Hartford Funds

Mark Fine

ROI Scaling Catalyst

How does a lean IT shop scale AI without a big team or budget? At Hartford Funds, Head of IT Mark Fine has built a phased playbook—pairing AI literacy and right-sized governance with hard questions about cost and ROI—to turn cautious adoption into real business impact.

Every IT leader is under pressure to adopt AI. For a smaller organization like ours, the harder question isn't whether to move—it's how to do it deliberately, without a big team or budget to absorb the missteps. That constraint has shaped everything: how we build literacy, how we govern, and how we measure whether any of it is actually paying off.

Drivers

• Efficiency & Scale - There is constant pressure to ensure we are not "falling behind." We position this as the need to operate more efficiently and scale without adding headcount — doing more with the same staff.

• Smarter Decision-Making - How can we make datadriven decisions throughout the organization? This includes empowering our teams with insights-driven, prescriptive guidance — enabling better prioritization, more targeted outreach, and more effective execution.

• Growth & Financial Professional Engagement - Beyond internal efficiency, AI creates an opportunity to strengthen how we engage with financial professionals. By personalizing interactions, anticipating needs, and delivering more relevant content at the right time, we can deepen relationships and drive growth in a highly competitive market.

The Journey

We approached this journey in phases, beginning a few years ago with AI literacy and the rollout of foundational tools. Building on Microsoft Copilot, we created a program that blended traditional training with knowledge-sharing and peer forums. And because the pace of change was relentless— Copilot itself evolved rapidly—we introduced “microburst” updates: short, frequent briefings that kept staff current on new capabilities and use cases as they emerged.

Goals & Measurement

 We don't chase AI to avoid falling behind—we adopt it with literacy, governance, and ROI discipline, so every deployment earns its place. 

Adoption was slow at first, with tools that were inconsistent and limited. The turning point came when we defined organization-wide AI goals and aligned them with individual staff objectives through a three-phase progression:

• Phase 1 - Complete training (measured by enrollment)

• Phase 2 - Record concrete use cases against a modest hourly-savings goal (measured by activity)

• Phase 3 - The frontier we are working through now: how do we measure true ROI?

Governance

Our governance model begins by categorizing AI according to how it enters the organization: third-party enterprise and SaaS platforms, internally developed tools with embedded AI, and traditional machine-learning solutions. Across all three, we apply a consistent set of governance criteria—risk classification, human accountability and validation, data and regulatory compliance, and financial oversight. For the foreseeable future, keeping a human in the loop remains essential.

One of those criteria—risk classification—drives how we vet every new tool and use case. We evaluate each through four lenses:

• Accuracy / Quality - Decisions being made on inaccurate information

• Financial - Cost exposure as AI usage scales

• Ethical & Legal - Potential for bias as well as IP infringement

• Security & Data Privacy - Potential for loss of data or intellectual property

Financials

As AI providers evolve their offerings, they are refining their monetization models—and costs are rising as a result. The central challenge for Phase 3 and beyond is demonstrating true ROI, and that will demand discipline on several fronts: better mechanisms to measure the real value of AI and to govern consumption, training that guides staff toward the models that deliver the best value, and financial oversight that moves from informal to formalized. None of this is simple. Traditional metrics such as headcount reduction don't apply, productivity is tricky to measure and largely anecdotal, and quality has to be factored into the equation. Solving for a credible, quality-adjusted measure of value is the work ahead.

Observations

A few reflections stand out. Copilot has matured into a robust platform, and our adoption is now relatively strong—the opportunity ahead is to grow our level of sophistication and, ultimately, our business impact. The biggest win has been within IT itself: our own adoption has opened the door to bespoke tools and process automation that would never have been feasible with traditional development approaches. The proof is concrete—within IT alone, we have saved thousands of hours in the development process, accomplishing in-house through AI agentic coding work that vendors had proposed as paid engagements. This is a rare case where the value is unambiguous, precisely because it is measured against avoided spend rather than softer productivity gains.