Shaz Khan is the CEO of Vroozi, a procurement platform helping businesses modernize how they manage spend and supplier relationships.
getty
Nearly every week, I talk to executives who are convinced they need an AI strategy. They're right. But most of them are asking the wrong first question. Instead of "What AI should we adopt?" the question should be, "Are we actually ready to adopt it?"
Most organizations treat those as the same conversation, and that's where things go sideways.
Over the past several months, I have worked with procurement and finance leaders at companies of all sizes. They've seen the demos, heard the pitches, understand something significant is happening and feel the pressure to move.
What most of them are missing is a clear sense of where to start and what has to be in place before the technology can deliver what the vendors are promising. So here's what I tell them, in order.
AI is only as good as what you feed it. If your procurement data lives in five different systems, if your supplier records are incomplete, if your approval workflows exist mostly in people's heads and email threads, no AI layer will fix that. Garbage in, garbage out—only faster.
The first step to meaningfully adopting AI is to spend months—often uncomfortable months—getting honest about the state of your data before anything else.
AI adoption fails at the top more often than most people expect. When a CFO and a CPO have different definitions of what success looks like, the people below them spend months navigating that ambiguity instead of making progress.
Before any vendor conversation, the executive team needs to agree on two things: What problem are we solving, and how will we know if we've solved it? Those sound like obvious questions. In my experience, most teams haven't actually answered them together in the same room.
The instinct, especially when there's board pressure to show AI momentum, is to launch broadly with pilot programs across multiple departments, lots of announcements and dashboards showing activity.
What that usually produces is a lot of activity and very little insight into what’s actually working.
Pick a structured workflow (e.g., sourcing a new supplier, creating a purchase order, processing and paying invoices) where the pain is most acute and the data is cleanest. Go deep on that one and measure it carefully. Let it generate enough confidence internally to earn the right to go broader.
Breadth can come later, and when it does, you'll have a proof point that actually means something.
This is the step most organizations underinvest in, and it kills more AI initiatives than bad technology does.
A 2025 Predictive Index survey of 1,000 U.S. professionals cited by HR Dive found that 68% of employees want more AI training opportunities, which ranks higher than job guarantees or promotions. Most of the workforce isn't afraid of AI, but they do want the skills to use it effectively.
The people closest to these business processes have genuine questions about what AI means for their jobs and their daily routines. If leadership doesn't address those questions directly and honestly, you'll get passive resistance that looks like adoption and isn't.
The best implementations I've seen take employee adoption as seriously as the technology itself. They invest in governance councils that help evaluate AI investments and establish guardrails around their use. Those who use these tools every day have to believe the technology is making their work better, and that they have a future in the organization using it.
Here's one thing that media coverage of AI mostly glosses over: The AI being developed today is moving faster and doing more than most public narratives reflect.
That's an opportunity that comes with a real risk. Many experts, such as University of Virginia economist Anton Korinek, who sits on Anthropic's Economic Advisory Council, are warning that most companies are not taking AI advancements seriously enough.
As these generative tools take over more of the mechanical and even analytical work, the humans working alongside them have to be disciplined to catch errors, ask the right questions and push back when something is off. Critical thinking is a skill that gets rusty when you stop using it. The companies I'm most impressed by right now treat that as an operational risk, because when the AI gets something wrong, and it will, you need people sharp enough to catch it.
When AI takes over a manual process, that time goes somewhere, and what you do with it is a choice worth making deliberately.
Companies that reinvest it in strengthening supplier relationships, solving business problems or tackling projects that never seemed to fit into the day will make the most of AI adoption. Companies that simply absorb it into cutting headcount and calling it efficiency will likely wonder later why morale didn't improve along with the margins.
Before you automate anything, decide what you want people to do with the time they get back. Everything else in the rollout follows from that answer.
None of this works if you skip ahead. What I keep telling clients is that the preparation matters more than the selection. By getting clean data, aligned leadership and a workforce that trusts the process right, the technology choice can almost take care of itself.
In thinking about where AI is headed, I feel both encouraged and a little bit unsettled. The technology is moving faster than most organizations are ready for. Closing that gap starts with the unglamorous preparation work most companies want to skip. A rushed pilot won't get you there. That's been true of significant operational change, and AI is no different.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。