How we use AI
We, Nextima, s. r. o., use AI in AIKOZO to help teams with software requirements, specifications, planning, design and delivery. When you use an AI Assistant or an AI generation or analysis feature, you are interacting with an AI system. Its responses are generated automatically and can be wrong, incomplete or unsuitable for your situation.
Depending on your enabled features and permissions, AI can help you interpret source material, draft and translate content, search for related information, identify gaps, suggest plans and designs, and prepare delivery artifacts. Our AI assistance supports your professional judgement and your organisation's review and approval processes.
What an AI request can include
We may assemble task context from your prompt, selected records, document extracts, related workspace items, retrieved information, templates and conversation history. The input can therefore include more than the text you type. Some workflows use translations or an English working version of your content; check that they preserve the intended meaning.
When a feature uses an external model, we send the relevant input to the configured provider. Generation, translation and embeddings can involve different services. Our Privacy Policy explains providers, processing locations, transfers, retention and model-training practices. EU application hosting alone does not mean all AI processing stays in the EU, and embeddings are not automatically anonymous.
Only provide information your organisation is authorised to process in that configuration. Avoid passwords, access tokens and unnecessary personal or sensitive information. Integrations and external agents may introduce additional recipients and actions within their granted permissions.
AI identification and records
Our AI Assistants provide a reminder beside the input: “AI can make mistakes. Review results before use.” The nearby About AI control explains processing and links to this notice and our Privacy Policy. You can read those details before submitting a request.
Where available, saved generation and conversation records provide source references, generated content, model information, timestamps or history. Some Assistant responses carry an AI-generated response label when supporting metadata is available. These details are not included in every view, copy or export; an unlabelled result is not proof that a person wrote it.
A source reference, generation record or human approval helps with review but does not prove that an output is correct or compliant.
Review and execution
Check generated or modified content against the original evidence, business intent and applicable requirements before relying on it. For code and delivery instructions, also check behaviour, security, dependencies, licensing and the scope of changes. Use the tests and approvals appropriate to the target environment.
Creating an Action Pack does not execute it. Where enabled and authorised, approved Action Items can initiate connected build or issue-creation workflows. An authorised run can perform several steps without asking again at every step. Check the target, scope and permissions before authorising execution. Other integrations or external agents may also change information within their permissions; do not assume every Assistant interaction is read-only.
Your review responsibilities do not remove our obligations under applicable law or your agreement with us.
Limits to keep in mind
- Errors and omissions: AI can invent facts, references or relationships, miss relevant evidence and make unsupported assumptions. A citation may not support the conclusion.
- Variation and bias: Similar requests can produce different results. Language, source material, model and configuration changes can affect answers and introduce bias.
- Scores and estimates: Similarity scores, confidence indicators and effort estimates are aids, not independently validated probabilities or commitments unless specifically documented.
- Security and rights: Source material can contain malicious instructions. Generated text or code can introduce vulnerabilities, unsuitable dependencies or third-party rights issues. Review suggested actions and changes before use.
We provide AIKOZO for software specification and delivery work. We do not present it as a system for deciding people's employment, creditworthiness, access to education or public services, or other matters with legal or similarly significant effects. Do not use an AI recommendation as the sole basis for such a decision. A different intended use can require its own assessment and safeguards. Generating requirements or code does not establish that the resulting system is compliant or safe to deploy.
Questions and concerns
Report errors, unexpected disclosure, harmful output or unintended actions through your organisation's support channel or to nextima.sro@gmail.com. Include the feature, approximate time, relevant artifact or run reference and a brief description. Do not send secrets or unnecessary confidential information in the initial report; we can arrange how to obtain further details.
We update this notice when material AI capabilities or processing arrangements change. Our Legal Notice identifies us; our Privacy Policy explains personal-data processing and your rights.