By Drew Garner, chief AI officer, Smartsheet

For two decades, the most valuable parts of enterprise software sat behind a barrier: workflows built around how the organization actually ran, automations eight steps deep, judgment calls baked into every exception. That's real depth, the kind serious work depends on and the kind that took weeks, months or longer to master.

• The barrier was never that you had to be technical. It was that you had to climb the curve to reach the value or sit next to someone who already had.
• Most people opened the tool, hit a wall, and went back to a spreadsheet. The value was real. It was just locked to the few who put in the time.

🔒 The outlook: AI picked that lock. It changed which side of the keyboard holds the complexity: You describe the outcome in plain language, and the agent translates it into the right steps. The depth stays. The climb disappears.

• We've now watched what happens when that barrier comes down at scale and across thousands of organizations, and the most important finding isn't about productivity. It's about who showed up.

The people who walked through weren't the experts

In our internal analysis of AI-enabled work across organizations, one pattern stood out: the people embracing these capabilities now weren't necessarily traditional power users.

Key numbers: 87% of users who had only ever viewed information — never contributing to it — became active creators once AI-enabled workflows were in place. Some of the most active builders today weren't building anything six months ago.

• Of these users, 89% used AI to create or update work, not just ask questions. The more of a system they reach into, the more they're building rather than querying.

Here's what else: The locked-out population was enormous. Many of the seats that came alive through an agent hadn't been touched in quarters.

• This was found across project managers, directors, owners, and CEOs rather than a new class of technologist, and mirrors broader industry findings.

This is the part that should reframe how every company reads its own usage data: The demand was never missing. It was dammed up behind a barrier, and most of it never showed up in a single dashboard.

Why your foundation is now the whole game

Here's what most organizations are missing in the rush to roll AI out.

When only specialists could reach your systems, a weak foundation was contained. A handful of careful experts navigated around the broken workflows, inconsistent data and ungoverned processes.

• They knew where the problems were and worked around them. The barrier that kept everyone else out was also protecting you from your own mess.

Okay, but: Remove the barrier, and that protection is gone. Every gap in your data, every disconnected workflow and every process that no one ever governed is now accessible to everyone in the building, including those who have no idea the gaps are there.

• AI didn't make your foundation less important. It removed the last thing standing between a bad foundation and your entire organization.

📈 The impact: This is where the familiar warning finally earns its weight. AI doesn't fix a broken process. It exposes it, and then it scales it.

• With a strong foundation underneath, removing the barrier is the best thing that's ever happened to how your organization works. The depth you spent years building is now within reach of everyone who needs it.
• With a weak one, you've handed the whole company a faster way to multiply the dysfunction. Same technology, opposite outcome. The difference is entirely what's underneath.

A foundation built for access, not navigation

The old foundation only had to be good enough for experts to navigate. The new one must be good enough for anyone to reach safely. That's a higher bar, and it changes what you build toward.

The strategy: It starts with a shared system of record where work, data, decisions and the people behind them actually connect.

• AI can't produce reliable answers when the work itself is scattered across disconnected tools. Without shared context, even the best model is guessing. Your data foundation matters as much as your models, and when access is broad, it matters more.

🚧 On that foundation, the guardrails that used to be optional become structural:

• Governed data, so that when anyone can query it, the answer is the one your experts would trust.

• Clear boundaries for what an agent can and can't do.
• Human checkpoints at the moments that matter, because broad access means more people are acting on outputs they can't independently verify.

The goal isn't to slow AI down. It's to make sure that when the whole organization can access your systems, what they access is worth accessing.

The experts moved

It's tempting to read all of this as expertise becoming obsolete. The opposite is true. Experienced judgment matters more now, not less. It just moved.

Here's how: The specialist used to be the gate because they were the only ones who could reach the depth, so everything flowed through them.

• Now an agent reaches it for everyone, and the expert's value moves up the stack: from operating the system to validating the judgment.

AI responds fast, but it can't tell you whether the response reflects the full picture, whether context is missing or whether someone asked the wrong question. A system can tell you why one project is over budget.

• An experienced person recognizes when the same problem is showing up across five of them, and what that pattern actually means.

Why it's important: It's critical to design systems around what we call the "expert in the middle." The agent analyzes and recommends; people stay responsible for validating the output and making the call, especially when the stakes are high.

• And these systems only get better when people actively shape them by connecting trusted data, correcting errors and refining output to match how decisions actually get made.

I use AI every day. It works for me because I've put in the time to train it inside my own environment. The barrier didn't remove the need for expertise. It freed experts to spend it where it counts.

The barrier is gone: The foundation is the decision

For twenty years, the difficulty of your systems was that they accidentally protected you from their flaws. That protection is gone, and it isn't coming back.

The takeaway: The question is no longer whether your people can use your software; they can, all of them, just by asking. The question is whether what they reach when they ask is something you'd be proud to scale.

• Get the foundation right, and removing the barrier is the best thing that's ever happened to your organization's operations. Build on sand, and you've just given everyone a faster shovel.

AI didn't make the foundation matter less. It made it the whole game.

Learn more.