What the NVIDIA AI Advisory Board Taught Me About Enterprise AI Adoption
79% of organizations are struggling with AI adoption right now, according to WRITER’s 2026 enterprise AI survey. I sit on NVIDIA’s Global Enterprise AI Partner Advisory Board, and I can tell you the number tracks with what I see across six industries, every week.
Most executives don’t have an AI problem. They have an adoption problem wearing an AI costume.
When AMSYS was named NVIDIA’s State and Local Government Partner of the Year this March, and I joined NVIDIA’s Global Enterprise AI Partner Advisory Board as a founding member in December, both came from the same place: 22 years of watching technology rollouts succeed or fail based on decisions that have nothing to do with the technology itself. The advisory board seat just gave me a wider aperture, a view across dozens of enterprises instead of the handful I touch directly through AMSYS.
Here’s what that wider view has confirmed, and what it’s changed my mind about.
The 79% Aren’t Failing at AI. They’re Failing at Sequencing.
That WRITER survey found something that should worry every CEO more than the headline number: 75% of executives admit their AI strategy is “more for show” than actual guidance. Nearly half, 48% — call their AI adoption a “massive disappointment.”
I’ve sat in enough advisory board sessions now to see the pattern. Companies buy the tool before they’ve defined the workflow. They announce the AI initiative before they’ve decided who owns the data quality problem underneath it. They chase the demo instead of the deployment.
At AMSYS, before we touch a client’s AI rollout, we ask three questions first: What decision gets faster or better because of this? Who is accountable when the model is wrong? What breaks if this goes down? If a client can’t answer those, the AI project isn’t ready, no matter how good the pilot looked.
So, what: the fix isn’t a better model. It’s better sequencing. Strategy, data readiness, and accountability come before the tool gets selected, not after.
Governance Is the Actual Bottleneck
Here’s a number from that same survey that gets less attention than the ROI stats: 67% of executives believe their company has already suffered a data breach tied to unapproved AI tool usage. That’s not a future risk. That’s a confession.
I’ve built a career on the cybersecurity side of this equation, AMSYS runs security operations for clients across healthcare, finance, energy, and government, sectors where a governance gap isn’t a headline, it’s a fine or a breach notification. What I’ve learned advising at the NVIDIA level is that this problem scales with company size, not shrinks. Bigger enterprises have more shadow AI, not less, because more employees are quietly running ChatGPT or a coding copilot against sensitive data with zero visibility from IT or security.
The advisory board conversations that matter most aren’t about which GPU architecture to standardize on. They’re about how enterprises build an approved-tools list fast enough that employees don’t route around it.
Key takeaway: the three governance moves I recommend to every client before their next AI rollout:
- Inventory shadow AI first. You cannot govern what you haven’t found. Most companies are surprised by what’s already running.
- Approve a short list, fast. A 90-day governance review kills adoption. A 2-week approved-tools list with clear data-handling rules doesn’t.
- Assign a single accountable owner per AI system. Not a committee. Committees don’t get paged at 2 a.m. when something breaks.
The ROI Gap Is a Measurement Problem, Not a Technology Problem
Only 29% of companies see significant ROI from generative AI. Only 23% see it from AI agents. Meanwhile 59% of companies are spending over $1 million a year on AI. That gap between spend and return is the conversation I have most often, whether it’s at an advisory board table or a client’s boardroom in Houston.
The pattern I’ve seen: companies measure AI ROI the way they measured software ROI a decade ago, adoption rate, license utilization, feature usage. Wrong metrics. AI ROI shows up in cycle time, error rate, and headcount that doesn’t need to grow as fast as revenue does. If you’re measuring logins, you’ll conclude AI doesn’t work. If you’re measuring how much faster a claims team closes a case, you’ll see the actual number.
This is also where the AMSYS multi-vertical view helps more than any single-industry playbook could. What works to prove ROI in a hospital billing department looks nothing like what proves it in an energy company’s field operations, but the discipline of picking the right metric before you start is identical everywhere.
What I’d Tell My Younger Self, and What I Tell Founders Now
Twenty-two years ago, I started AMSYS with $100 and a conviction that IT infrastructure was a trust business before it was a technology business. That’s turned out to be even more true for AI than it was for managed IT.
The advisory board seat hasn’t changed my strategy. It’s confirmed it, at a scale I couldn’t see from inside one company. The enterprises pulling ahead on AI right now aren’t the ones with the biggest budgets. They’re the ones who sequenced governance and accountability before they sequenced the rollout, and who are honest enough internally to admit when a pilot isn’t working instead of dressing it up for the board deck.
If 54% of C-suite executives say AI adoption is “tearing their company apart,” the other 46% are worth studying. In my experience advising across dozens of enterprises now, the difference usually isn’t the technology they picked. It’s the discipline they applied before they picked it.
Referenced: https://writer.com/blog/enterprise-ai-adoption-2026/, citing a survey with Workplace Intelligence of 2,400 global leaders.
Khalid Parekh
CEO, AMSYS Group
Houston, Texas