AGILE Framework

Created: 2026-05-17 Updated: 2026-05-17 Source count: 1

TL;DR

The AGILE framework is an organising structure for AI risk in financial services introduced at the Second Financial Industry Forum on AI (FIFAI II, 23 March 2026), convened by Canada’s OSFI and the Global Risk Institute. The acronym stands for Awareness, Guardrails, Innovation, Learning, Ecosystem Resiliency. It is industry-side organising language rather than a supervisory expectation, but is now widely cited as a practical reference for board-level AI governance in financial services.

TL;DR (detail)

AGILE complements rather than replaces the BCBS ten-step playbook. Where othp90 is supervisory and bank-facing, AGILE is industry-facing and explicitly board-oriented. Paul’s positioning treats AGILE as useful framing for assurance reviews and target-state governance designs [S-2026-03-23-fifai-ii-agile].

Key Points

• Awareness — board-level AI risk literacy and understanding of agentic AI in particular [S-2026-03-23-fifai-ii-agile].
• Guardrails — human oversight for high-impact decisions, data-integrity standards [S-2026-03-23-fifai-ii-agile].
• Innovation — supporting responsible deployment rather than blanket restriction [S-2026-03-23-fifai-ii-agile].
• Learning — workforce skilling and continuous adaptation [S-2026-03-23-fifai-ii-agile].
• Ecosystem Resiliency — third-party oversight and supply-chain resilience [S-2026-03-23-fifai-ii-agile].
• Industry forum convened by OSFI (Canada’s prudential supervisor) and the Global Risk Institute [S-2026-03-23-fifai-ii-agile].
• Published 23 March 2026; Fasken summary “Adopting AI in Financial Services: Key Takeaways from the Second Financial Industry Forum on AI” published April 2026 [S-2026-03-23-fifai-ii-agile].
• Reports that leading firms are 1.7× more likely to have a Responsible AI framework [S-2026-03-23-fifai-ii-agile].

Detail

Origin and authority

FIFAI II was convened by OSFI and the Global Risk Institute in March 2026 as the second Financial Industry Forum on AI. AGILE is the headline organising structure to emerge from the forum, communicated principally through Fasken’s April 2026 practitioner summary “Adopting AI in Financial Services” [S-2026-03-23-fifai-ii-agile]. The framework is industry-side, not supervisory — but the OSFI / GRI sponsorship gives it more credibility than a pure consultancy product.

Relationship to supervisory frameworks

AGILE intentionally complements (rather than replaces) the BCBS ten-step playbook. The pillars map approximately:

• Awareness → BCBS step 1 (interdisciplinary AI committee) and step 9 (workforce skilling).
• Guardrails → BCBS steps 2 (responsible-AI principles), 6 (risk and control assessments), 7 (ongoing monitoring).
• Innovation → No direct BCBS equivalent — AGILE-specific.
• Learning → BCBS step 10 (continuous review).
• Ecosystem Resiliency → BCBS d605 third-party risk principles.

Practitioner read

Fasken’s framing — that AI oversight sits at the intersection of three disciplines (model risk thinking, adversarial testing, technical model understanding) and no single existing function owns it — is the central qualitative finding [S-2026-03-23-fifai-ii-agile]. This connects directly to the where-does-AI-governance-sit tension.

Practical Applications

• Use AGILE as the organising frame for board-level governance reviews — five pillars are intuitive and translatable to board reporting.
• Map a client’s current controls to each pillar to surface gap areas in target-state design [S-2026-03-23-fifai-ii-agile].
• Cite the 1.7× statistic as a comparative benchmark in maturity assessments.

Related Concepts

• maps-to → Three Lines of Defence for AI — AGILE’s “Guardrails” pillar maps onto 1LoD / 2LoD coverage.
• relates-to → BCBS AI Governance Framework — complementary supervisory framework.
• relates-to → AI Governance Maturity Gap — AGILE is the convergent industry frame for the gap.
• relates-to → Model Risk Management and Agentic AI — connects via the Fasken cross-discipline framing.

Open Questions

• Whether the AGILE structure will be adopted by supervisors as a reference lens or remain industry-side guidance.
• Methodology behind the 1.7× statistic — not captured in ingestion.

Sources

• [S-2026-03-23-fif