Anyone in marketing knows the pattern. Surveys, comments, reviews and emails arrive through many channels. What fits into the week gets analysed. The rest waits.
What the numbers show
According to Eurostat, EU companies that use AI apply it most often in marketing and sales. The most widespread AI technology is text analysis, used by 11.8 % of enterprisesA. That matches the research: the Danish study by Anders Humlum and Emilie Vestergaard lists marketing professionals among the occupations with the largest time savings from chatbots.
The well-known study by Harvard Business School and Boston Consulting Group shows the other half of the picture. Consultants using GPT-4 on suitable tasks worked 25.1 % fasterS, completed 12.2 % more tasksS and delivered more than 40 % higher qualityS. Their output also became more uniform. For marketing, that is a warning sign: if the same model writes every text, you sound like everyone else.
A real case
The Portland Trail Blazers, an NBA team, analyse guest feedback with a Zapier workflow. According to Zapier, review time fell from 50 to 3 hours a weekH. Guests with a problem get a reply within 24 hours. Again, this is a company claim. The workflow itself transfers to almost any business.
The workflow
- 01DataFeedback from surveys, email, reviews, social media
- 02AIAI tags topic, sentiment and urgency
- 03CheckCritical case?yes: draft reply goes to service, answer within 24 hours
- 04AIWeekly report: topics, trends, examples
- 05PersonTeam decides on actions
- 06ResultActions into the project tool, check impact in the next report
Limits and risks
- Uniform copy: AI drafts need a clearly described brand voice in the prompt and an editorial pass.
- Personal data: Free-text feedback often contains names and contact details. Pseudonymise before analysis.
- Labelling: Art. 50 of the AI Act sets transparency duties for certain AI-generated content. Check with legal what applies to your channels.
- Copyright: Check generated images and texts for copied material before publishing.
How to start
- Pull all feedback sources into one spreadsheet or form export.
- Let AI form at most eight topics and assign every response.
- Check the assignment by hand on 30 examples before trusting the result.
- Set up a fixed weekly report that leads to real decisions.
Sources
- AEurostat: 20% of EU enterprises use AI technologies, 11 Dec 2025
- SDell'Acqua et al.: Navigating the Jagged Technological Frontier, HBS Working Paper 24-013, 2023 (report by Legal Dive)
- SHumlum, Vestergaard: Large Language Models, Small Labor Market Effects, NBER WP 33777, 2025
- HZapier: How the Portland Trail Blazers save 47+ hours a week with AI, 16 Sep 2025