5 Ways CIOs Set the Pace and the Foundation for an AI-Powered Business

If Mark Twain suddenly appeared in the C-suite, reincarnated as a chief information officer (CIO), he just might say, “The reports of my death are greatly exaggerated.” Think about it—back in 2000, in the article “Are CIOs Obsolete?” authors at Harvard Business Review agonized:   

[T]he role of the chief information officer is undergoing intense scrutiny. Should the CIO participate fully in strategy formulation? Is the position evolving from a technical manager to a general manager? If, in fact, the CIO role is becoming more strategic than technical, how much will the job overlap with that of the CEO?  

The good news is that the CIO position, like Mark Twain’s humor, is alive and well. This particular executive role has evolved into one that works closely with a host of C-suite peers: chief technology officer (CTO), chief analytic officer (CAO), chief financial officer (CFO), and yes, the chief executive officer (CEO).  

Today’s modern CIO is a driving force in the operationalization of AI. Through this enablement, the CIO is setting the pace, and building an enterprise-class, scalable foundations, for AI- and GenAI-powered business.  

So how do we do it? The technology foundation for Responsible AI and GenAI comprises five principles: 

1.Break Down Silos and Drive Alignment 

In my career as an IT consultant and executive, I’ve never seen any technology saturate the enterprise as fast as AI and GenAI. This tsunami affects business heads, CAOs and CIOs, chief data officers (CDOs), risk and legal, and, in FICO’s case, software engineering. The modern CIO is ideally positioned to create the alignment necessary to take the business to where it wants to go, while avoiding shadow AI initiatives and keeping innovation aligned with risk frameworks.  

CIOs are leading the way in establishing RACI matrices for AI/GenAI that specify who is responsible, accountable, consulted and informed. By defining these roles and responsibilities, it’s clear who owns standards, infrastructure, deployment, and ongoing monitoring. To make sure silos don’t creep back up, joint KPIs and joint roadmaps enforce all parties’ ongoing alignment between business value, technical feasibility, and risk governance. 

In sum, productizing AI delivery—by shifting from a scattershot “projects & pilots” mentality to an integrated, coordinated approach—can really help enterprises cut through the hype of GenAI with chargeback/show back financial accountability and clear adoption goals. 

2. Deliver Operational Excellence 

Highly reliable, high-performing technology have long been the common denominator of high performing organizations One of the cornerstones of success is site reliability engineering (SRE), an approach that infuses IT operations with the rigor of DevOps software development to automate IT operations, accelerate software delivery, and minimize IT risk.  

Most CIOs already have mature SRE practices, but Responsible AI requires more. Taking a platform approach, one that centralizes governance and observability, standardizes a common data foundation, and promotes asset reuse, to AI and GenAI deployment allows CIOs to merge the tenets of SRE with machine language operations (MLOps). Like DevOps, MLOps is a set of practices for streamlining and standardizing the processes (the ML pipeline) through which machine learning models are built, deployed, audited, and maintained. Applying SRE concepts like service level objectives (SLOs) to MLOps gives CIOs the levers to build resilience into AI and GenAI systems, while enforcing and operationalizing SLOs for model uptime, drift, and bias.  

Furthermore, a platform approach typically offers built-in business intelligence and reporting capabilities that let CIOs track the KPIs for ML model performance and reliability.  

3. Shifting from AI Experiments to Enterprise Value 

In addition to productizing solution delivery, the organizational alignment CIOs can drive is critical in transitioning from the “Wild West” of unbridled GenAI experimentation to measurable outcomes that deliver sustained business value. It’s crucial for CIOs to drive this transition right now, as Fortune reports that 95% of GenAI projects are stalled in pilot mode, “delivering little to no measurable impact on P&L,” according to a survey by MIT’s NANDA initiative. Time, and patience, are running short. 

Measuring ROI allows a multitude of pilot projects to be prioritized for funding and empowers CIOs to advance AI maturity within the organization by pivoting from GenAI hype to enterprise decision intelligence. But calculating AI/GenAI ROI is not a “one-and-done” task; measuring value extends beyond model performance, to ongoing evaluation of business KPIs that connect these initiatives to real outcomes. This includes mitigating the risk of rogue GenAI in particular, which may expose customer information or other protected data.  

And then there’s the cost. When innovative, risk-compliant applications are greenlighted for production, the next challenge is to operationalize them at scale, with ultra-low latency, and highly resilient, available resources. The costs can be eye-popping. Through established FinOps practices, CIOs have at their disposal a wide range of tools and techniques to help control spending, while running at scale. 

4. Take a Page from the Secure Infrastructure Playbook 

Ultimately, CIOs can tame the unpredictable performance and costs of production AI/GenAI applications by revisiting a page from a playbook they know well: how to deploy scalable, secure, compliant infrastructure. Transitioning from do-it-yourself pilot programs to enterprise-grade performance makes a powerful case for CIOs to invest in shared infrastructure to scale AI/GenAI deployments responsibly and cost effectively. 

In my experience, predictability is the #1 AI scalability challenge. CIO strategies that are designed for predictability cover a vast expanse of cloud and compute resources, real-time monitoring and more issues, including: 

• Capacity planning, workload isolation, and deterministic pipelines  

• An evaluation rubric for making build vs. buy choices about decisioning solutions and AI operations 

• Architecture decisions on data locality and “sovereignty by design” to ensure that multi-region, multi-cloud AI deployments meet regulatory requirements 

• Real-time observability for AI stacks through a “single pane of glass” 

• Applying FinOps to AI to observe costs for training, inference, vector databases, and agents, and then deciding which should be standardized 

Speaking of playbooks, it’s been said that “AI standards are the new cybersecurity,” which enterprises have decades of experience in operationalizing. Treating Responsible AI in a similar way, like an enterprise risk and resiliency framework, can bring clarity to the challenge. 

5. Implement Responsible AI as a Platform Strategy, Not a Project 

With over 2,000 tools available, GenAI has opened the floodgates for disparate technologies and tools to rush into the enterprise—and they have. At this stage it’s tough for any enterprise to put in place controls to govern this unwieldy technology patchwork, a critical first step toward Responsible AI.  

A platform approach addresses the need for controls and resiliency while unleashing innovation. It puts guardrails in place to protect against the potentially disastrous effects of biased models or unchecked model drift. Creating a safe AI/GenAI innovation space reduces the risk of data leakage from corporate intellectual property assets or other privileged information that may be used to train third-party models or otherwise be released “into the wild.”   

Adopting a platform strategy to operationalize AI and GenAI—I call it “platformization”—is a major topic for CIOs who are tasked with setting the pace (and building the base) for GenAI innovation. It’s such a big topic that I’ll be devoting my next blog to platformization entirely, covering these points: 

• The 50% ROI gains that enterprises are achieving today through AI platformization 

• How CIOs are hardwiring Responsible AI standards into a unified MLOps/decisioning platform 

• What a reference architecture looks like for Responsible AI at scale 

• Additional insights on consolidating model development, deployment, monitoring, and more 

How FICO Can Help You Advance in AI 

• Read FICO’s new State of Responsible AI for Financial Services report 

• Explore FICO’s posts on Responsible AI 

• Download our AI Playbook: A Step-by-Step Guide for Achieving Responsible AI