AI governance & agentic systems
AI agents can access data, call tools and take actions before policies or oversight catch up.
What you receive: AI control register, approved-use boundaries and agentic risk treatment plan.
Scope this capabilityEngagement model / practical delivery
davisAIgov brings senior leadership and qualified specialist support to technology decisions with material operational, regulatory, transaction or litigation consequences.
Start here
Choose the scope that fits your immediate decision and resources.
Fixed-scope technology-risk reviews, supplier vetting, transaction exposure or policy evidence gaps.
Governance operating models, SIEM-A designs, remediation and control validation.
Fractional CISO, continuous evidence review and retained specialist support.
Service portfolio
AI agents can access data, call tools and take actions before policies or oversight catch up.
What you receive: AI control register, approved-use boundaries and agentic risk treatment plan.
Scope this capabilityA written policy cannot show whether AI and automated processes follow it in production.
What you receive: Telemetry map, detection logic, exception evidence and operating dashboards.
Scope this capabilitySecurity programs struggle to keep pace with expanding attack paths, operational change and recovery obligations.
What you receive: Risk roadmap, incident and recovery playbooks and management reporting.
Scope this capabilityFormal policies may describe controls that were never deployed, were bypassed or have stopped working.
What you receive: Citation-to-control mapping, tested findings, evidence gaps and prioritized closure plan.
Scope this capabilityAfter an attack, disputed technical facts, missing records and inaccurate loss narratives create additional exposure.
What you receive: Technical chronology, ESI/discovery plan and defensible issue and loss-evidence narrative.
Scope this capabilityNew vendors, emerging AI tools and service providers can introduce hidden dependencies or control weaknesses.
What you receive: Risk-tiered qualification file, material gaps and conditional approval criteria.
Scope this capabilityKnow the cyber, AI, privacy and supplier exposure before an acquisition, financing event or critical vendor approval.
SERVICES / DAVIS & ASSOCIATES ADVISORY, LLC
New AI deployments can add vendors, permissions and data connections faster than teams can review them. We help identify the immediate business risk, reduce unnecessary approval delays and define testable controls, owners and evidence before unmanaged technical debt becomes an expensive incident or remediation project.
We help leaders move from a specific problem to a defined engineering or advisory engagement. We identify the exposed systems and business consequences, agree who owns each decision, test the safeguards and produce records that show what worked and what remains uncertain. Scope, expertise and cost are agreed before sensitive materials are requested.
Begin with a short introductory message: the business decision, jurisdiction, sector, operating problem, incident, transaction or regulatory question you need to address. Sensitive material should be shared only after scope, communication preferences and any necessary engagement terms are confirmed.
Services are scoped by written agreement. Prices are quoted after scoping; no checkout or recurring billing is enabled. Specialist legal, tax, assurance and managed-security services require the appropriate engagement and provider.
Provide a concise summary of the issue, affected business unit, time sensitivity and the outcome you need. This can be a board question, regulator request, incident, diligence need or implementation decision.
We review the issue, identify likely workstreams, confirm whether the matter fits the advisory model and clarify which disciplines are implicated: legal, cybersecurity, privacy, operations, resilience or evidence.
Following agreement on scope and engagement terms, appropriate arrangements can be established for further materials. Detailed or sensitive documents are not requested through the public contact form.
SPECIALIST PRACTICES / GOVERNANCE & EVIDENCE
Boards, examiners and courts may ask different questions about the same event. We test the underlying facts and maintain traceable records so conclusions can be explained without claiming more certainty than the evidence supports.
From the examination room to the courtroom: senior, hands-on judgment at the intersection of cybersecurity, AI governance and defensible evidence.
REGULATOR-FACING SPECIALIST
We help security and compliance teams turn policy into checks they can operate: owners, approved actions, monitored exceptions, recovery steps and records that can be produced during an examination.
AI governance has a security function that needs a named owner. We bring an active CISO’s perspective to making that responsibility operational.
