Wait, it can do what? Five customers put AI to work in Smartsheet at ENGAGE 2026

Dan Humphries of Lilly Kinsale had a clear message for anyone rushing into AI: it won’t fix inconsistent operations, it intensifies them. Anyone who has watched an AI pilot stall on messy data, or on a process no one fully owns, knows it’s true.

ENGAGE 2026 AI use cases at a glance

• Lilly Kinsale: Claude builds one live changeover view on top of Smartsheet, saving roughly 30 to 35 minutes per core team member every day.
• Swampfox: Claude proposes RAID log updates on a triage sheet, and the customer approves, denies, or reworks each one.
• Wizards of the Coast: a Claude skill checks up to 14,000 card faces on a schedule and reports only what’s missing.
• Compassion International: 35 sheets and 35 reports became 3 reports, rebuilt in 30 minutes.
• Vista: one workspace, 100+ users, and 75+ automations handle transfer agreements, with no AI needed to start.

Intelligence got cheap. Execution didn’t. 

AI made intelligence abundant. Models get better every few months, and almost anyone can now build something useful with them. But Claude, Copilot, ChatGPT, Gemini, Amazon Quick, and the rest have read the far corners of the internet, but not how you work. They don’t know your team’s process, what’s due Monday, who owns it, or what changed over the weekend. 

So work got faster, but nothing got more aligned. Everyone prompts on their own, outputs pile up that nobody has verified, and no one can say who signed off. Every ungoverned AI output is an unlogged decision, and left alone, AI does not compound value. It multiplies fragmentation.

The fix isn't a smarter model. It's what people design around it: who signs off, who's next to approve, where AI should stop. The question isn't which model you buy. It's where the work lives, and who's allowed to touch it.

Five customers showed us what that looks like in practice. It starts at enterprise scale, with Lilly Kinsale, where a complex manufacturing process now runs from a single shared view. Then four builders from the Smartsheet Community show how to get there, from the easiest first step to work where every number has to be right.

Five customers, five ways to make AI to work in Smartsheet 

One view of the work, before the morning meeting (Lilly Kinsale)

Saves roughly 30 to 35 minutes, per core team member, every day.

Lilly's Kinsale site in County Cork, Ireland makes active pharmaceutical ingredients. Every time it switches from one medicine to another, the changeover spans safety, cleaning, maintenance, automation, and operations. Before, the data lived in five systems and on paper, so planners gave three to four half-hour status updates a day, retelling the same story each time. Now Smartsheet holds every changeover plan, and Claude builds a live view on top of it through the Smartsheet connector (MCP, the plug between an AI tool and your sheets). Teams post updates, the view refreshes before the morning meeting, and the whole floor works from the same display. Dan's advice: standardize your plans first, then layer on AI.

Key takeaway: Standardize cross-team plan structures before adding AI.

Nothing changes until Will says so (Swampfox)

AI proposes. He decides.

Will Jeffords is a consultant and Smartsheet Community Overachiever who works inside his customers' programs. Will maintains a RAID (risks, assumptions, issues, and dependencies) log for one of his large customers, which used to eat hours every week. Now, he uses Claude to review his schedule sheet and suggest updates to the RAID log. These recommendations are recorded on a separate triage sheet, where he can approve, deny, or rework each one, stress-testing them further with the help of AI. The time savings from this have been significant, and now, sixty out of the seventy people on this project have Smartsheet licenses. 

Key takeaway: take one thing you update all the time, and let AI propose the updates on a triage sheet.

Up to 14,000 card faces, checked in the background (Wizards of the Coast)

Only what's missing, reported while the team gets on with other work.

Every Magic: The Gathering release has 8,000 to 14,000 card faces, and the fans notice every mistake. Jean-Alix Duplantier, a creative manager with 22 years as a graphic designer, and the team would manually compare release artwork in a Smartsheet sheet against the print server (tool/product/software?), line by line. Now a Claude skill (a saved set of instructions, so AI does a job your way every time) does it and reports only what's missing. Then the team asked, "Could this run by itself?" It now runs on a schedule. Wizards has a strict rule: no generative AI in creative work. AI checks. People create.

Key takeaway: put a routine manual check on a schedule with AI.

From two weeks to 30 minutes (Compassion International)

35 sheets and 35 reports were reduced to 3 reports.

Matt Austermiller has 15 years of experience with Excel and VBA and about 3 months with AI. Last December he built one leadership timeline across dozens of projects the hard way: 35 sheets, 35 reports, two weeks of work, and a fragile solution. This year his team gained access to Amazon's Kiro, connected it to Smartsheet , and he rebuilt the entire solution in 30 minutes. It’s now more robust and runs as a scheduled task (an agent that works on its own at a set time) at 8 a.m. every day with almost no upkeep. He borrowed the idea from Cicero, Smartsheet's Chief Evangelist, a builder from Day 1. Not all of the wins were attributed to AI: adding a new multi-select column replaced 35 checkboxes, ensuring better data quality and auditability.

Key takeaway: refresh your existing solution with minor optimizations and AI to make a major impact. 

Where close enough isn't close enough (Vista)

One workspace. 100+ users. 75+ automations.

Claire Williams leads a finance team at Vista Equity Partners, a private equity firm, where a wrong number has real consequences. Her team handles transfer agreements: one finished agreement sets off a chain of tasks across an organization of more than 750 people. Claire mapped the ideal process first, then tackled the hardest problem and added a few simple automations. It eventually grew into one connected workspace with controls, permissions, and links to Salesforce, eFront, and ledger data. Her point: complex finance work can be controlled and automated in Smartsheet alone, mostly with features people already have. She didn't start with AI. She started by putting the work, the rules, and the approvals in one place.

Key takeaway: map your ideal process first, then find the tool that fits it.

The job title that doesn’t exist yet

Drew Garner, the Smartsheet Chief AI Officer, asked customers at ENGAGE to watch for a job title that doesn't exist yet. He called it the “workflow architect", that is, the person who decides where AI can help, where a human stays in charge, and how it all leads to real results. 

These customers are workflow architects in their own right. Dan's team standardized the plan before adding AI. Will formalized the human sign-off process for AI’s recommendations. Jean-Alix set clear boundaries around what AI can and can’t do. Matt updated his solution configuration before setting up AI to handle recurring tasks. Claire mapped out her ideal workflow and rules before adding AI.

The people and teams best placed to succeed already know how work really moves, who approves what, where handoffs break down, and where data goes stale. The Execution Hub is what empowers them: agents draft and check, humans decide and approve—all in one governed place.

Start today

Choose one thing you update regularly and create a triage sheet for it. Let AI make suggestions, then decide what to use. You don't need to learn everything about AI first. Start simple. Not sure where to start or how to build it? The Smartsheet Builders Series walks through the options for choosing when and how to use Smart Assist, automations, MCP, and the API.

Missed the keynote? Learn more about the Smartsheet innovations announced on Day 1.