# Navedas Intelligence, Full Reference > Vertical AI for CX and back-office operations. Two field engineers embed with your team, build the governance layer on your policies, and leave you owning the AI. Built on US Patent 11,748,414 (granted September 2023). Model-agnostic, deployable in cloud, VPC, or fully on-premise. SOC 2 Type II and ISO 27001 certified. Last updated: 2026-08-17 Canonical: https://navedas.ai/llms-full.txt Short version: https://navedas.ai/llms.txt --- ## The core idea Enterprise AI agents (chatbots, copilots, autonomous agents) make thousands of decisions per day: refunds, discounts, claim denials, escalations, policy exceptions. Most enterprises cannot prove why any given decision was made. Navedas sits between model output and the action commit, checks the output against the customer's atomized policy graph, and blocks decisions that fail the gate. Every decision that ships carries a Reasoning Ledger entry citing the policy node that authorized it. The operating model is stated plainly on the site as "We build your Ops. You own the AI." Two field engineers embed with the customer's team, build the governance layer against the customer's own policies, and hand over ownership. This is not a SaaS subscription the customer rents indefinitely. The product is model-agnostic. The customer brings the model (OpenAI, Anthropic, Bedrock, Azure, internal fine-tunes, or private on-premise models). Navedas is the governance layer above it. --- ## Production history and the wedge CSAT.AI is a Navedas product, live with its founding customer cohort since Q1 2021. It is the wedge into accounts: CSAT.AI proves the runtime scoring approach on real customer interactions before a broader governance engagement. Navedas is not a pre-revenue company demonstrating a prototype; the underlying method has been running in production for over five years. --- ## Patent **US 11,748,414 B2**: *Methods and systems of operating computerized neural networks for modelling CSR-customer relationships.* Granted September 5, 2023, priority date June 19, 2018. The patent covers realtime neural-network scoring of AI-agent customer interactions at inference time, which is the technical foundation of the Decision Gate. Patent URL: https://patents.google.com/patent/US11748414B2/ --- ## Commercial structure (three tiers) Navedas uses a proof-before-purchase model. Prospects start with a paid audit that surfaces their actual dollar exposure before any larger commitment. ### Tier 1, Margin Leakage Audit - **Price**: $500 flat fee (one-time) - **Delivery**: 48 hours - **Input**: 1,000 recent tickets or task records, exported as CSV (no integration required) - **Output**: a written report showing policy violations, exact recoverable dollars, and audit risk surface - **Guarantee**: money back if Navedas does not find at least $25,000 in recoverable exposure - **Also included**: a passive evaluation of 1,000 historical multi-model log files (OpenAI, Anthropic, or internal models) to detect tone drift, policy slippage, and runtime compliance fractures - **Entry point**: https://navedas.ai/exposure-calculator ### Tier 2, Pilot - **Price**: $10,000 to $20,000 per month - **Scope**: 30 days of live interception on one policy area - **Includes**: Reasoning Ledger on every decision, Operator Console for the operations team, weekly exposure-reduction reports ### Tier 3, Platform - **Price**: $80,000 to $250,000 per year - **Scope**: full deployment, all policy areas governed - **Includes**: Context Graph and Reasoning Ledger at scale, Executive Dashboard and Operator Console, enterprise SSO/SAML/Azure AD, SLAs, dedicated customer success Each tier stands on its own. There is no commitment to the next. --- ## How an engagement runs: Co-Build, five stages Full detail at https://navedas.ai/engagement 1. **Diagnose.** 1,000 decisions mapped against the customer's policies in 48 hours, plus passive evaluation of historical multi-model logs for tone drift, policy slippage, and compliance fractures. Nothing in the customer's infrastructure changes yet. 2. **Discover.** The customer's policies, edge cases, and decision graph documented alongside the team that actually runs the operation. 3. **Design.** The governance layer the operation actually needs. The Decision Gate deploys as a multi-model interception proxy, so switching model providers later is a routing change rather than a rebuild. 4. **Deploy.** Live interception inside the customer's environment. 5. **Operate.