Many corporate boards are rushing to approve multi-year plans for artificial intelligence (AI) without asking how these tools will change daily operations. While long roadmaps feel safe to directors, locking in a multi-year plan rarely works when modern software shifts every few months. According to Victoria Savio, board members need to look past high-level promises and focus directly on the real work people do every day.
Focusing on Real Outcomes Instead of Multi-Year Plans
Long roadmaps look great in slide presentations, but they quickly fall apart when software changes month to month. When directors focus only on long projections, they risk funding tools that could be outdated before the team finishes rolling them out. Savio points out that boards need flexibility instead of rigid targets: “What are the outcomes that you’re focused on? Because the market is changing so fast in the AI space that to imagine that a plan created today in 2026 is going to look like reality in 2027 is simply not realistic. So we must continue to rapidly pivot, deploy new technology, evaluate, and understand where AI has been and where it is going. That means, roadmaps must be anchored in our business plans and outcomes.”
Financial returns are often checked first by directors, but that view can hide how software changes daily tasks. Rather than viewing new tools purely as a way to cut costs, leaders should look at how technology frees up staff to do better work. As Savio explains, “We can focus on return on investment, but that really focuses on the technology aspect of it, while what we should be focusing on is: ‘What’s the work that we expect to get done?’ and ‘How do we continue to add value with employees and bring about true transformation, rather than just implementing technology?’”
When companies automate basic data entry or routine handoffs, staff get hours back to solve harder business problems. Experienced workers can then spend their energy on judgment calls and customer issues that computers cannot handle on their own. Savio notes that the goal should be “giving them time back, automating manual processes, introducing an agentic process. And hopefully, a lot of organizations are going to take that and allow their employees to do higher-value-add work: the true thinking and understanding work that comes with their decades of business experience and their true acumen about their subject matter expertise.”
Demanding Specifics in the Business Case
Vague business cases often promise higher efficiency across an entire department without naming a single task that will change. Savio believes boards should push these broad proposals back to leadership until the steps are clearly defined. Directors need to know exactly which human tasks will stop and which steps will move over to software.
“I would focus on what process or business objective is going to change. So, I would send back a proposal if it doesn’t name the specific thing that is going to change,” Savio says. “If we’re just focused on those efficiency gains and productivity improvements and we’re not pointing to a specific task, approval process, or handoff that a human is no longer going to perform when we’re talking about an agentic AI, then nothing’s changing. You’re going to have somebody who is using a tool but very possibly still running the process in parallel. So what you want to know is: ‘What is the work that is going to fully shift to AI, rather than have a human and an agent running in parallel?’” Without that clarity, workers will keep doing the same job by hand just to feel safe about the result.
Running two systems side by side wastes company money and makes it impossible to see if the new tool is working. When employees do not trust the software, they quietly double-check every step and run the old steps anyway. Clear business proposals stop this problem early by setting firm rules for where human work ends and software takes over.
Looking Beyond Simple Usage Numbers
Many executives report progress to the board by showing how many people log into a tool each week. While tracking usage is a simple place to start, high login numbers do not prove that anyone is creating real value. Workers might simply open an app because their manager checks the team log every Friday afternoon.
“And we’ve all heard about ‘tokenmaxxing,’ and people are doing it. You’ve got Jensen Huang saying he’d be alarmed if his top engineers weren’t burning through hundreds of thousands in tokens a year, which is well and good, but if they’re not actually doing work that is adding value for the company, then you’re not getting the outcome of that,” Savio notes. Using expensive software to write simple emails or draft basic notes does not help the bottom line. As Savio points out, “We don’t want employees just sitting there and using tokens to write emails. That’s not an efficient use of the company’s resources at all.”
Holding managers accountable means asking what new work the team can handle once routine tasks are automated. Leaders should clearly point to projects that were previously out of reach due to lack of time or staff. Savio suggests boards ask leaders to show that “…because we’ve implemented an AI agent in this area of the business, we are now able to also do X, Y, and Z. And what are those things?”
Cleaning Up Broken Workflows First
Adding modern tools to a broken process will only make mistakes happen much faster. If a workflow lacks clear rules, automated tools will simply repeat those same flaws on a bigger scale. Boards need to make sure management fixes the manual process before spending money to automate it.
“If we already have a bad process, then AI is going to do that process faster and make more mistakes. And if we don’t have the right oversight of the work already, then AI is just going to replicate those issues,” Savio explains. “Does the work that we’re already doing manually with humans make sense in its current process? Is there a proper way to handle exceptions? Is there a human override? Are there issues where someone knows, ‘Yes, we ignore those error messages because they’re fake,’ or, ‘This doesn’t flag as an error, but we do need to look at it and take time to resolve it?’ These are the things that we’re looking for.”
In many offices, experienced staff fix hidden errors every day without ever writing down the steps. When these unwritten habits stay hidden, new software fails because it cannot handle everyday exceptions. Setting up oversight and human controls cannot be left as a task for later when problems begin to show up. As Savio emphasizes, “If that’s an afterthought and something you’re going to do once you see issues coming out, then you’re going to continue to have issues with the process and with the AI. But if you’ve got the governance in place already and you’ve got a good process with the proper exception handling, overrides, and oversight from a human, then you have a real opportunity to scale that, but only with governance.”
Follow Victoria Savio on LinkedIn for more insights on AI transformation, corporate governance, and aligning emerging technology with business operations.
