Good AI starts with a specific task.
Find knowledge. Understand documents. Simplify processes. Discover how meloki can help with your business tasks – with clear data flows and under your control.
From your needs to the result
What would you like to make easier?
Choose a scenario and follow your data’s journey. These examples show possible workflows, not live processing of company data.
How do I request a new software application?
- InputQuestion & knowledgeSoftware request + IT policy
- ProcessingFind text passagesPassages relevant to your question
- ProcessingDraft the answerLocal AI uses the passages
- AnswerYour answerProcess from the IT policy
With your configured knowledge base.
Explore this scenarioWhat are the most important points in this report?
- InputPDF + taskCondense the report to five points
- ProcessingCondense informationLocal AI summarises
- AnswerFive key pointsDraft summary
Check the draft against the original.
Explore this scenarioCan this request become a draft ticket?
- InputInternal request“Printer not responding”
- ProcessingDraft ticketAI: category + description
- HumanYour approvalCheck and approve the draft
- AnswerCreate ticketThrough the configured API
Example with a configured n8n workflow. The ticket is created only after your approval.
Explore this scenarioWhich tasks did we agree on in the meeting?
- InputAudio fileApproved conversation recording
- ProcessingTranscriptWhisper converts speech to text
- ProcessingAI analysisExtract key points and tasks
- AnswerDraft tasksCheck against the conversation
AI analysis adds to transcription as a separately configured workflow.
Explore this scenarioCan my AI support a defined workflow?
- InputYour taskPrepare the case information
- ProcessingPermitted toolsAI uses approved sources
- AnswerDraft resultInformation + open points
- HumanYour reviewFollow-up action only after approval
Access and approvals are configured for the workflow.
Explore this scenarioWhich information is relevant to this case?
- InputData sourcesERP or CRM, for example
- ProcessingInterfaceAPI reads permitted fields
- ProcessingLocal analysisBring together data for your question
- AnswerYour overviewInformation + visible gaps
With individually configured interfaces.
Explore this scenarioWhich local model suits my application?
- InputExample taskSummarise this report
- ProcessingTest modelsSame task, different models
- HumanCompare resultsQuality, runtime, memory
- AnswerChoose modelSuited to your task and hardware
Choose based on task, licence and hardware.
Explore this scenarioFind knowledge
Your knowledge. Ready for questions.
How do we handle this here? Your team asks questions and finds answers in your company’s approved manuals and policies.
Existing knowledge becomes easier to access at work.
What it can look like
- Your starting point
Your knowledge
Approved manuals and policies
How do I request new software?
- What meloki does
Find relevant passages
meloki searches the configured knowledge base.
- Your result · Example
An answer to your question
Use the internal IT form. Describe the purpose and the access you need. IT will review your request.
Clearly written, based on your sources
Setup & technical details
Usable with a configured knowledge base
- Suitable inputs
- Approved manuals, process descriptions, policies and internal documents.
- Technical building blocks
- Document preparation
- RAG / semantic search
- Local language model
- Requirements
- Current, machine-readable and approved documents
- A configured knowledge collection and a suitable model
- Suitable access permissions in the application used
- Integration
- File stores and business systems can be connected through suitable interfaces. Synchronisation and permissions must be configured for the specific source.
- Result and context
- A clear answer based on the information provided. Source references depend on the chosen application and configuration.
RAG makes knowledge searchable; it does not automatically train the model. Check sources and results for important decisions.
Understand documents
Read less. Understand more.
A long report before your next meeting? Get a clear draft of the key points and open questions.
You can identify more quickly which sections need a closer look.
What it can look like
- Your starting point
Your document
A readable report or documentation
What are the key points in this project report?
- What meloki does
Condense information
Local AI works on your specific question.
- Your result · Example
The key points
Implementation is in progress. The interface review and team approval are still open. Next step: review the results together.
A short summary you check against the original
Setup & technical details
Usable with suitable document processing
- Suitable inputs
- PDFs, technical documentation, manuals, policies, reports or, as supporting material, contracts.
- Technical building blocks
- Document processing
- Local text analysis
- Summarisation and structured extraction
- Requirements
- A supported, readable document format
- Document processing suited to the source material
- Enough context and suitable hardware for the workload
- Integration
- Automatically importing content from file stores or business applications requires a configured integration. Scanned documents may need additional text recognition.
- Result and context
- A summary or structured list of relevant information tailored to your question.
AI can miss or misrepresent content. Legally relevant documents and decisions based on them need human expert review.
Connect processes
Turn a request into a useful next step.
Read a request, classify it and record it as a ticket: connect these steps in a configured workflow – with your approval before the action.
Less manual copying. The decision stays with your team.
What it can look like
- Your starting point
An internal request
A report about a faulty printer, for example
The printer in the meeting room is not responding.
- What meloki does
Prepare ticket
AI and the workflow suggest a category and description.
- Your result · Example
You approve the action
Draft ticket: printer not responding · Category: IT issue. Review the draft and approve ticket creation.
Only then does the configured interface create the ticket.
Setup & technical details
Workflow requiring individual setup
- Suitable inputs
- An approved event, such as an internal request or a webhook with defined data.