Richard E. Davis serves as Chief Information Security Officer at RiteCheck, an NYDFS-regulated financial services business. We bring that operating perspective to AI governance: define the risk, assign responsibility, test the control and organize the evidence so an examiner can follow it.
The CISO leads AI-related cybersecurity risk. Legal, compliance, privacy and business owners retain their respective responsibilities; board oversight remains essential. This is shared accountability with a clear security owner, not a claim that every AI obligation belongs exclusively to the CISO.
NYDFS’s AI cybersecurity guidance applies the existing Part 500 framework to AI-related cyber risks; it does not create a separate AI certification. Applicability and exemptions must be assessed for the entity.
“Examiners will be human long after analysts aren’t.”
Richard E. Davis · A working premise for accountable governance
Wide reference table: scroll horizontally to read every column. Keyboard users can focus the table and use the arrow keys.
| The question | Engagement deliverable | What leadership can use |
|---|---|---|
| What AI is in use, and who owns its risks? | AI and vendor inventory; risk assessment; decision-rights and ownership matrix; applicable Part 500 control mapping. | A documented scope and named accountability, including material exceptions. |
| Do the controls operate as described? | Control-to-evidence register; targeted testing; supplier assurance review; AI DDR telemetry and evidence-coverage design. | Test results, source provenance and explicit coverage limitations rather than policy assertions alone. |
| Can management answer and act? | Examiner-response support; indexed evidence package; CISO and board reporting; remediation tracker and response rehearsal. | A coherent account of the risk, control, result and corrective action, with responsible owners. |
For: CISOs, compliance officers, general counsel, boards and NYDFS-regulated organizations. Engagement: readiness review, examination support or ongoing advisory. No examination outcome or regulatory endorsement is promised.
Discuss examination readiness ↗AI INCIDENT EXPERT WITNESS & LITIGATION SUPPORT
An AI agent may produce a convincing explanation while the actual tool call violated policy. We compare what was proposed, what was authorized, what ran and what changed to establish the limits of the available evidence.
When an AI-enabled process causes loss, a fluent account of events is not enough. Counsel needs an independent technical reconstruction that can be examined, challenged and explained.
We translate agent spans, tool calls, delegated identities, policy decisions and downstream records into a comprehensible account of system behavior. The analysis distinguishes the agent’s narration from independent evidence, and an observed sequence from a supported conclusion about cause.
AI DDR — AI Deviation Detection and Response — supplies the investigative frame: compare approved authority with observed behavior, identify control failures and preserve the evidence behind each finding. A trace ID links observations; it does not, by itself, authenticate a person or prove an outcome.
Expert opinions remain independent of the retaining party’s desired result. Alternative explanations, missing telemetry, sampling, clock uncertainty and limits on reproducibility belong in the analysis, not in a footnote after the conclusion.
Wide reference table: scroll horizontally to read every column. Keyboard users can focus the table and use the arrow keys.
| Workstream | What the work examines | Potential deliverable |
|---|---|---|
| Preservation & reconstruction | Model and tool versions; prompts where lawfully available; span context; identities; approvals; retrieval sources; original target-system logs and chain of custody. | Evidence inventory, preservation plan and a source-linked event chronology. |
| Authority, controls & causation | Approved scope versus executed actions; policy-gate and dispatcher records; access controls; evidence gaps; alternative explanations and downstream effects. | Reproducible technical analysis, control-failure assessment and clearly bounded opinions. |
| Reports, testimony & demonstratives | Technical issues within the agreed scope, with methods, assumptions and limits made explicit for the intended audience. | Consulting analysis or an expert report; deposition/trial support and testimony where retained and qualified. |
For: counsel, insurers, boards and investigators. Engagement: conflict check, written scope and agreed evidence access. Expert qualification and admissibility are case-specific and determined by the court; no outcome is guaranteed.
Discuss an expert engagement ↗Explore the method: inter-agentic threat atlas · harness telemetry to Elastic. The interactive resources are educational simulations, not evidence from a client matter. For the federal evidentiary framework, see the Federal Rules of Evidence, including Rule 702.
