HomeEnterprise AI

Human Ready, AI-Enabled: Enterprise AI Scales When Companies Redesign Workflows Across The Business

November 18, 2026

Scott Schemmel, Strategic Advisor at OAKS Advisory Group and former CIO at Construction Resources, on how AI returns show up when companies rebuild work from end-to-end, cut the handoffs, and govern the data their agents act on.

Human Ready, AI-Enabled: Enterprise AI Scales When Companies Redesign Workflows Across The Business
Credit: CIOnews

Get the latest from CIOnews.

Enterprise AI, governance, risk, and leadership insights for CIOs, CTOs, CISOs, and technology leaders.

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
"You'll have a product team, a data team, a sales team, a marketing team all moving the same data different ways, and no one is talking across those lines. The fix starts with alignment."

Scott Schemmel

Strategic Advisor
OAKS Advisory Group

Most enterprises have spent two years layering AI onto the work they already do, and the financial returns have stayed thin. Automating one step inside a broken process only produces faster waste, and it locks the broken process in place. The larger opportunity sits in the workflow itself: the handoffs between departments that no one owns, the manual rekeying that carries data from one system to the next, and the missing controls as autonomous agents begin acting on that data directly. At the enterprise level, the payoff shows up only when the workflow itself gets rebuilt.

Scott Schemmel is a Strategic Advisor at OAKS Advisory Group, where he works with software and technology-services firms on scaling and transformation. He spent most of the past decade as a technology executive, serving as Chief Information Officer at Construction Resources, where he built the company's first technology organization across ten product categories, a run that was recognized as a Georgia CIO of the Year. Prior to that, he served as Chief Technology Officer at Graham-Field Health Products. His earlier CIO roles spanned chemical manufacturing, healthcare technology, and telecommunications. That range, across manufacturing plants, distribution operations, and private-equity-backed turnarounds, shaped a firm conviction about where AI programs stall.

"I use the old framework of people, process, technology, in that order, and I put the people side first," Schemmel said. "AI should amplify your talent and your human judgment, not become a reason to shed two-thirds of your headcount." The sequence is the point. People who understand the work come first, the process they run gets simplified next, and the technology lands last. Reverse the order, and the tools scale whatever dysfunction was already there faster than they fix it.

  • Redesign before you automate: "At my last company, one division had a team of 13 spending 37 hours a day keying in orders by hand, and there was no value in it," he said. "We piloted a tool that read the orders however they came in, gave the team a confidence score to check, and took order entry from eight minutes to two. That moved the manual work from 37 hours to about 14." The pilot succeeded and still stalled. Leadership would not move forward unless the hours converted into headcount reductions, which Schemmel now treats as the smaller mistake. The larger one was scoping the project too narrowly. "The real miss was that there was a whole step before this, on the sales side, where people were filling out an Excel template and emailing it to the order entry team to type into the system," he said. "If we had looked at it from the sales side forward, we could have eliminated that entire handoff, and the value would have been dramatically bigger." Outside research points the same way: the strongest financial returns track with redesigning the work itself rather than layering tools onto the steps that already exist.

  • Cross-functional over departmental: The handoff problem is rarely technical. It comes from departments optimizing for their own goals without a shared view of the company's. "You'll have a product team, a data team, a sales team, a marketing team all moving the same data different ways, and no one is talking across those lines," Schemmel said. "The fix starts with alignment. We can't do ten priorities at once, so what are the three biggest ones for the organization? Go execute on those." He saw the CIO as uniquely placed to force that view, pairing domain experts with technology teams and letting the work get ranked by business impact rather than by which department pushes hardest.

  • Augment the experts: Amplifying people means handing experts information they couldn't see before, then leaving the decision with them. Schemmel pointed to a mining and manufacturing plant where three veteran operators, each with 30 to 40 years on the job, ran the same plant three different ways. "They couldn't see the ore grade in real time. It went to the lab and came back the next day, which was too late for what they were manufacturing that day," he said. "We approximated it with sensor data, ran it through machine learning, and gave them the optimal settings for those conditions, and they could accept it or adjust it. A change in ore grade could have put us down for 24 hours, and now they could see it coming." The system recommended; the operators still decided. That line between a recommendation and an action is where he wanted human judgment to stay, especially as the outputs start feeding other systems.

Keeping judgment in the loop gets harder as AI stops advising and starts acting. An agent that writes a wrong answer into a downstream system does more damage than a chatbot that answers a question wrong, because the error propagates before anyone sees it. Schemmel's view is that scaling safely comes from building the controls that let people trust the parts running on their own, so every process doesn't have to be slowed down with manual review. That work splits into two problems: giving employees a safe way to build, and governing the data the agents consume.

  • A governed path for shadow AI: "Shadow AI and shadow IT have always gotten a bad rap, but don't treat it as the enemy," Schemmel said. "You want to expand your capacity to innovate. The rule is that nothing gets deployed without coming through an automated process: a security scan, checked in, then deployed." The failure mode he described is the prototype that keeps running after its creator moves on, wired to live data, never properly security-reviewed, and eventually dropped on an IT team that never knew it existed. He cited Jack Henry, the financial-technology firm, as an example of the model working, where employees are encouraged to build and a single automated pipeline handles the security review and deployment for everything they produce. It lines up with leaders who secure rather than block, routing employee-built AI through controls instead of shutting it down.

  • Trust the data agents act on: The second problem is the data itself. Companies have chased master data management for years and never reached perfect data, Schemmel noted, and the gap matters more at machine speed. "It's one thing to ask AI a question and get a wrong answer. It's worse when you have agents writing bad data into a system, because that compounds," he said. He described tooling that scores data down to the individual element rather than the whole record: "It puts context around the data wherever it lives, links the same entity across different databases, and gives you a trust score on how recent and how accurate each element is. Then you have audit trails, so you can go back and review." That's what lets oversight become selective. Low-stakes agents that have proven reliable run with lighter monitoring, and the consequential ones keep a human in the approval path. Done well, it moves governance from written policy to enforceable controls, the kind of structure a published risk framework expects around automated decisions.

For Schemmel, every one of these examples comes back to the same discipline. A working pilot is rarely where programs break down. What decides whether AI pays off is whether the company redesigned the work around it, tied the effort to real priorities, and built enough trust in its data to let the system act.

"Next to the CEO, the CIO sees the whole business end to end," he said. "If you build those relationships and really understand the business, you can add a lot of value. But you have to look at the whole process, because it's going to cross every traditional functional boundary."

research report

From the Edge to the Core:
Bringing Agentic AI to the Heart of the Enterprise.