NIST AI RMF 1.1 — A 90-Day Implementation Playbook for 2026

TL;DR — NIST AI RMF 1.1 organizes AI risk into four functions: Govern, Map, Measure, Manage. The Generative AI Profile (AI 600-1) added LLM-specific risks. Most enterprise buyers now ask vendors to attest against it. Run a 90-day rollout: 30 days governance, 30 days inventory + risk-tier, 30 days monitoring + validation.

What the norm says

The framework is voluntary, sector-agnostic, and built around four functions:

• Govern — culture, accountability, policies that span the AI lifecycle.
• Map — context, intended use, stakeholders, downstream impacts.
• Measure — analyze, benchmark, and monitor risk; quantify trustworthiness characteristics.
• Manage — prioritize, treat, and document residual risk.

Seven trustworthiness characteristics anchor the model: valid and reliable, safe, secure and resilient, accountable and transparent, explainable and interpretable, privacy-enhanced, and fair with harmful bias managed.

The Generative AI Profile (NIST AI 600-1) layers in LLM-specific risks — confabulation, data leakage, dangerous content, model collapse — and is the version most procurement teams reference in 2026.

flowchart LR
 G[Govern] --> M[Map]
 M --> ME[Measure]
 ME --> MA[Manage]
 MA --> G
 G --> POL[Policies + roles]
 M --> INV[AI inventory + risk tier]
 ME --> EVAL[Bias + safety evals]
 MA --> MON[Continuous monitoring]

What this means for AI vendors

Procurement will ask three things:

1. Show the inventory — every model, prompt, tool, dataset, and deployment with an owner.
2. Show the evals — pre-deployment safety + bias scoring, post-deployment drift checks.
3. Show the kill switch — who shuts a model down, on what signal, in how many minutes.

The U.S. Treasury's February 2026 Financial Services AI RMF translates the four functions into 230 control objectives — that is the level of granularity bank buyers now expect.

CallSphere posture

CallSphere maps every product surface to the four RMF functions. 37 agents across 6 verticals, 90+ tools, and 115+ DB tables are inventoried with owner, risk tier, and last-evaluated date. HIPAA controls cover Govern and Manage; vertical eval suites cover Measure.

• Act — $49/mo · pre-flight safety evals included
• Orchestrate — $99/mo · coordinates processes across multiple systems · custom risk profiles per workspace
• Custom Build — $149/mo · bespoke integrations and unlimited custom development · NIST RMF attestation pack + quarterly review

Explore a live demo and compare current plans to find the right fit for your business.

Compliance checklist

1. Stand up an AI Governance Council with a named accountable executive.
2. Build an AI inventory — model, version, owner, vertical, risk tier.
3. Map each system to RMF risk categories and the GenAI Profile.
4. Define eval gates for safety, bias, privacy, and groundedness before release.
5. Publish a model card per agent with limits and recommended uses.
6. Wire continuous monitoring to a dashboard with drift, refusal-rate, and bias deltas.
7. Document a kill-switch playbook with named on-call.

FAQ

Q: Is NIST AI RMF mandatory? No, but federal acquisition, Treasury, and most Fortune 500 procurement now require attestation.

Q: Difference between RMF 1.0 and 1.1? 1.1 adds clarifications, the GenAI Profile, and updated trustworthiness characteristics. The four-function structure is unchanged.

Q: How long does adoption take? A 90-day phased rollout (30/30/30) is realistic for a focused team. Mature programs continue to refine for 12-18 months.

Q: Does it conflict with ISO/IEC 42001? No — they map cleanly. RMF is risk-led, ISO 42001 is management-system-led. Most teams adopt both.

Q: How does it apply to voice agents specifically? The GenAI Profile covers TTS and ASR risks (synthetic media, dialect bias, accent disparate impact). Add voice-specific evals to your Measure function.

Sources

• AI Risk Management Framework — NIST
• NIST AI 100-1 RMF 1.0 PDF
• NIST AI RMF Implementation Guide — RiskTemplate
• Navigating the NIST AI RMF — OneTrust

NIST AI RMF 1.1 — A 90-Day Implementation Playbook for 2026: production view

NIST AI RMF 1.1 — A 90-Day Implementation Playbook for 2026 sits on top of a regional VPC and a cold-start problem you only see at 3am. If your voice stack lives in us-east-1 but your customer is calling from a Sydney mobile network, the round-trip time alone wrecks turn-taking. Multi-region routing, GPU residency, and warm pools become the difference between "natural" and "robotic" — and it's all infra, not the model.

Serving stack tradeoffs

The big fork is managed (OpenAI Realtime, ElevenLabs Conversational AI) versus self-hosted on GPUs you operate. Managed wins on cold-start, model freshness, and zero-ops; self-hosted wins on unit economics past a certain conversation volume and on data residency for regulated verticals. CallSphere runs hybrid: Realtime for live calls, self-hosted Whisper + a hosted LLM for async, both routed through a Go gateway that enforces per-tenant rate limits.

Latency budgets are non-negotiable on voice. End-to-end target is sub-800ms ASR-to-first-token and sub-1.4s first-audio-out; anything beyond that and turn-taking feels stilted. GPU residency in the same region as your TURN servers matters more than choosing a slightly bigger model.

Observability is the unglamorous backbone — every conversation produces logs, traces, sentiment scoring, and cost attribution piped to a per-tenant dashboard. HIPAA aligned isolation keeps healthcare traffic separated from salon traffic at the storage layer, not just the API.

FAQ

Is this realistic for a small business, or is it enterprise-only? The IT Helpdesk product is built on ChromaDB for RAG over runbooks, Supabase for auth and storage, and 40+ data models covering tickets, assets, MSP clients, and escalation chains. For a topic like "NIST AI RMF 1.1 — A 90-Day Implementation Playbook for 2026", that means you're not starting from scratch — you're configuring an agent template that's already been hardened across thousands of conversations.

Which integrations have to be in place before launch? Day one is integration mapping (scheduler, CRM, messaging) and prompt tuning against your top 20 real call transcripts. Day two through five is shadow-mode running, where the agent transcribes and recommends but a human still answers, so you can compare side-by-side. Go-live is the moment your eval pass-rate clears your internal bar.

How do we measure whether it's actually working? The honest answer: it scales until your tool catalog gets stale. The agent is only as good as the integrations it can actually call, so the operational discipline is keeping schemas, webhooks, and fallback paths green. The platform handles the rest — observability, retries, multi-region routing — without your team owning the GPU layer.

Talk to us

Explore a live demo and compare current plans to find the right fit for your business.