Business leaders should know that AI is already practical enough to use, but only if they start with the right problem. Chris Whitney’s advice is clear: question what you think you know about the speed of work, pick a real but lower-risk use case, keep a human in the middle, and build from there.
In this episode of Up In Your Business, Jason Randall sits down with Chris Whitney, Chief Technology Officer at Questco, to talk about AI in the place where it actually matters: inside the operating rhythms of a real business.
The conversation starts with the frustration most leaders feel. AI is everywhere, but much of the discussion is either too technical to act on or too vague to trust. Jason names the simpler questions business leaders are really asking: How does this apply to my business? Where do I start? What does it do to my people when I do?
Chris answers those questions less like a futurist and more like someone who has had to build the thing, launch the thing, fix the thing, and get people to use the thing.
Business leaders often overestimate the complexity of getting started with AI. Chris says the barrier to responsible experimentation has changed quickly, and companies that assume AI requires massive investment, a large technology team, or a perfect long-term roadmap may delay too long.
Chris says his answer to this question would have been different six months ago. That alone is useful. The topic is moving so fast that static AI strategies can become stale before they are fully implemented.
His point is that AI tooling, agents, and model connectivity have lowered the barrier to entry. Leaders no longer need to treat every AI initiative like a major systems overhaul. Some projects still require deep technical oversight, but responsible experimentation is far more accessible than many companies assume.
“You should be questioning everything that you know,” Chris says, “in terms of just the speed of accomplishing work in this new AI and agent-driven world.”
That line carries much of the episode. The mistake is assuming yesterday’s project timelines still apply. A process that once required months of research, vendor selection, and development may now be testable in weeks or days. The danger is less about choosing the wrong AI theory and more about getting stuck in planning mode while the tools keep changing.
Chris puts it simply: take calculated bets, move, and iterate.
A business should start by giving employees a responsible way to experiment with AI in their daily work. Chris recommends equipping individual contributors and leaders with tools that help them generate documents faster, assess problems faster, and bring AI into normal workflows before pursuing larger transformation projects.
That guidance sounds modest, which is why it works.
AI conversations often jump straight to the grand use case: fully automated workflows, intelligent agents, software that replaces weeks of manual work, and systems that connect hundreds of steps. Chris is not dismissing those possibilities. He is saying most companies need a place to begin before they try to automate the most consequential parts of the business.
His first question for leaders is practical: Do you have a mechanism for responsible experimentation?
That mechanism does not have to live only with traditional IT staff. Chris specifically mentions curious business analysts, curious interns, and existing employees who can lead and experiment. The first AI champion may not be the person with the deepest technical resume. It may be the person who understands the workflow, asks good questions, and is willing to test without pretending the first version will be perfect.
Questco built BETH, an internal AI agent that helps the benefits team answer plan-related questions faster. Before BETH, researching a benefits question could take 20 to 30 minutes. With BETH, the same research can happen near-instantaneously, with a specialist still validating the answer before responding.
This is the strongest part of the conversation because it gives the AI discussion a concrete object.
BETH was not born from a vague desire to “do AI.” Questco had a specific problem. Benefits specialists were spending time searching through standard operating procedures, benefits guides, plan details, and other documents. The information existed. The library was already curated. The issue was retrieval.
Jason describes it as a “needle in a haystack problem.” In a PEO environment, those answers matter. Benefits questions can be time-sensitive and emotionally charged. Facts matter. Speed matters. A client or employee waiting on a benefits answer is not experiencing an abstract technology problem. They are trying to make a decision.
Chris saw that Questco already had a strong data repository. The information was public-facing and not sensitive personal client information. That made it a strong first AI use case. The team could point AI at a trusted document library and create a conversational way for specialists to query the data.
The result was not AI replacing benefits expertise. The result was AI removing search friction so the specialist could spend more time validating, responding, and serving the client.
The team solved one thing well, shipped quickly, and kept the work tied to the company’s service mission.
Questco learned that AI cannot simply be clipped onto an old process. Chris says the team had to create a new AI-driven process, with the specialist using AI to retrieve an answer, validating that answer, and then responding to the client. Adoption and change management also took longer than the technical build.
Chris is blunt about what they got wrong. It is tempting to look at an existing process and imagine AI replacing one step inside it. Client asks a question. Specialist searches the library. Specialist reads the document. Specialist sends the answer. Just point AI at the library and the process is solved.
Chris says it was not that simple.
“You have to inject a new AI-driven process,” he explains.
In the BETH example, the process became: specialist goes to AI, AI provides an answer, specialist validates the answer, specialist responds to the client. That sequence is different because it acknowledges what AI can do well and where human judgment still belongs.
