How Quicksilver uses AI to strengthen analysis, judgement and delivery, without giving up professional responsibility.
Quicksilver Business Innovation Group and Quicksilver Software Development (together, “Quicksilver”) use AI to strengthen analysis, judgement and delivery, not to replace professional responsibility or diminish a client's ability to understand and govern the work.
These standards apply to AI we use, recommend, develop or integrate in advisory services, research, data work, software engineering, automation and internal operations. They also guide work performed on our behalf.
We start with the problem and the intended outcome, not the technology. AI is appropriate only when it offers a defensible benefit relative to its risks and alternatives. We favour approaches that clients can understand, govern and sustain.
The safeguards required for a particular engagement depend on the nature and consequences of the work and are agreed with the client where relevant.
AI-generated material is not evidence simply because it is plausible or persuasive. We verify material claims, distinguish facts from inference and uncertainty, and evaluate outputs against their intended purpose and context. Our recommendations remain the product of professional judgement, informed by evidence and open to challenge.
We consider whether AI use could introduce unfair bias, exclusion or other adverse effects, particularly where people may be affected by its results. We aim to make the systems and guidance we deliver understandable and usable by their intended audiences.
We protect information entrusted to us, including client data, source code, business knowledge and personal information.
We do not submit sensitive or confidential client material to AI services unless that use is authorized and appropriate safeguards are in place. Client material is not made available for training third-party AI models or unrelated secondary uses without explicit written client authorization.
We assess relevant legal, contractual, privacy, security and jurisdictional obligations, as well as provider practices for access, storage, retention, deletion and data use. We use information only to the extent needed for the agreed purpose and respect third-party intellectual property and licensing obligations.
AI-assisted code and AI-enabled software are held to the same engineering standards as other software. Generated code is not presumed correct, secure, maintainable or ready for production.
We apply proportionate architecture review, code review, testing, security assessment and release controls according to the requirements and risks of the system.
Where systems use AI to access information, make recommendations or take actions, we assess their data boundaries, permissions, dependencies and failure modes. Automation is authorized within defined limits, with appropriate monitoring, traceability and means to intervene, reverse or stop its actions.
We explain material AI use in our work and in the systems we deliver, including its purpose, limitations and the verification required. We do not represent unverified AI-generated content as established fact or overstate what an AI system can reliably do.
We remain accountable for the work we deliver. Responsibilities for decisions, approvals and ongoing operation are made clear with clients.
Where consequences are significant, people must retain meaningful authority to review, question, override or stop AI-supported decisions and workflows.
AI tools, models and risks change. We review relevant tools and practices, maintain the skills needed to use them responsibly and, for deployed solutions within an engagement's scope, agree how performance, emerging risks, incidents and changes will be monitored and addressed.
We are prepared to modify or discontinue an AI-enabled approach when it no longer meets its purpose or acceptable risk.
This policy states our standards of practice, not a claim that any AI system is infallible or that a tool alone establishes legal compliance. Engagement-specific controls and responsibilities are documented as appropriate.
For questions about this policy or AI in a Quicksilver engagement, contact [email protected].
For information about this website's handling of personal information, see our Privacy Policy.