.
  • Retail and CPG: 42% growth in e-commerce sales with a recommendation engine.

These are just some examples of the value that data teams and their members are generating across industry sectors. So, as a data team leader or member of a data team, how do you knit this all together to create a bigger story? 

In many cases, it simply comes down to identifying a beach head or landmark use case and, in consultation with a business sponsor, have them validate the value. In every vertical, the business sponsor and the data science team should be in lock step when it comes to these initiatives and socializing them to boost internal growth. 

Accelerating Your ROAI With Dataiku Solutions 

One surefire way for teams to quickly bring an AI-driven approach to key business challenges is via Dataiku Solutions — pre-built projects and ready-to-use templates designed to accelerate AI use case delivery. We have a comprehensive Solutions Catalog for the myriad of out-of-the box solutions available across key industry verticals, business functions, or departments.

Dataiku solutions

As data team managers and leaders, the pressure is on to deliver tangible business value through AI initiatives. Here are some reminders to help navigate this challenge:

  • Focus on ROAI: Shift your narrative from platform capabilities to business outcomes. Quantify the value of your AI projects in terms that resonate with leadership and shareholders.
  • Craft compelling value stories: Develop concise, impactful narratives that showcase how your teams work directly to address business challenges and generate value. These stories should be your team's "elevator pitches."
  • Embrace enablement: Use your value stories to drive awareness, understanding, and ultimately, wider adoption of analytics across the organization.
  • Collaborate closely with business sponsors: Ensure alignment between data science efforts and business objectives by working hand-in-hand with sponsors to validate and communicate the value of data-driven initiatives.
  • Leverage pre-built solutions: Consider using tools like Dataiku Solutions to accelerate AI use case delivery and deliver quantifiable value gains.

By implementing these strategies, you can position your data team as a critical driver of business success, effectively communicate your impact, and maximize your organization's ROAI. 

Remember, in today's AI-driven landscape, it's not just about spending on AI — it's about spending smartly and showcasing the results.

You May Also Like

From Vision to Value: Visual GenAI in Dataiku

Read More

Understanding the Why and How of the LLM Mesh Architecture

Read More

The Ultimate Test of ChatGPT

Read More