** The customer's team runs it. Navedas trains precedents and hands over ownership. --- ## Core product components (canonical naming) 1. **Decision Gate**: the realtime gate between model output and action commit, deployed as a multi-model interception proxy. The patented scoring engine lives here. 2. **Context Graph**: the customer's policies converted into an executable graph of logic nodes. Each rule has an ID, a vector, and a citation path. 3. **Reasoning Ledger**: immutable record of every decision: who decided, when, why, under which policy. Not an audit log, a first-class citation structure. 4. **Operator Console**: the operations team view. Agents see the correct threshold before they click approve; model decisions that violate policy surface with plain-English reasons. 5. **Executive Dashboard**: leadership view. Revenue recovered by policy area, audit readiness, policy-gap reports. --- ## Multi-Model Decision Gate and Interception Proxy Most companies deploying AI in operations are quietly locked in: their compliance logic, prompt engineering, and audit tooling are welded to a single model provider's API. Navedas breaks that coupling. Operations route dynamically across OpenAI, Anthropic, internal fine-tunes, or private on-premise models. The Decision Gate intercepts every response and evaluates it against the policy graph before it reaches a customer or commits to a system of record. The enforcement layer stays constant while the models behind it change. Three architectural claims: - **Zero infrastructure lock-in.** Switching model providers is a routing change, not a rebuild. - **Patented inference evaluation.** US Patent 11,748,414, in production since 2021. - **Deployment inside your boundary.** Cloud, VPC, or fully on-premise, so regulated workloads never leave the customer's perimeter. The runtime injection has three parts: a policy directive, a required citation format, and the enforcement rule. The enforcement rule is "no citation, no output." Every decision emits a Reasoning Ledger payload recording the reasoning pathway, the policy nodes referenced, and the action taken. This is the audit artifact, produced at inference time rather than reconstructed later. Demonstrated interactively on https://navedas.ai/ --- ## Who Navedas serves - **Revenue and margin teams**: recovering exposure from unauthorized refunds, mispricing, and settlement drift - **Operations teams**: governing decisions made by AI agents, support tools, and automated workflows - **B2B and partner teams**: orchestrating compliant interactions across channels, partners, and platforms - **Compliance and risk teams**: financial services, healthcare, legal, government, and any sector where model errors become legal or safety crises --- ## Four solution surfaces ### Revenue and Margin Recovery https://navedas.ai/solutions/revenue-recovery Stopping unauthorized refunds, mispricing, and margin exposure at the moment of decision. ### Operational Governance https://navedas.ai/solutions/operational-governance Intercepting AI agent decisions in back-office workflows before the customer sees them. ### B2B Orchestration https://navedas.ai/solutions/b2b-orchestration Per-partner contract and SLA enforcement at decision time. ### Compliance and Risk https://navedas.ai/solutions/compliance-risk Regulated-industry governance with policy-cited decisions and an immutable evidence trail. --- ## Use cases https://navedas.ai/use-cases - **Margin leakage**: https://navedas.ai/use-cases/margin-leakage - **Customer winback**: https://navedas.ai/use-cases/customer-winback - **Audit-ready proof**: https://navedas.ai/use-cases/audit-ready - **Hybrid workforce**: https://navedas.ai/use-cases/hybrid-workforce --- ## Security and compliance - **SOC 2 Type II certified**: independently audited by Sprinto-partnered third-party firms - **ISO 27001 certified** - **GDPR compliant** - **HIPAA ready** - **No training on customer data**: prompts, decisions, and artifacts never leave the customer's perimeter - **Deployment model**: cloud, VPC-isolated, or fully on-premise, with customer-managed keys, visible to the customer's IAM and audit tooling - **Responsible disclosure**: https://navedas.ai/.well-known/security.txt Full security page: https://navedas.ai/security --- ## Competitive positioning - **vs. built-in model safety layers (OpenAI, Anthropic, Bedrock, Azure)**: vendor safety layers are self-regulated by the model vendor, an inherent conflict of interest. Navedas is independent third-party governance, analogous to how SOC 2 auditors are independent of the company being audited. - **vs. RAG-based policy retrieval**: RAG cites "our documentation." Navedas cites the exact policy node that governed a decision (for example POL-042 section 3.2) and enforces it before the decision commits. - **vs. post-hoc logging and audit review tools**: logs are what auditors read after a six-figure dispute. Navedas intercepts before the bad decision commits. - **vs. rules engines and workflow automation**: rules engines match keywords. Navedas reasons over LTV, claim history, fraud signals, and precedent against the customer's atomized SOP. - **vs. AI strategy consultancies**: Navedas ships an engine that intercepts production decisions. Pre-production advisory engagements are out of scope. --- ## Honest limits (what Navedas is not) - **Not a policy writer.