- Technical building blocks
- n8n workflow
- Local classification and text generation
- API connection and approval step
- Requirements
- Configured n8n service and defined workflow
- An accessible, supported interface and suitable permissions
- Defined error handling, approvals and logging
- Integration
- This example requires a custom workflow; it does not promise a native ticketing integration. External APIs may process data outside your infrastructure.
- Result and context
- A prepared draft ticket. It is created in the target system only through a workflow explicitly configured and approved for that purpose.
Decide in advance which steps may run automatically. Limited permissions and deliberate approvals are useful for actions that write data.
Work with audio
From conversation to next tasks.
What did we discuss? Convert a permitted conversation recording into text and use a suitable workflow to prepare the next tasks.
Conversations become readable and can be worked on together.
What it can look like
- Your starting point
Your audio recording
A permitted, approved conversation recording
Which tasks did we agree on in the conversation?
- What meloki does
Speech becomes text
Whisper transcribes. A language model can structure the text.
- Your result · Example
Transcript and draft tasks
Draft tasks: review the proposal, collect open questions and agree the next meeting. Please check against the conversation.
Check names, statements and tasks against the conversation.
Setup & technical details
Local transcription with a configured service
- Suitable inputs
- A lawfully recorded, approved audio file in a supported format.
- Technical building blocks
- Local Whisper transcription
- Text structuring
- Summarisation with a local language model
- Requirements
- Permitted recording and suitable audio quality
- Configured transcription service
- Language model and, if needed, a workflow for summarisation
- Integration
- The workflow starts with an audio file you provide. A direct connection to video conferencing systems needs to be evaluated and configured separately.
- Result and context
- A transcript and, with a suitable workflow, a draft summary and task list.
Check names, numbers, technical terms and tasks against the conversation. Speaker recognition and automatic meeting participation are not assumed here.
Extend your applications
Your task. Your tools. Your boundaries.
Does your task need custom tools? Add a tailored application that retrieves selected information and supports defined work steps.
You decide which tools the application may use.
What it can look like
- Your starting point
Your task
A specific goal and explicitly permitted tools
Prepare the information about this case for me.
- What meloki does
Your application follows
The model uses only the access configured for it.
- Your result · Example
A result for your review
The application collects approved case data and highlights open points. Follow-up actions that write data require your approval.
Follow-up actions that write data need your defined approval.
Setup & technical details
Custom integration and development
- Suitable inputs
- A defined task with permitted data sources, tools and decision steps.
- Technical building blocks
- Local model
- Custom application or workflow
- Restricted tool and API calls
- Requirements
- Design and build a suitable application
- Model support for the required tool calls
- Limited permissions, approvals, tests and error handling
- Integration
- Agents and custom applications are integration scenarios. This does not promise a universal autonomous agent feature as part of the standard package.
- Result and context
- An application tailored to your process, with its scope defined by the model, tools and configuration.
Limit tool permissions and check input from external sources. Critical actions or actions that write data need defined human approvals.
Analyse business data
See the connections. Across your systems.
The information is in your ERP, the context in your CRM? Bring permitted data extracts together through configured interfaces to answer your question.
Related information comes together in one overview.
What it can look like
- Your starting point
Your business data
Approved fields from ERP, CRM or a database
Which details are still missing from this order?
- What meloki does
Bring information together
Interfaces provide the data for local analysis.
- Your result · Example
Your overview
The data extract provided has no delivery date. The payment status is available. Check the current state in the source system.
A draft result with relevant information and open points
Setup & technical details
With individually configured interfaces
- Suitable inputs
- Defined data extracts from ERP, CRM, a database or an accessible REST API.
- Technical building blocks
- Interfaces and data preparation
- Local text and data analysis
- Structured results
- Requirements
- Suitable, available interfaces
- Data mapping, suitable permissions and a defined refresh cycle
- Expert review of data quality and results
- Integration
- ERP, CRM and database connections are configured for the specific system. Universal native connections and real-time data are not assumed.
- Result and context
- An analysis or explanatory draft limited to the connected data.
The analysis is only as current as the data provided. Reliable metrics and calculations should be checked using suitable business logic.
Choose your models
The right AI for each task.
Short texts, long documents or custom applications: choose a suitable model for your task and adapt your selection later.
Your model choice stays open to new tasks and requirements.
What it can look like
- Your starting point
Your example task
A typical task from your working day
Which model suits our summaries?
- What meloki does
Compare models
You assess quality, runtime and the resources needed.
- Your result · Example
Your suitable selection
Try the same task with Gemma 4 and Devstral, for example. Compare results and runtime; you can adjust your selection later.
A suitable starting point you can change later
Setup & technical details
Model selection depends on compatibility and hardware
- Suitable inputs
- A specific task with the required language, context size, quality standards and permitted data flows.
- Technical building blocks
- Local model serving with Ollama
- Task-specific model selection
- Modular AI services
- Requirements
- Compatible model format and suitable model service
- Sufficient GPU memory, RAM and storage
- Reviewed licence and tests using your own subject-specific examples
- Integration
- Additional services and custom applications can be integrated after technical review. Compatibility with any model or container is not guaranteed.
- Result and context
- A model and service configuration suited to your task and hardware.
A larger model variant does not automatically guarantee higher quality. Check resources, licences and suitability for your actual use case.
Ready to meet your own AI?
Discover in a personal demo how meloki can support your company knowledge, applications and workflows.
Arrange a demo