The AI assistant (beta) is your smart assistant on your platform. The assistant helps you quickly and easily find the right information without having to search through all the documents and information pages yourself. Thanks to artificial intelligence (AI), the AI assistant can search all the content within your platform and provide relevant answers to your questions. The assistant can also help you write or translate content.
The AI assistant is not enabled by default.
Would you like to activate the AI assistant? Email support.plek@bcs-hr.com.
The AI assistant uses Azure AI to provide smart answers to questions within our platform. This is done through various AI services:
- Azure OpenAI: Generates answers based on the available information.
- Azure AI Services: Ensures secure storage and optimisation, so that answers are as accurate as possible. OpenAI is part of this.
If no relevant information is available within the platform, the AI assistant bases its answers on general knowledge from public datasets.
What can the AI assistant do?
The AI assistant currently has two functionalities:
-
Search for information
Quickly and efficiently find answers to your questions without manually browsing through all the information pages and documents. -
Help with creating messages
Whether you need inspiration, translations or improvements, the AI assistant supports you in writing engaging, high-quality content.
Searching with the AI assistant
1️⃣ Ask a question
Type your question in the chat with the AI assistant.
Example: “How do I request leave?”
Or use one of the preset prompts.
2️⃣ Search the platform
The AI assistant scans all information pages and documents to find relevant information.
3️⃣ Formulate an answer
The assistant provides a clear and concise answer based on the information found.
4️⃣ Sources
You receive an overview of the sources used, so that you can click through to the original documents.
Using the AI assistant for messages
When creating a message, you can open the AI assistant using the Assistant button below the text editor. The assistant appears in the top right of your screen.
Creating a message: start with a prompt
Not sure how to get started? Use a prompt to create a draft.
Example:
"Create a message announcing that two new employees, Ella and Nathan, will start as Sales interns from March 2025."
Happy with the result?
- Click Copy to copy the text to your clipboard.
- Click Play to place the text directly in the post editor.
Improving a message
Do you already have a draft but want to refine it?
- Write your draft in the text editor.
- Click the AI assistant button and choose a prompt to improve your text.
- The assistant rewrites, simplifies or optimises your text.
ℹ️ Please note: The assistant always works with the content of the text editor. Make sure the latest version of your text is in it.
What information does the AI assistant use?
✅ Platform-wide documents and information
✅ Documents and information pages from groups you are a member of
If no relevant information is found, the assistant uses general knowledge from public datasets (up to a certain point in time).
Is my data safe?
Yes. The AI assistant was developed with Azure AI, which complies with strict security and privacy standards:
- Your data remains yours: Microsoft does not use your data to train AI models.
- Strong encryption: Data is always encrypted during transmission and storage.
- Secure access: Only authorised people have access through Azure Active Directory.
- Complies with GDPR and ISO certifications: Including ISO 27001 and SOC 2.
- Full transparency: Plek can always check how data is processed and stored.
Requirements for the AI assistant
- Up-to-date content: All information must be up to date.
- Sufficient information pages: A well-stocked knowledge base ensures better answers.
- Logical structure: Clear main and subpages help the assistant understand connections.
Limitations of the AI assistant
❌ Cannot search images (text only).
❌ No access to external documents (such as SharePoint or Google Drive).
❌ Answers are based on the available information and the question asked; no rights can be derived from them.
Tips for an effective PROMPT
...
1. Be specific and clear
The more specific your question, the better the assistant can respond. Avoid vague or overly broad questions.
Poor example: "Tell me something about sickness absence."
Good example: "How should I report myself sick according to the policy in my department?"
2. Use context
Give the assistant sufficient context so that the answer better matches your question.
Poor example: "How does it work?"
Good example: "How does the process for requesting leave work in our HR system?"
3. Ask one question at a time
Assistants can have difficulty processing multiple questions in a single prompt. Therefore, focus on one question at a time.
Poor example: "How do I request leave and how do I change my working hours?"
Good example: "How do I request leave in our system?"
4. Keep your question short but informative
Long, complex sentences can cause confusion. Formulate your question clearly and concisely.
Poor example: "I was wondering if you could perhaps explain how the new safety procedures introduced last month work?"
Good example: "Can you explain the new safety procedures introduced last month?"
5. Ask for specific formats or styles
If you want an answer in a particular format or style, make this clear.
Poor example: "What are the benefits of working from home?"
Good example: "Can you list the benefits of working from home in a list of five points?"
6. Avoid ambiguity
If a question can have multiple interpretations, provide clarification.
Poor example: "What are the rules?"
Good example: "What are the rules for requesting leave in my department?"
7. Use examples
If you expect a specific type of answer, you can provide examples to help the assistant understand what you mean.
Poor example: "What are the guidelines for working from home?"
Good example: "Can you list the guidelines for working from home, such as working hours and availability requirements?"
8. State a goal
If you have a specific goal in mind, state it to get more targeted answers.
Poor example: "How can I improve my presentations?"
Good example: "How can I improve my presentations to engage my audience more?"
With these tips, you can ensure that the AI assistant understands your questions better and can provide more useful and targeted answers!
FAQ
Does my data remain within my environment?
In short: Your questions do not, but your data does. Your data never leaves the application and is not used to train AI. The AI assistant is hosted via Azure Bot Services. When a user asks the AI assistant a question (a 'prompt', in technical terms), it is sent to these services. Azure uses this data temporarily to generate answers. We have configured our Azure services with Microsoft-Managed Keys for encryption and they do NOT retain data for any other purposes than processing our requests (according to their privacy policy). The source material – information pages and documents – is stored securely in our Elasticsearch instance, which means that the content remains private and protected.
Is my data visible on other platforms?
No, your data remains on your environment. An index is run for each platform, where the answer to your question is searched for in your content using the AI Search Index. When using AI Search Reindex in our system, OpenAI (via Azure AI Services) is used to create vector embeddings for the document content. These vectors are then stored in our own Elasticsearch instance. When a user's search query is processed, OpenAI is used to convert the search query into a vector embedding using the same embedding model. This embedding is then used to search for the most relevant content in Elasticsearch.
Vector storage is specialised data storage that allows information to be indexed and retrieved based on vector representations. These vectors, called embeddings, capture the semantic meaning of the data that is embedded. Vector storage is often used to search unstructured data, such as text, images and audio, to retrieve relevant information based on semantic similarity rather than exact keyword matches. Elasticsearch is vector storage.
How does the AI assistant use our data to answer questions?
This is how it works:
- Data Indexing: When using AI Search Reindex in our system, OpenAI (via Azure AI Services) creates special data representations called vector embeddings for your content.
- Data Storage: These vectors are stored securely in our Elasticsearch instance.
- Query Processing: When you ask a question, OpenAI converts it into a vector embedding. This embedding is then used to find the most relevant content in Elasticsearch.