From Hype to ROI: What’s Actually Working in Enterprise AI

A Turning Point for Enterprise AI

On March 5, 2026, SEI Seattle brought together more than 80 leaders to answer one question: What’s actually working in enterprise AI?

“From Hype to ROI: What’s Actually Working in Enterprise AI” wasn’t a night of vendor talking points. It was a practitioner’s field guide, forged from lived experience, hard failures, and real wins — featuring executives and practitioners from Google, Microsoft, AIGovOps Foundation, and SEI.

AI Is Not a Tool — It’s a Tectonic Shift

Most organizations are still treating AI like software. Something to layer onto existing processes. That’s where things break down.

Only 5% of AI pilots deliver meaningful impact. Not because the technology fails, but because the approach does.

“If you’re looking at AI as a tool, you’re missing a giant mark. Imagine sitting in 1999 trying to bolt ROI calculations onto the Internet. You would have absolutely missed the mark.” — Antonio, Practice Lead AI & Technology, SEI

The panel drew a clear parallel to the Industrial Revolution, the internet age. AI is larger in scope and faster in speed than anything that’s come before.

Ravi reinforced the scale of the moment, drawing on 25 years watching technology reshape Microsoft.

“I’m in a consequential role, in a consequential company, at the most consequential time in history. How could you not be excited? And if you’re not also a little terrified, you’re living under a rock.” — Ravi, Corporate Vice President, Microsoft IDEAS

Fix the Foundation Before You Scale

Most organizations are trying to scale AI on top of weak data foundations.

Even at Microsoft’s size, Ravi shared how teams ran into inconsistent definitions, missing context and data not designed for machine use.

“Data is the fuel for AI. Most companies never actually invested in it. The starting line has moved way ahead, and they’re not going to catch up without fixing the data layer first.” — Ken, Founder, AIGovOps Foundation

Without a solid data layer, governance unravels.

Ken shared two real-world cases, not born of bad intentions, but of inadequate structure: 

• Litigation revealed that an insurer’s human-review step averaged just 1.2 seconds per claim.
• An autonomous agent deleted a production table, added synthetic data, and altered logs to hide the error.

What this means:

• Clean, structure, and add semantic context to your data
• Define owners and require human review for AI outputs
• Set up monitoring for your deployments to catch and resolve issues quickly

Trust and Adoption Come Down to People

Even the best AI fails if the experience doesn’t hold up — and if your culture isn’t ready for it.

Alix watched real users type a single word, “table,” expecting sophisticated data retrieval. The gap between what designers assumed and what users actually needed was significant.

“You get one shot. If your agent ships and doesn’t work well, users won’t come back. Make sure whatever you release does that one thing really, really well.” — Alix, Agentic AI & AI-Powered UX, Google

The deeper challenge the panel kept returning to was unlearning.

Ravi was direct: “We are obsessing about the code. We are not focusing enough on the culture.” 

Antonio pushed on this further, asking the audience how many use AI to write outgoing emails and how many use it to summarize incoming ones: “A lot of what we do in the enterprise is accumulated debt dressed as process.”

What this means:

• Focus on real user needs
• Rethink workflows, not just automate them
• Keep human judgment at the center

What This Means for Your Organization

The panelists closed the evening by distilling their experience into actionable guidance. Across their different vantage points — product, governance, data infrastructure, and delivery — five clear themes emerged:

Focus on one outcome first

Resist the temptation to let a thousand experiments bloom. Pick an entity, a kernel, a use case — and get it right. Success compounds.

Fix your data before you scale your AI

Semantic richness, freshness, quality, and governance are not post-launch concerns. They are prerequisites.

Govern from the start, not as an afterthought

Accountability structures, risk classification, and compliance integration are what separate one-time pilots from trusted, scalable capabilities.

Instrument everything and build for learning

Treat every deployment as Version 1. At the end of every AI session, ask the model how you can accomplish the same outcome in fewer steps.

Keep humans at the center

There will always be roles that are irreplaceably human: judgment, relationships, reading a room, holding the line. Protect that. Invest in people.

From Strategy to Execution, End to End

AI strategy without execution doesn’t deliver value. Execution without strategy creates waste. SEI brings both — and a proven methodology to get you there.

The SEI AI Transformation Approach:

01: Define a Path Forward Rigorous AI assessment and strategy — evaluating readiness, identifying high-value use cases, and building a clear roadmap aligned to your business goals.

02: Prepare the Organization Building AI literacy, managing culture change, and ensuring your people understand the real value and real limits of AI before you scale.

03: Experiment & Innovate Turning strategy into production-ready solutions — custom agentic workflows, vendor evaluation, and the data infrastructure to support each use case.

04: Sustain Value Embedding intelligent automation into critical processes, governing AI agents with rigor, and building the feedback loops that improve performance over time.

Across all four phases, SEI brings full-spectrum capabilities, allowing us to serve as a single, accountable transformation partner rather than a collection of specialized vendors.

AI & Technology • Concept to Delivery • Data & Analytics • Security, Risk & Compliance • Strategy & Operations

SEI Seattle: Where Strategy Meets Execution

Since opening in 2023, SEI Seattle has built a team focused on solving complex, real-world AI challenges across the Pacific Northwest and beyond. Seattle was a deliberate choice — it’s the epicenter of technology innovation in North America, and its entrepreneurial spirit matches our own.

This event reinforced what we see every day: organizations don’t need more AI ideas. They need partners who can help make AI actually work.

If you’re on your own AI journey and want to be part of this dialogue, we invite you to connect with the SEI Seattle team. Let’s talk!