BROADER ADVISORY
Separate tools and approvals often create expensive handoffs, duplicated reviews and gaps in ownership. We connect cyber, AI, privacy and resilience work around business processes so risk controls support delivery instead of becoming a parallel paperwork system.
Specialist work connects to the full range of AI governance, cyber resilience and executive advisory services below.
Monthly retainer
A virtual CISO helps define security priorities, decision owners, risk reporting and incident readiness. Agree on operational authority and test recovery, not just policy completion.
Executive security leadership without adding a full-time CISO role.
Strategy and oversight; 24/7 monitoring and hands-on remediation require separate scope.
Discuss this engagement ↗Project or retainer
Identify the AI system, intended use, owner and permitted actions before deployment. Then test data access, human oversight and the ability to stop harmful activity.
Move useful AI into operation with clear authority, testing and oversight.
Illustrative 4-6 week planning scope; dependent on systems, access and stakeholders.
Discuss this engagement ↗Project
Identify which rules actually apply, who owns each control and what records prove it operated. Avoid treating policy documents as complete audit evidence.
Understand which questions to resolve and assemble evidence that supports a defensible response.
Not a guarantee of compliance or a substitute for a jurisdiction-specific legal opinion.
Discuss this engagement ↗Project or retainer
Set escalation triggers, decision authority and recovery roles before an incident. Rehearse realistic failures to reduce confusion and prolonged service interruption.
Prepare leaders to contain disruption, preserve evidence and coordinate decisions under pressure.
No emergency or 24/7 response service is implied without a separate agreement.
Discuss this engagement ↗Project or retainer
Map sensitive information, its permitted uses, retention and sharing. Test technical restrictions so privacy depends on enforced controls, not promises alone.
Know what information matters, where it goes and who is accountable for its use.
Local counsel and specialist tax/privacy advice may be required; no automatic DPO appointment.
Discuss this engagement ↗Project
Join AI activity with independent permission and outcome records. A monitoring alert is a lead for investigation, not proof that a control was effective.
Turn operating signals into evidence leaders can use.
Design/advisory scope; implementation, licensing and managed operations are separately agreed.
Discuss this engagement ↗Project or retainer
Inventory outside services, libraries, data sources and agent tools. Check version, origin, access and maintenance so a supplier weakness does not become a hidden internal risk.
Understand the risk inherited from critical technology and business providers.
Assessments rely on available evidence; testing requires written authorization.
Discuss this engagement ↗Session
Explain the business decision, possible harm, control owner, available evidence and choice required. Technical detail should support—not obscure—the decision.
A decision-focused session on the risk, choices and evidence leadership needs.
Illustrative planning range; preparation and stakeholder scope to be agreed.
Discuss this engagement ↗Workshop
Map a real workflow through data, model, tools, permissions and evidence. Identify where an independent check must prevent an unauthorized business effect.
See where a proposed AI action becomes a business consequence.
Educational/advisory workshop, not a production implementation or certification.
Discuss this engagement ↗Project or retainer
Separate common technical controls from jurisdiction-specific legal duties. Confirm the operating facts and local requirements before relying on shared documentation.
Connect headquarters oversight with the obligations of the markets, data and operations you actually touch.
Local licensed counsel, tax and invoicing specialists engaged where needed; language capacity confirmed in scope.
Discuss this engagement ↗Specialist research
State the question, evidence sources, assumptions and decision supported. Label uncertainty so management can distinguish a finding from an informed hypothesis.
A source-linked research starting point tailored to the organization and its priorities.
Bespoke research and update schedules are agreed separately. All published website resources are freely accessible.
Discuss this engagement ↗Davis & Associates Advisory, LLC · Professional advisory
Bring a specific decision, incident or system change to the first conversation. We can define the required outcome, affected data and systems, minimum controls, test cases and evidence before proposing a delivery team or fee.
Describe the challenge, business objective or assurance question. An introductory discussion can determine whether advisory services are a good fit.