AgenticFides logoAgenticFides

Assess. Underwrite. Aggregate Agentic AI Risk exposure.

AgenticFides builds Assess™, Workbench™, and Horizon™ — product infrastructure for organizations evaluating, insuring, advising on, and governing deployed Agentic AI systems.

Agentic AI Risk infrastructure
Three products, one proprietary method.
Assess™
Risk assessment for deployed agentic systems, workflows, and agent estates.
Workbench™
Underwriting evidence, reviewer workflow, decision records, and renewal support.
Horizon™
Portfolio and accumulation analytics across shared AI dependencies.
Powered by proprietary AI Agent Risk Classification Framework (AARCF™), Layered Risk Score (LRS™), and Insurability Matrix™ methodology each following a detail scoring logic, thresholds, and calibration rules.
01 — Agentic AI Risk

There is a new risk category.

Agentic AI Risk emerges when AI systems can plan, act, use tools, access enterprise systems, influence customers, affect legal or financial outcomes, or operate across vendors and portfolios.

Law firms need to advise on it. Consultancies need to assess it. Enterprises need to govern it. Carriers and reinsurers need to underwrite and aggregate it. AgenticFides builds the product infrastructure for all of them.

Existing AI policies, model-governance programs, cybersecurity controls, and professional-liability frameworks remain important. They are not enough, on their own, to evaluate a deployed agentic capability in practice.

Common Agentic AI Risk pathways.

i.
Autonomous decision failure
Faulty operational, financial, customer, claims, pricing, or triage decisions.
ii.
Unauthorized tool or workflow execution
Tool, API, workflow, or permission use outside the intended envelope.
iii.
Data, privacy, and confidentiality loss
Improper access, disclosure, transfer, retention, or use of sensitive information.
iv.
Manipulation and adversarial misuse
Prompt injection, poisoned context, retrieval manipulation, or social engineering.
v.
Faulty delegation of authority
Agents granted, assuming, or exercising authority beyond the intended mandate.
vi.
Multi-agent or vendor cascade
Interacting agents, tools, models, or vendors amplifying errors or dependencies.
vii.
Regulatory, contractual, or claims error
Wrong notices, commitments, rights determinations, coverage statements, or legal positions.
viii.
Operational resilience failure
Unreliable performance, weak fallback, failed recovery, or disrupted business services.
ix.
Portfolio accumulation and concentration
Correlated exposure across model families, vendors, tools, jurisdictions, or business functions.
02 — Products

From assessment to underwriting to portfolio accumulation.

Assess™ is the risk assessment layer. Workbench™ is the underwriting and evidence workflow layer. Horizon™ is the portfolio and accumulation analytics layer.

AgenticFides Assess™ · Example Agentic AI risk assessment
Assessment object
Customer-service agent estate
Use cases: law firm diligence · consultancy review · carrier submission · governance review.
Structured risk profile
Autonomy and authorityelevated
System and data accessmaterial
Adversarial exposurehigh
Oversight and evidencedeveloping
Business impactmaterial
Assessment outputs
  • Risk profile
  • Evidence-gap view
  • Control recommendations
  • Insurance-readiness view
  • Assessment report
Illustrative Assess™ product view. Public examples intentionally omit scoring logic, thresholds, and calibration rules.
Private beta
AgenticFides
Assess™
Law firms · Consultancies · Carriers · Insurers · Reinsurers · Enterprise risk teams

Standalone risk assessment for deployed Agentic AI systems, workflows, and agent estates. Assess™ produces a structured risk profile, evidence-gap view, control recommendations, and assessment report for advisory, governance, diligence, insurance submission, or underwriting review.

Private beta
AgenticFides
Workbench™
Carriers · Syndicates · MGAs · Reinsurers · Underwriting teams

Underwriting and evidence workflow for Agentic AI Risk. Workbench™ supports pre-bind and renewal review, evidence collection, reviewer workflows, underwriting files, control conditions, decision records, and exports.

In development
AgenticFides
Horizon™
Carriers · Reinsurers · Syndicates · Retro · ILS diligence

Portfolio and accumulation analytics for Agentic AI Risk. Horizon™ helps identify correlated exposure across models, vendors, tools, data sources, jurisdictions, and business functions before concentration builds.

Claims / Incident Feedback Loop

A cross-product capability that maps AI-related incidents into renewal, remediation, claims, and portfolio signals. It helps teams learn from real events without treating the illustrative examples shown here as actual claims or coverage determinations.

03 — Method & Advisory

Methodology.

Standards are the starting point. ISO 42001 and the NIST AI RMF confirm one thing: a company has an adequate management system. They don't tell you whether a specific agent is insurable, how autonomously it runs, or what it could cost when it fails. Those standards are entry criteria. Agentic Fides is powered by a proprietary approach to evaluating deployed Agentic AI systems in practice pursuing a holistic approach.

i.