The second lesson was change management. Chris says an AI solution is not something you “build it, set it, and forget about it.” It is “a living thing that changes.” The launch is only the beginning of the management work.
Chris argues that leaders should use AI to amplify people, especially by removing heavy manual work. The better question is which tasks consume time that employees could spend on client response, relationship-building, process improvement, and judgment-heavy work.
The replacement question is unavoidable. Jason brings it up directly because every AI conversation eventually gets there. Are we replacing people? Are we augmenting them? What is actually happening?
“If you’re a leader and you’re purely looking at how can I replace roles, this is just the wrong thinking.”
At Questco, the goal is to empower specialists and teams to move heavy manual work into AI-supported workflows. That could mean faster research, faster document creation, or faster problem assessment. For the benefits team, it meant spending less time hunting through documents and more time responding to clients.
AI adoption works better when leaders can name those tasks clearly. The question for employees, Chris says, is what AI could do for their role. Can it make them quicker? Can it help them operate better every day? Can it remove enough manual drag that their more valuable work gets more attention?
A good first AI use case should matter to the business without carrying the highest possible risk. Chris recommends starting with something a human can review, such as AI-assisted standard operating procedure creation. Do not start with payroll, compliance, or another core process where a bad answer creates serious exposure.
Chris introduces “vibe coding” as part of this discussion. His definition is simple: rather than traditional developers or engineers writing and deploying code, AI helps do it for you. That shift changes the pace of building.
But speed does not mean every process is a candidate.
“You’re not going to vibe code payroll,” Chris says. “You’re not going to vibe code compliance.”
Start where the work matters. Avoid starting where one wrong answer creates immediate damage.
Leaders should separate two kinds of trust: trusting AI with company data and trusting AI to perform the work correctly. Before using AI, a company needs to know what client data, company data, sensitive data, restricted data, and PII it has. Then it needs human validation inside the process.
Chris treats trust as an operating requirement, not a vague cultural aspiration.
Before a company puts AI on top of data, leaders need to understand what kind of data they are working with. Client data. Company data. Sensitive data. Restricted data. Personally identifiable information. Those categories need to be defined and controlled before the AI layer is introduced.
Then comes the second trust question: Will AI do what we want it to do?
Chris’s answer is human validation. Put a person in the middle. Have that person check the output, validate the information, and apply practical judgment to what comes back. As the AI gets better at the specific process, the workflow becomes more efficient. The person stays because the work still requires judgment.
“All roads lead,” Chris says, “to just making sure that your people are equipped to be able to measure that trust coming out of AI.”
That may be the cleanest operating principle in the episode. AI trust is not a feeling. It is a process. It is built through data control, validation, and repeated use inside a defined workflow.
Business leaders should pick one focused AI use case, assign responsible owners, define the data involved, keep human validation in the workflow, and iterate. AI does not require every business to become a technology company. It does require leaders to stop treating speed, trust, and people as separate issues.
Chris ends with a broader challenge. The types of problems leaders once assumed required more people can now be reconsidered in an AI-driven world. Even technologists who once said they did not code can challenge that assumption.
His phrase for this is “super architects.” With the skills professionals already have, powered by what AI can now do, people can build, test, and improve in ways that were not available before.
That does not mean every leader should run toward the biggest possible AI project. The lesson from Questco is more disciplined than that. Choose a real problem. Use data you already trust. Keep a person in the middle. Learn from the process itself. Then move again.
Business leaders should know that AI is already useful for focused operational problems, especially where teams spend too much time searching, drafting, or moving information. Chris Whitney’s advice is to start with responsible experimentation, move quickly, and keep human validation in the workflow.
BETH is an internal AI agent Questco built to help benefits specialists answer plan-related questions faster. It uses a curated repository of benefits documents and allows specialists to ask conversational questions instead of relying only on keyword searches.
A company should choose a first AI project that matters but does not carry extreme risk. Chris recommends starting with something a human can easily review, such as internal SOP drafting, rather than starting with payroll, compliance, or a process where a wrong answer creates major exposure.
AI does not have to replace HR or benefits specialists. In Questco’s BETH example, AI removed manual search work while the specialist stayed responsible for validating the answer and responding to the client. Chris frames the goal as amplifying the team.
Human validation matters because AI can produce useful answers while still getting details wrong. Chris recommends keeping a person in the middle to check the output, apply business context, and measure whether the answer can be trusted.
Companies should understand and control their data before putting AI on top of it. That means defining client data, company data, sensitive data, restricted data, and personally identifiable information. Data governance should come before broad AI deployment.
For more on how Questco built BETH, what Chris Whitney learned from the deployment, and how business leaders can think more clearly about responsible AI adoption, listen to the full episode of Up In Your Business.