The race to generate ROI from AI is on. Boards are asking hard questions, budgets are shifting, and CIOs are under pressure to show tangible value from big bets on automation and generative models.
For as long as I’ve worked in tech, I’ve seen the same pattern in middle management: a weak manager hands out a vague request, a strong employee quietly fixes the brief, fills in the gaps, and makes the project succeed.
The work lands, the manager gets credit, and their shortcomings stay invisible.
AI breaks that pattern.
As generative AI takes on more of the “hero individual contributor” work, like drafting, summarizing, analyzing, and even coding, bad managers lose their safety net. The model won’t push back, ask clarifying questions, or intuit what the boss meant to say. It will do exactly what it’s told, at speed and scale.
AI is only as good as its manager, and managers who struggle with people will struggle with AI.
AI Is a Mirror, Not a Magic Wand
Most AI conversations in the enterprise start with infrastructure: data quality, model selection, privacy, compliance, integration. Those matter. But they’re not where most AI initiatives fail.
AI isn’t only an IT problem. It’s a management problem.
AI is a new kind of “teammate” that takes instructions literally, lacks context unless you provide it, doesn’t challenge your assumptions, and produces something that looks confident, whether it’s right or wrong.
Clear thinkers who communicate well, test assumptions, and stay curious will succeed. Vague thinkers who over‑delegate and under‑review will get burned.
A human employee will translate vague directions into something actionable. They ask questions, push back, and keep iterating until the task is clear. AI takes a vague prompt, mashes up whatever seems relevant, and produces something that looks slick but may be irrelevant or misleading.
If managers can’t set clear expectations for humans, they won’t magically get better at it with machines.
Bad managers also let AI heavily influence key decisions or allow AI‑generated content to go live without vetting for accuracy and tone. In a human‑only environment, employees identify problems; AI won’t. Good managers understand that judgment can’t be automated and humans must be squarely responsible for decisions that affect people, brand, ethics, and strategy.
In a human‑only team, a great employee can rescue a bad brief. In a human‑plus‑AI team, the AI will happily scale that bad brief across the organization.
How CIOs Can Enable Manager Success
Right now, the temptation is to solely focus on tools, platforms, and roadmap. However, if you want AI investments to pay off, you must treat manager capability as part of your AI stack.
Audit your management layer
Analyze the performance of managers, identifying those who consistently turn fuzzy problems into clear, structured work, naturally test and iterate, and take responsibility for decisions, rather than hiding behind “the system” or “the model.”
Those are the people who will quietly make AI successful for you. The others will create invisible drag or visible damage.
Train managers in human+AI workflows
Most “AI training” today is tool‑centric. This was a good start a year ago, but things have grown since then. CIOs now need to promote management‑centric AI training.
Every skilled worker is becoming, to some degree, a manager of AI: writing prompts, reviewing outputs, and deciding when to override the system. Managers now sit above that hybrid layer. Their job is to design and govern socio‑technical workflows where some team members are human, and some are not.
Therefore, CIOs need to teach managers how to decompose work between humans and AI, design prompts that reflect business context, constraints, and risk, set review standards for AI outputs, and communicate changes to teams. AI needs to feel like an extension of their capabilities, not an existential threat.
Implement AI into governance and incentives
AI needs to show up in how you govern and reward. This means defining where AI is mandatory, where it’s optional, and where it’s prohibited. It also requires human review for specific categories of outputs (e.g., legal, HR, high‑stakes customer communications) and measurement of quality, risk incidents, employee engagement, and time‑to‑decision.
Success isn’t just found in the number of AI tools deployed. When managers know they’re accountable for how AI is used and what it produces, they behave differently than when AI is treated as an ungoverned experiment.
Accept that some work should stay human
You can’t automate everything.
Some processes are too sensitive, too ambiguous, or too relational to hand to AI today. Strong managers know when to say, “We do this with humans only,” or “AI drafts, but a human always finishes.”
Bad managers assume AI can handle everything, right up until it can’t.
AI Initiatives Make or Break Businesses
Right now, companies are scrambling to return on their AI investment. With rumors of a bubble bursting escalating, there’s no point in spending the money and not enabling managers to use it correctly.
Most companies aren’t struggling because AI is “not good enough.”
They’re struggling because they’ve underestimated the human element – particularly amongst their managers.
AI amplifies what it’s given.
In the hands of strong managers, it scales good judgment, clarity, and curiosity. In the hands of weak managers, it scales confusion, bias, and risk.
If you want your AI investments to pay off, don’t start with the model catalog. Start with the people you’ve put in charge. Equip your managers to lead hybrid teams of humans and AI, or be prepared for AI to show you, very quickly, who shouldn’t be managing at all.
Trusted insights for technology leaders
Our readers are CIOs, CTOs, and senior IT executives who rely on The National CIO Review for smart, curated takes on the trends shaping the enterprise, from GenAI to cybersecurity and beyond.
Subscribe to our 4x a week newsletter to keep up with the insights that matter.
