An Operator’s Perspective on AI
-- by Erica Akitani-Bob
Almost every subway ad in NYC is occupied by a personal injury firm or a newly funded AI startup. A few years ago, every conversation seemed to orbit crypto. Now it’s AI founders, AI copilots, and AI dinners. Somewhere between an NYU lecture on artificial intelligence and a Chief of Staff event downtown, I found myself comparing notes with operators across media, venture capital, private equity, and AI companies about how these tools are actually showing up in their work today.
I didn’t go to these events solely to understand how other operators were using AI, but after enough conversations, it became clear that no two organizations were approaching it in quite the same way. In some places, AI is already embedded into daily operations. In others, companies are still trying to determine where these tools fit operationally, culturally, or from a governance perspective. Most seemed less focused on finding a perfect tool than on figuring out where AI meaningfully improved processes and execution.
Below are perspectives from operators across a range of industries and company stages, each offering a candid look at how AI is beginning to reshape their work.
Director of Operations at a Next Gen Law Firm (that’s me!): “We’re very open to exploring tools that could improve the way we work, but ironically, some of the riskiest products we evaluate aren’t legal AI tools. They’re operational platforms with AI embedded into them. The deeper these systems integrate with our workflows, the greater the potential exposure to privileged information. Most products look impressive in a demo. The harder questions come later: where does the data go, what gets retained, who has access to it, and how does adopting this tool change our broader risk profile? At the same time, it’s difficult to ignore how quickly the technology is improving and how materially it can reshape internal operations when implemented well. Many publicly available AI tools still aren’t designed with highly regulated industries in mind, which makes adoption far more difficult in practice.”
Executive at a publicly traded media company: “Your ability to get real value out of an AI model is entirely dependent on the data you’re feeding it and whether you actually understand the system you’re trying to improve in the first place. We implemented an AI platform for automated creative development that reduced a process that historically took twelve weeks down to a matter of days. What the implementation also revealed, though, was how much of our original timeline had very little to do with the actual work itself. A lot of the delay came from fragmented ownership, unclear handoffs, and people brute-forcing projects across the finish line. In some ways, the AI didn’t just accelerate the creative process — it forced us to finally understand how the work was getting done in the first place.”
Operator at a fintech company: “AI is speeding up our engineers, but it’s also expanding who counts as one. We recently ran an internal AI usage survey, and 60% of non-engineers had already built something reusable. People in ops, finance, and customer-facing roles are shipping dashboards, automations, and internal tools without ever opening a ticket. My favorite recent example: someone on one of our ops teams (not an engineer) built a small internal tool that solved a recurring process issue we'd been working around with engineer time for a while. It took it off the engineering team's plate completely. It wasn't an obvious fit for that team to own. They just did.”
Operator at an AI company: “I’m one of the few non-technical people at a company largely made up of AI engineers working closely with some of the leading AI labs. A lot of my day is spent inside our own systems, using agents to audit how I’m spending my time, surface friction points, and identify which parts of a workflow are repetitive enough to automate next. Tasks that used to take days can sometimes be reduced to hours, which completely changes the rhythm of the work. But the work never really stabilizes. Once one process becomes more efficient, attention just shifts somewhere else: hiring, coordination, process design, edge cases, escalation paths. There’s always another layer of work that suddenly becomes the bottleneck.”
Operator at a private equity firm: “AdOps and routine service delivery work in general are ripe for automation. Last year, one of our portfolio companies built an agent that now handles three quarters of their operational tasks. It’s allowed the Client Services team to spend far more time on strategy and higher-complexity work rather than repetitive execution.”
Operator at a celebrity-led venture capital firm: “I think the conversation around AI often focuses on big, transformative use cases, when most of the immediate value I’m seeing is far more operational. Honestly, I still think a lot of companies, including ours, are underutilizing it. In my role, it’s been most useful for synthesizing information, organizing internal knowledge, and automating repetitive tasks that would otherwise slowly accumulate. None of it feels especially flashy, but removing that kind of operational drag compounds quickly over time.”
What stayed with me after these conversations was that none of the operators described AI as eliminating the need for human involvement. If anything, the opposite seemed true. As execution becomes faster and more compressed, the importance of judgment, coordination, prioritization, and operational clarity only becomes more visible. The repetitive parts of the work may shrink, but the responsibility of deciding what matters, what gets escalated, what gets automated, and how systems actually function still sits with people. In many ways, AI didn’t seem to be removing operators from the equation so much as making good operators even more important.
