Capability Over Cost Guides How One Enterprise CIO Weighs Every New AI Investment

The biggest return on AI may come from work that never made it onto anyone’s desk. By giving existing teams capacity they could never hire for, companies can pursue overlooked revenue, buried risks, and long-deferred improvements while keeping payroll flat.

Rob Spellman is the Chief Information Officer of Leidos QTC Health Services, a medical exam healthcare services company that provides disability and occupational evaluations for government agencies including the Department of Veterans Affairs. He started as a software developer and moved into technology leadership from there. Long before the current wave of AI tools, he had teams using machine learning and natural language processing on medical data. That history shapes what he asks of a new system before LQTC buys it. "If you're bringing in AI just for what it can save you, you're going at it the wrong way," Spellman said.

• Beyond the box: AI had already reshaped Spellman's own day. It gave back hours once lost to building slide decks and reading long email threads, and it opened room for decisions he had not always been able to reach. Security work showed the same pattern. Alerts had always piled up faster than the team could read them, and the technology helped analysts find the signal buried in the noise. The change reached how people saw their jobs, too. For years, workers had treated a role as a fixed box with clear edges. "I think that mentality is just getting crushed right now," Spellman said.

• Capability first: Cost was where most AI conversations started, usually framed around how much a tool could save. Spellman led with a different question and asked what a system could do that his teams could not do already. If the answer was strong enough, the financial return came on its own, through new revenue or lower operating costs. That approach also gave LQTC a way to stand apart from competitors, since a task done faster or better carried value even when it never showed up as savings. "We get more enamored by the capability than the potential cost savings," Spellman said. A clear mission held the whole effort together, and everyone at LQTC could point to the same target of growing the business without growing the payroll.

• Skeletons in the closet: During a migration to Amazon Connect, LQTC tried to switch on Agent Assist, which pulls answers from a knowledge base and suggests them to support agents in real time. The data it needed was scattered across formats and systems, with images stored apart from the text they belonged to. Pulling usable answers out of that mix turned into slow, manual work, and the effort put a floodlight on tech debt that had built up quietly for years. "If you can have code that writes code, you can have AI that fixes tech debt, especially when it comes to data issues," Spellman said.

Cleaning up the data was the part LQTC could control, a knowledge base it could untangle on its own timeline. The harder work sat with people, and it kept to nobody's schedule. Customers had to approve any new system before it touched their data, and employees had to fold the same tools into their daily work.

• Governance takes time: For LQTC's healthcare clients, any new system had to clear formal review before it could go live, and the word AI tended to raise the stakes right away. Customers pressed on where their data would travel and how it would be kept safe. Questions about bias in the models came up as well. Spellman's team met that caution head-on, sitting with clients to walk through the architecture and show how the information was handled. The approvals ran long enough to reshape a project's schedule. "That takes anywhere from 45 to 90 days, so you have to build that into the roadmap," Spellman said. Starting that clock early kept later stages from backing up.

• Lost in the fear: Spellman built each rollout around how it got explained. He framed a new tool around the work it absorbed and the hours it freed, telling staff the goal was to lift a task off their plate so they could focus on higher-value work. A communications plan and a training plan ran alongside the choice of a tech stack, since a new system changed how people worked and they needed time to adjust. "From a leadership level, they see all the advantages, and then as you go out to deploy that and you go through the organization, people question how it impacts their day-to-day. That message gets lost in the fear," Spellman said.

• Meet Iris: New tools still felt foreign to staff, so the team at Leidos gave its AI assistants names and personalities. Iris fielded IT requests for employees across the company, and another related agent helped people find the benefits they qualified for. The familiarity chipped away at the hesitation that met each rollout, and made the interactions feel like working with a coworker.

Spellman said he expected the line between human and automated coworkers to keep thinning until it no longer mattered which was which. His shorthand for that future was a colleague named Chad. "Chad may be a bot, or Chad may sit in Virginia. I think we're going to get there a lot faster than people realize."