KrelunaCompany knowledge
Find and explain information from approved documents and procedures.
Kreluna ecosystem
Business AI connected to real operations
Kreluna Core introduces AI through a controlled, step-by-step approach: define the need, use authorised sources and produce results people can review.
Kreluna is in active development. Early access, feature availability and integration coverage are confirmed individually for each request.
Where business AI can help
Operational support
Prepare drafts, summaries and analysis while retaining human review.
Integrations
Connect gradually to existing tools where supported by the provider.
More than a blank chat window
Kreluna is designed to support research, planning, comparison, drafting and structured problem-solving. Straightforward questions can be handled directly, while demanding tasks can be broken into assumptions and options for review.
- Summarise and organise information
- Prepare drafts and plans
- Compare options without hiding uncertainty
Start with a focused use case
Define the business need, authorised sources, reviewers and a measurable outcome. Kreluna confirms feature and integration availability before any workflow is activated.
- Approved information sources
- Clear reviewers and responsibilities
- Human approval for important actions
Frequently asked questions
Does Kreluna make business decisions?
No. It can prepare analysis and content, while accountable people review important outputs and decide what happens next.
Which data can it use?
Only data and documents the organisation is permitted to use and that fall within the agreed scope.
Are integrations already available?
Compatibility with each product is checked before a connection or pilot is confirmed.
Built around control and clarity
Every project starts with a defined objective, authorised information and a clear review process. Capabilities are enabled progressively and important actions remain subject to approval.
- Defined scope and responsibilities
- Human review for important actions
- Clear limits and measurable outcomes
Practical applications
Where AI can support an SME
A useful project starts with real work and authorised information, not an empty chat box. These are workflows to assess, not claims that every feature is already available.
Company knowledge
Search approved procedures, manuals and documents while showing the internal source and respecting role-based access.
Sales preparation
Prepare account briefs and follow-up drafts from authorised CRM data while leaving price, terms and sending to the responsible person.
Administrative documents
Classify incoming files, extract fields for checking and flag missing information before anything reaches the official system of record.
Service operations
Route requests, suggest answers from approved content and escalate sensitive or unusual cases to the right operator.
Project method
From use case to an evidence-led pilot
The choice between rules, conventional automation and AI follows observation of the workflow. Not every problem needs generative AI and not every step should be automated.
- Define the outcome
Describe the problem, who performs the work today and which observable result should improve.
- Map data and exceptions
Identify sources, permissions, manual hand-offs, unusual cases and points where a person must decide.
- Test a narrow scope
Run the pilot on authorised sample data against acceptance criteria agreed before the test.
- Measure and decide
Compare the workflow with its baseline, then extend, revise or stop it on evidence rather than enthusiasm.
Data and accountability
Decide integrations and data before choosing a model
CRM, ERP, email and internal repositories can only be connected through authorised interfaces and suitable permissions. Feasibility is confirmed against the real system.
- Use only defined sources and purposes
- Separate reading, suggesting and acting on a system
- Apply least-privilege identity and access
- Require human review for material decisions, messages and changes
- Document providers, retention, subprocessors and the complete data path
Measuring value
Measure value without inventing a percentage
Value depends on the baseline. Measures are agreed before the pilot and interpreted alongside quality, risk and exceptions.
Cycle time
Time from incoming request to completion, including waiting and review.
Rework
Tasks returned because information, rules or quality requirements were missing.
Source coverage
Outputs that can be traced to current, authorised business information.
Exceptions
Cases that require escalation or a fully manual route, and why they occur.
Before you begin
Questions that define scope
Which use case should we start with?
Choose frequent, bounded and checkable work with accessible data and a named owner. Clarify an unstable or ownerless process before automating it.
Does AI replace staff?
The proposed goal is to remove repetitive preparation and improve hand-offs. Decisions, exceptions and accountability remain with defined people.
What determines cost and timescale?
Sources, integrations, risk and acceptance criteria. Kreluna defines scope before making a proposal and does not publish unsupported estimates here.
References for responsible design
The AI Act and GDPR can apply together: duties and safeguards depend on the organisation’s role, the data and the system’s actual use. Assessment therefore comes before configuration.
- European Artificial Intelligence ActOfficial text of Regulation (EU) 2024/1689.
- EDPB — Data protection by design and by defaultSME guidance on safeguards, minimisation and accountability.
- NIST AI Risk Management FrameworkVoluntary framework for governing, mapping, measuring and managing AI risk.
- OWASP — LLM Prompt Injection PreventionRisks and defences for AI applications connected to documents and tools.
These sources help frame the work; they do not replace legal, privacy or security advice for a specific situation.