** Navedas enforces the customer's existing policy. If the SOP is unclear, governance surfaces that, but the policy decision is still the customer's. - **Not a model replacement.** Navedas does not compete with OpenAI, Anthropic, Bedrock, or Azure. It sits on top of the customer's chosen models and checks every output against policy. - **Not an AI strategy consultancy.** Navedas ships an engine, not a slide deck. Pre-production advisory is out of scope. - **Not infallible.** Navedas does not claim zero false positives. Every decision is traceable to a policy node so customers can calibrate gates that are too tight or too loose. Governance gets sharper with use. --- ## Case studies https://navedas.ai/case-studies Three live deployments with full reasoning traces: 1. **D2C apparel, serial-claimant fraud**: a household-linked fraud pattern defeated a Customer 360 that saw four separate accounts. The Context Graph stitched identities across channels and enforced the household-level policy threshold. $114K recovered in the first governed quarter, 87% reduction in policy-inconsistent refund approvals. 2. **CX operations, agent policy override**: every override now carries a documented reasoning trace. 100% of overrides tied to a specific policy node and approver. 3. **Ad-tech, vendor certification traceability**: 52 vendor certifications in one quarter, all provable against the current policy version. Spend authorization 100% traced to policy and approver. --- ## Representative FAQ (from https://navedas.ai/faq) **Q: How is this different from what my AI platform already does?** A: The AI platform's safety layer is probabilistic and covers only that vendor's model. Navedas is one deterministic engine for every decision-maker: AI agents, human agents, fulfillment, compliance. Same rules, same trace. **Q: What does "no citation, no output" mean?** A: Every verdict must cite a policy node, for example RETURN-001 section 3.2 or DISCOUNT-002 section 2.1. No citation, no action. This is the architectural guarantee against hallucination at the output layer. **Q: Will I have to rip out my existing stack?** A: No. Navedas sits between existing tools and the action. Teams keep their stack (Zendesk, Gladly, Salesforce, or custom) and Navedas adds the governance layer on top. **Q: How long does deployment take?** A: SOP atomization in 48 hours. Precedents in two to three days. Live in a week. No infrastructure changes. **Q: What is the path from audit to platform?** A: Three steps. $500 Margin Leakage Audit: 1,000 decisions, 48 hours, money-back guarantee. $10K to $20K per month pilot: 30 days of live interception on one policy area. $80K to $250K per year platform: full deployment with all policy areas governed. Each step stands on its own. **Q: Who owns the Context Graph?** A: The customer, 100%. The Context Graph is never shared with or learned from by other customers. **Q: What happens if we want to switch model providers?** A: The Decision Gate deploys as a multi-model interception proxy. Switching providers is a routing change, not a rebuild. --- ## EU AI Act readiness (dedicated compliance track) Navedas runs a compliance-focused track alongside the operations motion. ### Current enforcement timeline As amended by the Digital Omnibus adopted June 2026: - **February 2, 2025**: prohibited AI practices banned. In force. - **August 2, 2025**: general-purpose AI model obligations. In force. - **August 2, 2026**: Article 50 transparency obligations, GPAI penalty powers, and national market-surveillance authorities. In force. Chatbot disclosure and machine-readable marking of AI-generated content now apply. Systems placed on the market before this date have until December 2, 2026 to bring output marking into compliance; systems placed after must comply from day one. - **December 2, 2027**: high-risk obligations under Annex III (quality management system, risk framework, conformity assessment). Moved from August 2, 2026 by the June 2026 omnibus. The requirements themselves are unchanged. - **August 