For the past few years, the corporate playbook for artificial intelligence has focused on making individual employees faster. Leaders armed workforces with generative assistants, expecting a swift bottom-line transformation. Instead, they hit a wall.

Data from MIT's Project NANDA found that 95% of generative AI pilots delivered no measurable impact on the bottom line. Only 5% produced real organizational value.

Here's why: Enterprises have been investing in ad hoc intelligence (tools making an individual faster at a task) when they should be building institutional intelligence (the compounding capability making an entire organization smarter over time).

What Smartsheet is saying: "Sure, getting individual tasks done has gotten easier, but working across systems and across teams still hasn't," says Drew Garner, Chief AI Officer at Smartsheet.

• "AI exposed a structural debt enterprises have been carrying for years. The fragmented stack was always a liability. People were the connective glue that papered over the architectural gaps. Now that AI needs connected context to be useful, and enterprises need AI to stay competitive, the debt has come due."

The reality check: This isn't a future-state problem. Employees are already pasting company work into consumer AI tools, with or without IT's blessing. The AI they're using can't tell what's current from what's stale, what's approved from what's a draft or what that person is even allowed to see.

That leaves leaders with a false choice: Lock AI down and forfeit the value, or open it up and forfeit control. The organizations getting past the 95% wall are the ones rejecting the trade-off and connecting AI to a governed picture of their work, where every read is scoped to the person asking and every action leaves an auditable trail.

The challenge: The illusion of productivity

When you inject powerful AI models into an organization without a centralized governance standard, individual wins stay personal. Instead of transformation, you get two distinct workplace liabilities:

• The Hoarder: This person quietly masters prompts, ships work twice as fast and teaches no one. But a private superpower is just a single point of failure with a good attitude. If they take a week off or quit, that speed walks out the door.
• The Slop Cannon: This employee mistakes volume for value. Armed with a chat window, they fire off endless markdown, half-built HTML and 12-page documents nobody asked for. AI didn't make them better; it just made them faster at being wrong.

Both are symptoms of the same root cause: AI with access but no governance, no shared standard for what's true, what's allowed and what good output looks like.

The strategy: Building institutional intelligence

To move past fleeting productivity spikes, organizations must capture and unify three layers of knowledge that traditionally lived only inside employees' heads:

• Context: The live map of your work — knowing what is connected to what, who owns which asset and the exact state of things across systems.
• Intent: The "why" behind execution — understanding what outcomes actually matter and identifying early signals when work drifts from the plan.
• Judgment: The nuanced decision frameworks carried by your most experienced leaders — knowing when to escalate, when to wait and when the rules say one thing but the right answer is something else entirely.

What you're missing: As time goes on, this data usually evaporates. AI deployment without this connected architecture cannot compound; it resets to zero the moment a project ends or a team reorganizes.

The solution: Turning the playbook into software

Historically, corporate best practices went to die in unread wikis or forgotten SOP documents. Everybody wrote the playbook; nobody ran it. Rather than asking employees to remember the playbook, organizations can build it directly into the tools and processes they use.

How it's done:

• Smartsheet brings order to complex work by helping teams enforce corporate best practices directly in the work management platform. That structure makes context and intent clear — giving both people and AI the foundation they need to execute work consistently.
• Smartsheet shifts this paradigm with skills: packaged, proven intelligence about the most efficient way to get work done in our platform, built and served by Smartsheet as shared, server-side infrastructure. Rather than forcing individuals to figure out prompting in a silo, this expertise is applied automatically — running invisibly in the background for every connected human and AI agent, so the whole team works from the same standard by default.

• When a task begins, the skill goes to work: it hands the model its role, reads what you actually mean, picks the right tool, formats the output to a consistent standard, and applies the rules for how the work should be done. An expert in the room, supporting you in the background on every task.

Even better: This expertise is combined with the platform's live map of relationships and agent memory. It doesn't offer generic best practices in a vacuum; it delivers the exact right move for a specific project, accounting for active dependencies and team history.

The impact: A collective surface

Organizational velocity dies during the handoff. Work is a relay sport passed from human to human, human to agent and team to team. A genius operating in a silo is still a silo.

For a team to move faster, speed must survive the handoff. That requires a single, accurate view of work where humans, agents and external vendors see the same state at the exact same time. While AI models are probabilistic, the underlying enterprise record must remain exact — the row updated, the approval sent, and an auditable record left behind.

Key numbers: This set of activities is already operating at an immense scale. On Smartsheet, more than a quarter-billion automation rules fire quietly in the background. In a single month, the platform processes over 170 million handoffs across more than 300,000 distinct organizations.

• When an AI agent joins this environment, it doesn't face a blank canvas. It plugs directly into a living graph of work that already runs the business.

The takeaway: True AI return on investment won't come from fragmented productivity tools or the models themselves. It comes from governing the connection between AI and your work: AI that sees what's true, respects who's allowed to see it and acts inside the rules you've already set. That intelligence compounds in your governed work state, not in anyone's model. Swap models tomorrow; the intelligence stays. It's yours.

Get more details from the experts.