Assess-led risk review

For law firms, consultancies, insurers, and enterprise risk teams evaluating deployed Agentic AI systems, workflows, or agent estates.
ii.

Insurance-readiness review

Helps insureds and advisors understand evidence gaps, control expectations, authority boundaries, and risk posture before insurance submission or renewal.
iii.

Underwriting and policy design

For carriers, MGAs, syndicates, and reinsurers designing underwriting questions, evidence expectations, control conditions, exclusions, endorsements, and renewal triggers.
iv.

Portfolio accumulation review

Uses Horizon™ to evaluate shared model, vendor, tool, jurisdiction, data-source, control-pattern, and business-function dependencies.
v.

Incident and claims simulation

Maps realistic Agentic AI events into causation, remediation, renewal, claims, and portfolio signals.

Boundary. Recognized AI, cybersecurity, privacy, operational-resilience, and insurance frameworks may support governance evidence. They do not replace deployment-level assessment or carrier-specific underwriting judgment.

AgenticFides tools support risk assessment, underwriting, governance, and portfolio decisions. Final pricing, policy wording, coverage, claims, legal, regulatory, and board decisions remain with the responsible organization.

Team

One lawyer. One AI researcher.

The same two principals lead the methodology, advisory work, and product direction behind AgenticFides.

Co-Founder, Legal & Governance
Dr. Matthias Artzt
Dr. Matthias Artzt
Financial services, AI & data protection
Matthias has spent his career as an attorney and in-house counsel in the financial services industry, specializing in artificial intelligence and data protection law. He is a frequent speaker and the editor and author of numerous specialist publications in both fields. His work spans governance frameworks, policy structures, and risk management.
Co-Founder, AI & Technology
John deVadoss
John deVadoss
Frontier AI, machine learning & cloud computing
John is a frontier-AI researcher and start-up founder with a successful enterprise-AI exit and a multi-decade background in enterprise software engineering and cloud computing. He has authored multiple books on software architecture and speaks frequently at global technology forums. His work spans machine learning, agentic architectures, and multi-agent systems.
Research & writing

On the record.

Published in: IEEE Spectrum · Solicitors Journal · The Hill · Int'l In-house Counsel Journal · Kluwer Law International · arXiv · SSRN

Op-ed
Agentic AI and the law
Artzt & deVadoss · Solicitors Journal, Jul 2026
Research
Designing Intelligent Enterprise Agents: A Capability-Aligned Multi-Agent Architecture
deVadoss · arXiv, May 2026
Op-ed
Agentic AI is acting, but the law still assumes humans
Artzt & deVadoss · Solicitors Journal, Apr 2026
Op-ed
Don’t Regulate AI Models. Regulate AI Use
deVadoss · IEEE Spectrum, Feb 2026
Journal article
PromptCraft’s Gray Zone: Owning Agentic AI Workflows Without Owning Ideas
deVadoss · Int’l In-house Counsel Journal, Dec 2025
Research
Agentic Economics: A Novel Macroeconomic Theory, Production Architecture and Dynamics of AI
deVadoss · SSRN, Nov 2025
Research
A Hypothetico-Deductive Approach towards Human-AI Interaction: Conjectures, Refutations, and Design Implications
deVadoss · SSRN, Aug 2025
Op-ed
Congress should create a public AI wealth fund
deVadoss · The Hill, Aug 2025
Journal article
Legal and Governance Issues regarding Agentic AI in the EU and Japan
Artzt, Ahn & Otake · Int’l In-house Counsel Journal, Jun 2025
Research
A Byzantine Fault Tolerance Approach towards AI Safety
deVadoss & Artzt · arXiv, Apr 2025
Op-ed
AI safety: why a new approach is needed
Artzt & deVadoss · Solicitors Journal, Mar 2025
Standards report
AI Convergence Report — GSMI 6.0: Toward Decentralized & Open-Source AI: Transparency, Privacy, Security, and Reliability
deVadoss (working-group chair) · Global Blockchain Business Council, 2025
Book
International Handbook of AI Law: A Guide to Understanding and Resolving the Legal Challenges of Artificial Intelligence
Artzt (co-editor) et al. · Kluwer Law International, Dec 2024
Journal article
Can blockchain technology help mitigate the black box phenomenon of AI applications?
Artzt & deVadoss · Int’l In-house Counsel Journal, Oct 2024
Op-ed
How the EU and UK diverge in regulating AI
Artzt · Solicitors Journal, Aug 2023
Get in touch

Request access.

We are onboarding a limited number of design partners and qualified users across law firms, consultancies, carriers, insurers, reinsurers, brokers, MGAs, and enterprise risk teams. Initial conversations are confidential and can be made subject to an NDA.

Please do not submit confidential evidence, client materials, insured data, or privileged information through this form. Qualified users receive a secure intake process after onboarding.