Haidi – the clever BIPA bot that knows every shelf secret

by | Sep 30, 2025 | Customer Success, Employee Satisfaction, Uncategorized | 0 comments

By Editorial

When the branch becomes a jungle of knowledge – that’s when Haidi comes in.

Imagine: You are a BIPA employee standing in the BIPA store, a customer looks at you questioningly, the cashier is ill, your colleague is on her lunch break. You’d like to refresh your knowledge of how exchanges work or what the difference is between eau de parfum and eau de toilette – but where can you look it up?

These are the moments when you try to remember: ‘What was that exchange policy again? Do we need the original packaging or just a smile?’ The thick manual? Does anyone have time for me? That’s when Haidi comes in and saves you.

1. Who is Haidi – and why does she speak Turkish?

Take BIPA, add a pinch of Google technology, and spice up the partnership with a generous helping of voice intelligence. The challenge: ‘We need a smart bot for our branch employees.’

The result is Haidi, the AI assistant who, since August 2025, has been making internal knowledge readily available in all branches throughout Austria.

And the great thing is: Haidi not only speaks German, but also English, Turkish and Hungarian – so she can also translate internally.
This means that colleagues whose native language is not German don’t have to laboriously translate technical terms and guess their meaning, but can find their way around more quickly.

So when Haidi soon starts waiting for conversations in every BIPA branch, it could change the way employees learn, advise and feel even more confident when dealing with customers forever.

2. How it works – not a toy, but smart technology

Haidi is located directly on the BIPA intranet and has access to structured knowledge. Employees type in questions in everyday language – for example: ‘How do I make a complaint?’

And instead of providing endless lists of results, Haidi responds in complete sentences, suggests follow-up questions and links to relevant documents.

The technology behind it?

Architecture & Technology Stack

Basis: ‘Haidi’ is a RAG assistant based on Google Vertex AI Search.
Knowledge source: Haidi works with a curated knowledge base (e.g. process manuals, cash register guides, technical instructions).
Indexing: Content is provided in cloud storage, Vertex AI Search creates indexes and embeddings for fast semantic search.
Front end & runtime: Containerised application on Cloud Run, secured via Keycloak. Access only from branch networks.

Performance & Security

Zero data retention: User entries are not stored.
EU hosting: All services run in European regions.
Access control: Authentication via Keycloak, IP range restrictions.

Multilingualism & Voice-Output

Haidi understands and responds in multiple languages. Translations are performed dynamically to support diversity within teams.

Voice Intelligence Haidi currently only works with text. Voice control (e.g. via Google Cloud TTS) is already on the roadmap.

3. The internal revolution – with heart and byte

What BIPA has planned for Haidi is more than just a nice gimmick. New colleagues should be integrated into the branch team more quickly. Internal processes, rules and procedures that used to be tedious to look up can now be accessed via a question-and-answer system.

In a matter of seconds, not minutes or hours. Haidi could thus become a talking company encyclopaedia – precisely the revolution that voice intelligence now makes possible.

In other words: Wiki is out – Haidi is in!
Google is already saying that this type of voice intelligence should serve as a model for other companies – a kind of knowledge revolution in retail.

4. A few humorous thoughts (with instructions)

Haidi, tell me the truth: when you say that the new perfume is “super seductive”, does that mean “super good” or “so strong that you have to run away”?

Language mix-up: When someone asks a question in Hungarian and Haidi answers in German – is that a translation error or a secret test?

Follow-up question trap: ‘How do I make a complaint?’ → Haidi: ‘What kind of complaint do you mean?’ → You: ‘Electronics.’ → Haidi: ‘Model?’ … You could go on like this for hours.

Drama surrounding exchanges: Previously: you had to search through the manual, ask managers, comb through Excel files. Now: just ask Haidi! Simple, elegant, no more handwritten notes.

(C) 2025 REWE International AG

5. A step into the future – with a smile

Haidi ist kein bloßer Bot, sondern der revolutionäre Versuch, das interne Wissen von BIPA “sprechbar” zu machen – so fragt niemand mehr ins Leere. Wenn Haidi gut funktioniert, könnte sie MitarbeiterInnen das Leben um einiges leichter machen – und der Konkurrenz so manches Kopfweh bereiten: „Wie hat BIPA das bloß gemacht?“

If you think about voice intelligence the way BIPA does, further expansion stages are probably already in the pipeline. Why not roll out knowledge transfer across the entire group, why not standardise the onboarding and offboarding (!) of employees via VI? In any case, it will be interesting to see what this revolution in knowledge transfer has set in motion.

6. Making knowledge interesting again

One overlooked advantage: Voice Intelligence makes knowledge human again. Instead of reading long manuals, employees can interact with a voice assistant:

‘Tell me how we solved the problem with delivery delays last year.’

‘Explain to me the three most important insights from the last product launch.’

‘What are the most common customer complaints about product X?’

This conversational interface dramatically lowers the barrier to entry — no training required, no fear of “using the wrong search term.” People simply ask questions naturally and get meaningful answers.

(C) 2025 REWE International AG

7. Why such a bot cannot simply be created overnight

As smart as Haidi seems, getting there was no walk down the shampoo aisle.

There are some real challenges behind the project, which Overmind.one has highlighted in its 4C Framework:

• Context – Taming knowledge chaos: Previously, countless old manuals, PDFs, FAQ documents and PowerPoint slides had to be collected, updated and structured. It was a bit like spring cleaning the company archive – only with thousands of pages.

• Code – Mastering technical implementation: German, English, Turkish, Hungarian – but this requires a lot of training and fine-tuning to ensure that the bot does not suddenly say ‘Yes’ when it actually means ‘Maybe, but only with branch management’.

• Culture: Creating acceptance within the team: New tools are great – when they are used. So training, testing and a lot of persuasion were needed to ensure that employees really saw Haidi as a helper and not as digital competition.

• Compliance: Ensuring data protection and security: When internal processes and sensitive information end up in the bot, everything must be GDPR-compliant and still remain quickly accessible. No employee wants to suddenly have to read the 2018 data protection notice during the exchange process.

In short: behind the friendly chat window lies a huge organisational and technological effort that shows the direction we are heading in from now on. But if the end result is a new culture that can better support employees in their day-to-day business, then the effort will have been worth it.

(C) 2025 REWE International AG

Milan Veskovic CEO at Overmind

Jump Start free of charge

First, we invest in you by offering you our Project Jump Start (PJS). We analyse your readiness for AI and suggest the first concrete steps. Get in touch with us now!

Visit our e-book and learn about the challenges and opportunities of AI. Just click on the image.

Next generation AI for corporate challenges

Stopping the Knowledge Drain: How Voice Intelligence Turns Lost Know-How into a Competitive Advantage

Every company has a memory problem — and most don’t even realise how big it is. Over the years, teams hire and offboard employees, roll out new tools, merge departments, and shift strategies. Every change means one thing: knowledge — the practical, experience-based,...

read more

0 Comments

Submit a Comment

Your email address will not be published. Required fields are marked *