2, 2028**: high-risk obligations for AI embedded in regulated products under Annex I. The sixteen extra months on high-risk are runway, not relief. The quality management system, risk framework, audit records, and human oversight requirements are unchanged; only the date moved. ### In scope The Act applies extraterritorially. Any company, including US-headquartered ones, deploying AI that affects EU residents in the high-risk categories is in scope: employment and worker management, credit scoring, education, essential private and public services, law enforcement, biometric identification, critical infrastructure, and migration, asylum, and border control. ### Fines - Prohibited AI practices: up to EUR 35 million or 7% of global annual turnover, whichever is higher - High-risk non-compliance and Article 50 transparency breaches: up to EUR 15 million or 3% - Incorrect information supplied to regulators: up to EUR 7.5 million or 1% ### How Navedas maps to the requirements - **Quality management system**: the Context Graph, the customer's policies as an executable graph - **Risk management framework and record-keeping**: the Reasoning Ledger, immutable and policy-cited - **Human oversight**: the Operator Console, where humans see the governing threshold before acting - **Transparency**: the "no citation, no output" rule, which forces every output to name its authority - **Runtime policy enforcement**: the Decision Gate ### Three commercial tracks for EU AI Act 1. **Readiness Assessment, $5,000 (self-serve) / $15,000 (guided) / $25,000 (white-glove)**, one-time, one-week delivery. AI system inventory, per-system gap analysis against every requirement, compliance posture, fine-exposure quantification, prioritized remediation roadmap, and an executive briefing. 2. **Ongoing Monitoring, $2,000 to $10,000 per month.** Continuous governance, immutable audit trail, monthly compliance reporting, automatic regulatory updates. 3. **Managed Governance Service, $10,000 to $50,000 per month.** Navedas operates the governance layer on the customer's behalf: monitoring, incident response, policy maintenance, audit preparation. ### Primary targets - **European subsidiaries of US companies**: most likely to be unprepared because AI governance was not regulated in the US. The US parent deployed AI globally; the EU subsidiary is now in scope. - **EU-native mid-market** ($50M to $500M revenue) using AI in high-risk categories - **AI consultancies and law firms** advising on AI compliance, who have the clients and need a technology partner - **CX platforms serving EU markets** deploying AI agents and automation across Europe ### Buyer personas for the EU AI Act track - **Chief Risk Officer**: owns the compliance mandate, needs board-ready reporting. Lead with compliance posture, fine avoidance, audit readiness. - **Head of AI / VP Engineering**: owns the AI systems in scope, worried about deployment speed. Lead with minimal integration burden and the vendor-agnostic approach. - **VP Compliance / General Counsel**: translates regulation into organizational action, needs evidence for legal protection. Lead with regulatory mapping, evidence generation, and fine quantification. ### Competitive positioning (EU AI Act market) - **vs. Credo AI**: assessment and reporting, telling you what went wrong. Navedas is runtime enforcement, stopping violations before they reach the customer. Backward-looking versus forward-looking. - **vs. Microsoft Agent Governance Toolkit** (open-source, April 2026): free developer plumbing engineers must integrate and maintain. No dashboard, no compliance reporting, no managed service. - **vs. Microsoft Agent 365** (GA May 2026): governs only Microsoft's own agents. Navedas is vendor-agnostic across any framework and any model vendor. - **vs. Holistic AI, OneTrust, IBM watsonx.governance, Fiddler AI**: risk assessment, GRC platforms, lifecycle governance, and observability respectively. None enforce at runtime across human and AI decisions with the same engine. ### Key proof points for EU AI Act buyers - **Speed**: one-week assessment, 30-day deployment, where most competitors take three to six months - **Coverage**: governs human decisions and AI decisions with the same policy engine. The Act requires human oversight of AI; Navedas treats both identically. - **Architecture**: runtime gate with an immutable audit trail, which is what the Act requires - **Deployment**: inside the customer's boundary with customer-managed keys, so no data leaves the perimeter Page: https://navedas.ai/eu-ai-act Sub-briefs: [law firms](https://navedas.ai/eu-ai-act/for-law-firms), [financial