Transparency
AI & Service Limitations
Trust requires clarity about what technology can and cannot do. Raeburn Consulting avoids unsupported absolute claims such as “100% accurate”, “completely secure”, “fully autonomous” and “guaranteed compliant”.
AI outputs require validation
AI-generated or AI-assisted outputs may be incomplete, inaccurate, ambiguous, stale or misleading. Human validation is required proportionately to consequence, particularly for client-facing, legal, financial, security, privacy, employment and other high-impact use cases.
No 100% accuracy claim
We do not promise that AI, automation, discovery or analytical outputs are 100% accurate. Results depend on available data, access, configuration, third-party services, model behaviour and the quality and completeness of inputs.
No absolute-security claim
No responsible technology provider can guarantee absolute security. We apply risk-based security controls, secure-development practices, vulnerability management and incident response while recognising residual risk remains.
Autonomy has boundaries
Automation may execute defined tasks without manual intervention where appropriate, but not every service is fully autonomous. Human approvals, exception handling and escalation are retained where risk, client requirements or service design require them.
Client approval matters
Raeburn does not assume authority to make material business decisions on a client's behalf unless expressly agreed. Client approval may be required before production changes, consequential communications, material workflow actions, data migration or other defined activities.
Compliance is contextual
Technology and consulting services can support compliance processes but do not guarantee compliance with every law, regulation, standard or customer obligation. Compliance depends on scope, jurisdiction, implementation and the client's own controls and responsibilities.
Discovery can have constraints
Systems may be constrained by permissions, unavailable APIs, legacy technology, unusual networks/VPNs, proprietary protocols, hardware interfaces, incomplete documentation or client restrictions. Where direct inspection is not possible, limitations are recorded rather than silently inferred away.
Third-party dependencies remain
Some services depend on cloud, SaaS, infrastructure, communications, data or AI providers. Availability, model behaviour, processing location, retention and functionality can be affected by provider configuration, terms and service changes.
Our preferred language is evidence-based: designed to, tested against, supports, monitored, risk-based, proportionate and subject to validation. Material limitations, incidents, false positives/negatives and provider/model changes feed into continuous improvement.