Dude Wipes typed three words. Rippling AI built a flight risk model from scratch — and nailed the number one risk.

The Challenge

Dude Wipes — the Chicago-based consumer brand best known for its flushable wipes and a Shark Tank appearance that landed Mark Cuban as an investor — has grown fast. When Tyler Qahhaar, Dude Wipe’s VP of People + Culture, joined in August 2023, the company had around 15 employees. Today it has nearly 60, with a culture that has remained deliberately low-ego and collaborative through the growth.

The people function that supports all of that is a team of two: a VP of People + Culture and one full-time People Business Partner. The upside of that lean structure is that everyone is close to the business. The constraint is that there's not a lot of bandwidth for manual analytical work — which means that important but non-urgent tasks tend not to get done until they become urgent.

Flight risk analysis is a good example. Dude Wipes has been unusually lucky on retention. Turnover has been low, and for a long time, the team was young enough in tenure that the typical attrition pressure points hadn't arrived yet. But as the company has grown and employee tenure has climbed to three and four years, new layers of org structure have appeared, managers have changed, and the conditions that typically correlate with departure have started to accumulate.

The manual alternative — pulling compensation data, performance history, engagement survey results, tenure records, and manager change logs, downloading them, stitching them together, and then analyzing the result — was technically possible. But getting to a starting point for analysis alone would have taken three to four hours. For a two-person team running the full people function at a 57-person company, that's a project that keeps getting scheduled for later.

The Solution

The prompt was as simple as it gets:

“Who’s a flight risk?”

No rubric. No defined criteria. No uploaded data. Just the question, asked of a system that already had everything it needed to answer it.

Rippling AI first checked whether a flight risk score or custom field already existed in the system. It didn't. So rather than returning an error or asking for more context, the system identified what data was available and built its own rubric on the spot: compensation ratio, employees who hadn't seen a title or level change in two or more years, engagement survey results, whether a manager had changed recently, and whether the employee was in the two-to-four year tenure window — the period where attrition pressure typically peaks. It surfaced the top 25 highest-risk employees ranked against that rubric, unprompted, in about five minutes.

The result was accurate. The one employee the team had already internally flagged as the most likely to leave was at the top of the list.

That's the point that matters most. The prompt was three words. What made it work was that the answer already existed inside Rippling — compensation, performance history, tenure, engagement data, organizational change — all in one place, all automatically accessible, none of it requiring a manual export or a trip to a separate system. Most companies attempting the same analysis with a generic AI tool would first spend hours assembling the data. Here, that step didn't exist.

There's a second dimension to why the context layer matters: consistency. Because everyone at Dude Wipes is working from the same shared data inside Rippling, asking the same question returns the same answer regardless of who asks it. That predictability is part of what builds trust in the output.

And because the analysis happened inside Rippling rather than in an external tool, there was no round-trip. The insight and the action lived in the same place. Tyler was already thinking about whether the flight risk scores could be written back as custom fields on employee profiles — so that the analysis wouldn't just be a one-time report, but a living data point that compounds in value over time.

A second prompt during the same session illustrated the same dynamic at smaller scale. Dude Wipes was sending branded merch to remote employees and needed a quick t-shirt size list. A simple query returned 18 names — four more than expected. It turned out those employees had updated their home location to Chicago but still had a stale "remote" work location tag on their profile. The discrepancy would have been invisible if the data had been pulled manually, profile by profile. Instead, it surfaced immediately, and was fixed right inside Rippling AI without navigating to a single individual record.

The Impact

The flight risk analysis that would have taken about four hours without Rippling AI took five minutes. The rubric was self-constructed, the ranking was accurate, and the output was immediately usable for a conversation about retention priorities and top performer overlap.

But the more durable impact is structural. For a two-person people team running HR at a fast-growing company, the difference between a task that takes half a day and one that takes five minutes isn't just time savings — it's the difference between something that gets done proactively and something that only gets done in response to a problem. Flight risk analysis was in the second category. Now it isn't.

Half a day to five minutes. A flight risk analysis that would have required assembling data from multiple sources over four hours was returned as a ranked list of 25 employees in under five minutes — with no rubric provided and no data uploaded.

The rubric built itself. With no predefined flight risk framework in the system, Rippling AI identified the available data — comp ratio, tenure, title changes, manager changes, engagement scores — and constructed its own scoring model. The number one result matched the team's internal assessment.

Stale data surfaced automatically. A routine t-shirt size query returned four more remote employees than expected, revealing location tags that hadn't been updated when employees relocated to Chicago. The discrepancies were found and fixed without leaving the platform.

Insight and action in the same place. Because the analysis happens inside Rippling rather than in an external tool, there's no export-analyze-return loop. Insights can be actioned immediately — and written back to employee profiles, where they compound in value for future decisions.

Most conversations about AI focus on the model. Tyler's take cuts to something more fundamental: the model is only as useful as the data underneath it. For HR teams sitting on years of people data spread across disconnected systems, that's the problem. For teams running inside Rippling — where compensation, performance, org history, and employee lifecycle data all live in one connected layer — the problem is already solved. The prompt is almost beside the point.