Shiseido uses Gemini to take busywork off its beauty consultants
At a Google Cloud Japan event, Shiseido Japan described AI tools that handle study, scheduling and social posts so staff can focus on hospitality.
The short version
- Training slides become podcast-style audio that staff can play while commuting or setting up the store.
- AI predicts demand by store and day, then assigns consultants to stores.
- Three Gemini-based agents help consultants draft, check and review social media posts.
- Nine in ten staff who tried the audio learning called it very useful in a survey.
- Next steps: an after-hours AI advisor and a live-stream comment helper, with humans still leading service.
Why it matters
Many customer-facing jobs are crowded with preparation and compliance work that pulls people away from customers. Shiseido's approach is to name the human part worth protecting, then automate the tasks around it. The figures come from company speakers on a stage, not an outside audit.
The problem: too much besides welcome
Shiseido's beauty consultants, now called Personal Beauty Partners (PBPs), have built their work around omotenashi, a Japanese word for wholehearted hospitality, and empathy for the customer. But the product range keeps growing and customer tastes keep diversifying. Staff must learn a lot, and keep re-checking details.
Some PBPs also post on Instagram and YouTube. Because posts stay online, they need careful wording and a check against Japan's pharmaceutical affairs rules (薬機法), which takes time. As itmedia.co.jp reported, two Shiseido Japan digital staff described the response at a Google Cloud Japan event.
Study by ear, shifts by algorithm
For in-store staff, the company turns internal PowerPoint decks into podcast-style audio. Staff in the field do this themselves, with no engineers. They load slides into Gemini Notebook (formerly NotebookLM), draft a dialogue script, fix details in Google AI Studio, then review and publish. AI Studio is used because small edits, such as one phrase or one pronunciation, are hard to make in the first tool.
A second tool plans staff placement. Consultants used to work at one fixed store, which left gaps when demand jumped. The system forecasts demand by store and day from past sales and campaign information, works out the headcount needed and assigns people. It respects days off, training and a limit on consecutive working days.
Shiseido says nine in ten PBPs who used the audio learning rated it very useful. It expects the shift tool to save about 5.3万 hours of work within the year.
Three agents for social posts
For consultants who post online, Shiseido built three agents, meaning AI helpers that carry out a task step by step, on Gemini Enterprise. One drafts posts from short notes and copies each person's habits, such as emoji and sentence endings, because the team did not want AI to flatten individual style.
The second compares a draft with internal rules and past compliance notes and flags what needs changing. The third connects to social media results, sorts out which posts did well and discusses why with the consultant. The agents draw on five internal data sources.
Where the line is drawn
Next, Shiseido plans an "AI PBP" that answers customers outside store hours and an "AI Liver" that helps respond to comments during live streams. If a customer talks to the AI first, the context would pass to a human consultant, so nobody has to explain things again.
The company says it will not let AI complete service on its own. It is looking for value in things that are hard to count, such as how moved a customer feels, rather than only shorter service times.
What to watch
- Will Shiseido publish results on time spent with customers, not just time saved?
- How will the after-hours AI advisor hand customers over to human consultants in practice?
- What does the rollout cost, and how will it be judged beyond the survey and the forecast?
Try it
Write down the one human thing your team must protect, then list the surrounding tasks that eat its time. Pick one, such as turning a training document into audio, to test first.
Sources and links
First reported by itmedia.co.jp. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.