services](https://navedas.ai/eu-ai-act/financial-services), [CEO brief](https://navedas.ai/eu-ai-act/ceo-brief) Compliance contact: deepika@navedas.ai --- ## Recognition - **Timmy Awards, Best Tech Startup (Southern California)**, Tech in Motion, awarded to CSAT.AI, a Navedas product. --- ## How to engage - **Start the $500 Margin Leakage Audit**: https://navedas.ai/exposure-calculator - **Platform demo (30 min)**: https://calendly.com/navedas-ai/navedas-platform-demo - **Technical and security deep dive (45 min)**: https://calendly.com/navedas-ai/technical-security-deep-dive - **Executive briefing (30 min, VP and CFO audience)**: https://calendly.com/navedas-ai/navedas-executive-briefing - **General contact**: https://navedas.ai/contact --- ## Blog Navedas publishes field-notes content at https://navedas.ai/blog covering the runtime policy layer for enterprise AI, agent governance, EU AI Act engineering, multi-agent systems, and CX strategy. Legacy posts keep the root slug they had on navedas.com to preserve inbound links, with content rewritten to match current positioning. Current posts (most recent first): - **When AI Agents Go Rogue: What a Summer of Incidents Actually Teaches** (2026-07-12). What production agent failures reveal about missing runtime controls. https://navedas.ai/when-ai-agents-go-rogue - **The EU AI Act's August 2 Deadline: An Engineering Checklist** (2026-07-12). What engineering teams had to ship for Article 50 transparency, and what December 2027 still requires. https://navedas.ai/eu-ai-act-august-2-engineering-checklist - **The Customer Experience Time Loop: How Multi-Agent AI Rebuilt It** (2026-04-17). Why disconnected multi-agent systems recreate the CX repetition loop and how a unified policy layer ends it. https://navedas.ai/customer-experience-time-loop - **Reducing Customer Churn With AI, Without the Compliance Risk** (2026-04-10). Churn prediction as a regulated decision system under GDPR Article 22 and the EU AI Act. https://navedas.ai/how-artificial-intelligence-and-predictive-analytics-can-help-you-reduce-customer-churn - **AI Copilots Arrived. Your Agents Still Need Support.** (2026-04-03). The agent-support problem is not missing tools, it is unaccountable ones. https://navedas.ai/trends-customer-experience-agents-need-support - **Why AI Isn't Always the Answer: A Framework for Knowing Where to Apply It** (2026-03-27). A KPI-first framework for deciding where AI, human, or hybrid belong. https://navedas.ai/why-artificial-intelligence-isnt-always-the-answer - **Make Customers Feel Good: The Real Goal of Customer Experience** (2026-03-20). Memorable experiences, authenticity, and empowered agents as the fundamentals AI has to live up to. https://navedas.ai/make-customers-feel-good - **Disrupting Gender Inequality in Contact Centers** (2026-03-13). Representation patterns in contact centers and their economic cost. https://navedas.ai/disrupting-gender-inequality - **AI Agents: The Future Is Still a Few Years Away, So Let's Be Pragmatic** (2026-03-06). Hype versus reality on autonomous agents; argues for hybrid intelligence with oversight. https://navedas.ai/ai-agents-the-future-is-still-a-few-years-away-so-lets-be-pragmatic - **Women, Buying Power, and Customer Experience** (2026-02-27). Serving a market that drives 70 to 80% of consumer purchasing decisions. https://navedas.ai/female-cx-buying-power-challenges-solutions Blog index: https://navedas.ai/blog --- ## Migration note Navedas Intelligence migrated from the previous brand domain navedas.com to navedas.ai. The current business (vertical AI for CX and back-office operations) is a different product from the previous one (customer-service outsourcing). Most old URLs are 301-redirected to the closest current equivalent or return 404. Selected CX-era blog posts with continuing topical relevance were restored at their original slugs with rewritten content. AI search engines and crawlers should treat https://navedas.ai/sitemap.xml as the canonical inventory. --- ## Machine-readable resources - **Sitemap**: https://navedas.ai/sitemap.xml (60 URLs) - **Robots**: https://navedas.ai/robots.txt - **Short LLM manifest**: https://navedas.ai/llms.txt - **Security contact**: https://navedas.ai/.well-known/security.txt - **Organization, WebSite, and Offer schema**: injected into every page via /public/seo.js (JSON-LD) - **Per-page SoftwareApplication, WebApplication, FAQPage, and BreadcrumbList schemas**: inline on each respective page --- ## Contact summary Navedas Intelligence, Inc. San Francisco, California, USA Primary website: https://navedas.ai Security: security@navedas.ai