Smarter AI Starts with Human Know-How

AI gets sold as a shortcut to efficiency. But the uncomfortable truth is that many companies don’t even know how their own work gets done. Processes live in silos, creating operational blindspots and data integration challenges that plague marketing teams, and all of that expertise walks out the door during turnover.

That’s where Vanessa Liu comes in. Before founding Sugarwork, a consultancy that helps companies turn employee knowledge into reusable intel, she spent years at McKinsey diagnosing organizational issues. She’s seen first-hand that if you want to improve performance, you have to start by talking to people.

It’s as true today as it ever was. Before you roll out AI, capture how work really gets done. Not the tidy swimlanes in your intranet, but the messy, human pathways where decisions actually happen.

“So much of how things are done is not necessarily captured in systems or in documents. It’s a result of how people interact with one another,” Vanessa explained. “Turning tacit knowledge into explicit knowledge is a really critical step because without that you’re literally shooting in the dark.”

Why AI Projects Can Fail Before They Begin

MIT’s recent finding that 95% of AI deployments never make it to production echoes her point: the problem isn’t the technology, it’s the lack of business context.

Vanessa has seen firsthand how organizations falter when they rush into AI without understanding the human foundation of their work.

“Knowledge doesn’t live in documents,” Vanessa told us. “It lives in people—and in the interactions between them.”

In other words, the failure is not because the models don’t work, but because the business context isn’t there. It’s like hiring a super-powered agent and never giving it orientation.

For marketing and analytics leaders seeking to transform agency operations with AI, Vanessa's advice is simple: treat every AI deployment like onboarding a new teammate.

“Treat any AI deployment like you’re onboarding a person,” Vanessa said. “You wouldn’t throw someone into the role without giving them context. Why would you do that with AI?”

Knowledge transfer, she argues, isn’t optional. It’s the scaffolding AI needs in order to be effective. Without it, companies risk wasted investments and automation projects that collapse under their own weight.

How to Bottle Operational Wisdom (Without a 3-Month Workshop)

Here’s how Vanessa’s team captures what’s in people’s heads—fast.

1. Map the work via interviews, not hunches. Sit down with leaders from each department. Ask, What’s your part? What inputs do you need? Where does it break?
2. Record everything. Drop an AI notetaker into Zoom/Meet/Teams. Transcribe, then synthesize across subject-matter experts.
3. Ship real artifacts. Turn the interviews into a 15-step process map, RACI, and checklists—in hours, not months.

One of Vanessa’s recent projects was with Supergoop, the sunscreen company. The COO asked her team to map the entire commercialization process—how an idea becomes a product on shelves.

On paper, Vanessa discovered each department knew its piece. But nobody had ever stitched the end-to-end process together. So Vanessa’s team ran structured interviews with department heads, recorded and transcribed everything with AI note-takers, then synthesized it into a 15-step process map complete with checklists and a RACI model. The result was a shared playbook that made cross-department dependencies obvious—and fixable.

“In the past, this would take three months,” Vanessa noted. “With AI, you can get there in hours.”

‍

Resistance, Turnover, and the Hidden Bill of Lost Know-How

Of course, documenting operational wisdom isn’t always easy. Not everyone loves the idea of documenting what’s in their head. Some fear it’s a step toward redundancy, especially during restructuring or times of change.

Others just don’t have the time. But when the “why” is clear, people lean in.

“It’s about sharing why you’re collecting this information. When people see it makes their lives easier—fewer delays, smoother collaboration—they’re eager to participate.”

‍

The bigger risk isn’t resistance, she suggested, but waiting until it’s too late. One anecdote stuck with her: interviewing a 70-year-old executive who kept stacks of binder-clipped notes, each representing seven figures in potential revenue. “No one else knew what was in them,” she recalled. “If he walked out tomorrow, it would all be gone.”

And make no mistake—the costs are real. Research pegs the loss of a single subject-matter expert’s knowledge at $350,000. That doesn’t include the rework, the projects dropped midstream, or the months it takes to ramp up a replacement.

The Future: AI Agents With Context

Looking ahead, Vanessa sees a parallel between today’s AI moment and GE’s Six Sigma revolution under Jack Welch. Then, companies unlocked efficiency through rigorous process mapping. Now, AI can accelerate that work—but only if it’s built on a foundation of human insight.

She imagined a future where companies run on “collective brains”—centralized repositories of operational wisdom that both humans and AI can tap into.

“The things that create value are implicit, hidden interactions. What we do is help surface them, make them visible, and then show you where automation can make the biggest impact,” Vanessa said.

‍

The payoff isn’t just speed or savings. It’s resilience—ensuring that when someone “wins the lotto” and leaves, their expertise doesn’t walk out the door with them.

Slow Down to Speed Up

The rush to adopt AI often skips over the hardest part: asking the right questions, listening to employees, and documenting what makes work actually work. Vanessa Liu reminds us that AI isn’t magic—it’s an amplifier.

“It’s the slow down to speed up mentality,” she said. “Invest the time up front, and you’ll get the payoff down the road.”

‍

For marketers frustrated with fragmented data and under pressure to prove ROI, it’s a mindset shift worth making. Efficiency isn’t about skipping steps. It’s about getting the foundation right so the next step actually sticks.

‍

Listen now: Spotify | YouTube | Apple

Guest Links: Vanessa Liu on LinkedIn | Sugarwork website

Subscribe to What Gets Measured for more conversations at the intersection of marketing, data and AI.