      
      case studies
      /
      tech companies
    

google ads, localized to 27 languages, written in minutes.

how a european e-commerce retailer turned a nine-hour weekly job - rewriting its google ads for 27 countries, each in its own language - into a ~10-minute job an ai agent now does, with zero mistakes since launch.

sectore-commerce

scopegoogle ads localization

footprint27 countries

what we didmap · build · embed

        google ads · 27 countries
        fig. 01
      

       the outcome
    

~10 min to do the weekly job, down from ~9 hours

468 hrs of an expert's time given back each year

0% mistakes across 6 weekly runs and 27 countries

~€9.8K saved a year

 the challenge

nine hours a week, lost to copy-paste.

every week, the retailer's google ads needed fresh text for 27 countries - the extra links, the short selling lines, and the sale labels. one person copied and pasted it all across 27 spreadsheets. it took about nine hours.

they also had to translate every offer, count the letters by hand to fit google's limits, and match each ad to the right account.

mistakes - text too long, or the wrong account - only showed up at the very end, after half a day of work was already gone.

 what we did

a system, not a dashboard.

map

first we watched the whole job, step by step: the weekly plan with one tab per country, pulling the offers, writing the copy, checking the length, matching accounts, and getting it ready for google. then we tested it in three countries - czechia, slovakia, and the uk.

watch the job
3-country test
find the errors

build

one ai agent does five things in a row, once a week: it reads the weekly plan, pulls each country's offers, writes the ads in that country's language, checks the length and the account, then hands a ready-to-load file to the expert to review.

one weekly agent
writes in-language
ready to load

embed

we grew it to all 27 countries in three weeks. every time the expert fixes something, the agent learns from it - so it gets a little better each week.

27 countries
learns from fixes
better each week

 before → after · time per week

before

~9 hrs

all by hand, every week: 27 spreadsheets, translating and counting letters - with mistakes found only at the very end.

after

~10 min

the agent writes, checks, and gets everything ready. the expert just looks it over - and every fix makes it smarter.

 in their words

end-to-end, it just works.

- performance marketing lead

 what changed

zero mistakes, every week.

the agent has been running since launch. across six weekly runs and all 27 countries, it hasn't made a single mistake that needed fixing.

a nine-hour job now takes about ten minutes. that's 468 hours of an expert's time back every year, and about €9.8k saved - and the two things that used to go wrong most, the length limits and the accounts, are now handled for them.