The promise of Artificial Intelligence (AI) to transform the investment management industry is immense, from productivity gains and enhanced decision making to superior client service. But with great power comes great responsibility. Firms must adopt AI in a way that is ethical, compliant, and secure.
Imagine entrusting critical financial decisions to a black box, an AI system operating without transparency or oversight. Blind faith in such an opaque technology can lead down perilous paths, yielding erroneous results that erode trust, invite regulatory scrutiny, and ultimately damage your firm’s reputation and bottom line. Without a robust Responsible AI Framework, your organization is vulnerable to biased algorithms, data privacy breaches, and decisions made without clear accountability. Like navigating a minefield in the dark, proceeding with AI adoption without a structured framework is a gamble your firm cannot afford to take.
At Olmstead, we’ve seen a tangled web of regulations and standards, each with good intentions but sometimes difficult to navigate. To help our clients cut through the complexity and achieve AI success, we’ve developed a single, unifying framework: SPARTAH.
Much like how ancient Spartans were disciplined, our SPARTAH framework provides the structure and principles necessary for a firm to build and maintain a strong, responsible AI program. It’s a pragmatic approach designed to ensure that by following one comprehensive framework, your firm will be well on its way to compliance with guidance and regulations like the EU AI Act, the OECD AI Principles, and the NIST AI Risk Management Framework – just to name a few.
SPARTAH is an acronym for the seven pillars of a disciplined, responsible AI strategy:
S – Supervision
A successful AI strategy requires more than just good intentions; it demands corporate oversight. We advocate for enterprise oversight, like a Center of Excellence (CoE) model, to foster appropriate and responsible AI adoption. Staffed with specialized AI expertise, it drives innovation, verifies adherence to policies, and manages costs. It ensures that the firm’s approach is consistent, strategic, and aligned with business goals.
P – Privacy
Data is a critical ingredient for AI, but its use must be handled with the utmost care. Responsible AI solutions must respect data privacy and security, adhering to a “need-to-know” principle. This means one customer’s private data must never be used in another customer’s analysis, and firms must implement robust controls to prevent data leakage. Protecting client information is not just a regulatory requirement—it is a foundational principle of trust.
A – Accountability
AI solutions can feel like a black box, but when something goes wrong, you cannot simply “blame the AI.” Every AI solution requires a clear chain of accountability that ultimately leads to a specific person. That individual must ensure the solution operates correctly, is safe to use, and is consistent with all internal policies and standards. This personal accountability ensures that the firm has skin in the game and that there are clear consequences for negligence or policy violations.
R – Risk-based
Not all AI solutions are created equal. For example, an AI-powered chatbot used internally has a different risk profile than an AI solution making real-time trading decisions. A responsible AI framework must classify solutions by risk, with controls and governance processes that scale accordingly. This risk-based approach ensures that resources are allocated effectively, with the highest-risk applications receiving the most rigorous oversight and validation.
T – Transparency
Transparency transforms an opaque black box into an understandable and trusted system. AI solutions should be explainable, showing how inputs lead to outputs. They should be documented and log their decision-making chains. This level of transparency is not just for regulators; it builds internal confidence and allows for independent verification and auditing of the system’s behavior.
A – Awareness
For a firm to manage its AI ecosystem effectively, stakeholders must be aware of how and where AI is used. Firms should maintain an inventory of AI solutions to keep track of every AI application. People need to be aware when results are AI generated to avoid blind faith risking the enterprise. This awareness is also critical for external parties because firms may need to disclose their use of AI to regulators, clients, and prospects, fostering a culture of openness and trust.
H – Human in the Loop
AI is a powerful tool, but it cannot run amok. The Human in the Loop (HITL) pillar ensures that AI solutions cannot operate uncontrollably. Instead, systems are a collaboration between humans and AI with humans actively involved in guiding, correcting, or making decisions alongside the AI. For example, AI might draft a report, but a human professional must review and approve it before it is sent to a customer, accepting the responsibility for its veracity.
Conclusion
The right guardrails unlock immense value and competitive advantage by enabling your firm to move fast and avoid risks. The SPARTAH framework provides a strategic, disciplined way to manage AI risk, foster innovation, and build a program that is both compliant and ready for the future. Like the Spartans who stood as a bulwark against their enemies, SPARTAH stands as a comprehensive guard, protecting your firm’s reputation, data, and future from the complex risks of the AI era.