Purple Innovation runs its commercial conversation analysis through AI Workers

Purple Innovation is a science-based comfort company. The brand designs and sells mattresses, pillows, and other sleep products, with comfort engineered from its own materials. Customers reach Purple through its website, through retail showrooms, and through an inside sales team that handles chat, phone, and email.

Every one of those conversations carries signal. Customers say why they hesitate, what they think of a promotion, where a product falls short, and what went wrong on delivery. Reading that signal at the speed of the business is the job Purple's analytics and customer care teams set out to solve.

The challenge: insight locked inside the interactions

Purple's customer conversations always held the answers. The constraint was access. On the company's prior speech and text analytics tool, getting insight out meant downloading every result, working call by call, and assigning people to mark findings on spreadsheets. A single deep dive ran close to a month.

The old tool also demanded a specialist. As Jane Greiner, Director of Operations of Customer Care, put it, the team "needed to have someone who almost had a degree in how to use the tools that we had." Purple wanted analysis that more people across the company could run without a dedicated expert.

Why Purple chose Level AI

Purple evaluated several vendors. Two things decided the call. The first was direct access to insight rather than a manual path to reach it. The second was an interface people across the company could use without specialist training.

"We felt like it surfaced really great insights. It was simpler, and we were able to use it." — Jane Greiner

How it works: one worker, plain language, precise filters

Most of Purple's weekly analysis now runs through a single Conversation Research Worker. Angie McDonald asks a question in ordinary language, scopes it to the right team, channel, and time period, and reads the result the same day.

The work that once needed a tool expert now sits inside a plain-language request. Analysts can separate sales conversations from support, focus on a single channel, or look at one week, without learning a query syntax. That accessibility was the capability Purple set out to buy.

Reading the full funnel from conversations

Purple uses the worker across the whole commercial journey, from first interest through the return window. A handful of recurring reads show the range.

Demand

The team starts upstream, at interest. A standing analysis tracks how often customers raise a given product in sales conversations, measured against the same period a year earlier. The report tells Purple whether buying interest is climbing or cooling, in the customers' own words.

Promotions

Every promotional period gets the same treatment. The team gets scheduled AI Worker generated reports on what customers say about pricing and offers during a sale, and groups the themes. Each promotion is a hypothesis, and the read tests it against what customers said.

The output is a decision. Ahead of Black Friday, this analysis showed customers holding off in case a better deal landed on the day. Purple answered with a buy-now commitment: purchase today, and if a better Black Friday offer appears within 30 days, the company honors it.

Why sales stall

Cause is the harder question, and it is where AI Workers are working overtime. The Worker analyzes sales conversations, surveys, and CRM data to answer the "why" for customer churn. The questions that once took a month to answer now take an afternoon.

Operations and shipping

Purple points the AI Worker past the storefront. The team watches whether conversations handed from sales to support are trending up or down from one month to the next. It also reads why customers contact support about shipping and summarizes the reasons. Delivery problems surface in conversations well ahead of a dashboard.

Returns

The bottom of the funnel holds the costliest signal, the return. The team reads why customers send a mattress back inside the 100-night trial and summarizes what did not work for them. The weekly e-commerce review runs on the same worker, opening with a ranked breakdown of the week's sales conversation topics.

A single outbound campaign shows where this leads. Purple called recent buyers of its high-end mattress and used the worker to surface what else came up. Customers kept raising the mattress height. At 17.5 inches, paired with bases that delivery partners were setting to the highest position, the bed sat too tall for shorter customers. Purple corrected the install guidance and the showroom presentation.

Results

A single worker now covers demand, promotion response, conversion blockers, bundle interest, routing, shipping, and returns. All of it comes from one source, the conversations. Each read ends in a decision the team can act on.

The resourcing math follows. A deep dive that once occupied a team for most of a month now takes one analyst a few hours. Purple delivers the same depth of insight with fewer people than the work required before.

"In Level, I'm doing something similar now and can do it in a couple of hours. We have restructured our operation, I have less resources than I had in the past, and I can still deliver the insights with Level that we could before with less people." — Angie McDonald

What's next

Purple is moving toward AI coaching and custom workers built on its own evaluation questions. The team keeps sending Level AI feedback on what would help, on the strength of how fast that feedback has turned into shipped capability.

"I would absolutely recommend Level AI. It's been easy to adopt, and we've seen value come out of it really quickly." — Jane Greiner, Director of Operations of Customer Care, Purple Innovation