Podcast
One real enterprise AI deployment, dissected in full. Every Tuesday at 7 AM CT.
Host: Elise · Powered by AI Insight Lab · New episodes every Tuesday at 7 AM CT
24 of 26 episodes available now
Episodes
The Black Box Audit: What Big Four AI Tools Are Doing Inside Your Audit — and What Your Audit Committee Hasn't Asked
KPMG Clara runs analytics across 100% of journal entries — not a sample. EY Astra drafts audit memo language from flagged conditions. Deloitte Omnia surfaces anomalies before the engagement team reviews them. PwC Halo processes contracts and board minutes with GenAI. All four Big Four firms have announced Microsoft Azure AI partnerships. Your engagement letter may predate these tools. The PCAOB has signaled that AI-assisted audit procedures carry the same documentation requirements as human-performed ones. This episode dissects what your audit committee needs to ask before the next engagement — and what it cannot assume the engagement letter covers.
The Freight Intelligence Bet: What Carrier AI Routing Means for Every Logistics Director Locked Into a Contract That Assumed Human Decisions
UPS ORION routes 21M+ packages per day. FedEx committed $2B+ to its DRIVE AI network redesign. Maersk AI manages dynamic ocean freight pricing across 380+ ports. Your enterprise shipper contract — governing rates, data rights, routing accountability, and dispute resolution — was almost certainly written before any of these systems were in production. Carrier AI is training on your shipping data, applying dynamic pricing overlays within contracted headroom, and making routing decisions your SLAs don't account for. This episode dissects the data rights gap, the SLA accountability problem, and what every logistics director must add to their next carrier agreement.
The Copilot Code Gap: What Engineering Leaders Haven't Decided About AI-Written Code in Production
GitHub Copilot is active across 77,000+ organizations. Independent security research finds 36–40% of AI completions contain exploitable vulnerabilities in OWASP Top 10 categories. Copilot's IP indemnification is conditional on Enterprise tier and the code matching filter being enabled — most enterprises haven't confirmed filter status. Copilot sends code context to US-based Azure infrastructure by default, a data residency violation for European enterprises without explicit geographic configuration. SOC 2 auditors are now asking about AI code generation in SDLC documentation that predates the tool. This episode dissects the security, IP, and compliance gaps most engineering leaders haven't closed.
The 510(k) Gap: What Hospital Radiology Departments Haven't Resolved Before Their Next AI Model Update
Viz.ai is deployed in 1,100+ hospitals with 20 FDA clearances. Aidoc covers 1,200+ health systems. FDA has cleared 950+ AI/ML-based SaMD through Q1 2025 — 75%+ in imaging, most under the 510(k) predicate pathway that validates a specific algorithm version. When vendors update their radiology AI models, hospitals continue running them under the original clearance. Most radiology departments cannot confirm what version is in production, what the false positive rate is on their patient population, or whether alert threshold configuration matches their clinical protocols. This episode dissects the version staleness problem and the governance posture every health system needs before the next model update.
The Hiring Algorithm: What HireVue's 750 Enterprise Clients Haven't Filed Since NYC Local Law 144 Took Effect
HireVue is deployed at 750+ enterprises. NYC Local Law 144 took effect July 5, 2023 — requiring an annual independent bias audit, public posting of results, and candidate notice for any automated employment decision tool. Approximately one in four employers using AI hiring tools had filed. The EEOC's May 2023 guidance on AI adverse impact is in active enforcement posture. This episode dissects the compliance gap, why AI scoring embedded in standard ATS platforms is the most common uncovered exposure, what model change staleness means for the audit on file, and the four governance postures every CHRO needs to evaluate before the next hiring cycle.
The AI Drug Discovery Audit: What Recursion's Platform Bet Means for Every Pharma R&D Leader Committing Budget to an Algorithm's Hypothesis
Recursion acquired Exscientia in 2024, absorbing Sanofi's $1.2B AI drug discovery deal. An AI-designed drug completed Phase II. FDA reviewers are asking about AI methodology in pre-IND meetings. The governance gap: pharma R&D leaders are committing Phase II budgets to AI confidence scores without a protocol for validating what the algorithm actually claims — or for documenting it when FDA asks. This episode dissects the training data bias problem, the FDA documentation gap, what happens when your platform partner is acquired, and the four governance postures every pharma R&D function needs to evaluate before the next Phase II allocation meeting.
The Facilities Intelligence Bet: What CBRE's AI Deployment Means for Every Corporate Real Estate Leader
CBRE manages 2.1 billion square feet. JLL's Falcon AI covers 1.1 billion square feet. The operational results — 15–30% energy savings, predictive maintenance — are real. The governance questions most corporate real estate leaders have not answered: who owns the building intelligence your FM provider has accumulated over the contract term, what GDPR Article 28 requires when employee occupancy data flows to your facilities manager's AI platform, and what your FM agreement actually says about data portability at contract end. This episode dissects each of those questions and the governance posture every enterprise real estate team needs before the next FM renewal.
The Certification Gap: What Enterprise AI Upskilling Programs Get Wrong When Employees Complete Training But Don't Change How They Work
Accenture targeted 700,000 employees. Walmart built an AI Academy for 1.6 million associates. Deloitte ran AI University for 450,000 staff. Completion rates: 70–80%. Tool adoption at 90 days: 12–18%. Enterprise AI upskilling has a measurement problem — and the wrong metric is driving billions in L&D spend. This episode dissects the structural reason the certification-to-adoption gap persists, why corporate L&D platforms are designed to report the wrong number, and what CLOs and CHROs need to rebuild before the next budget cycle.
The Grid Intelligence Bet: What Duke Energy's AI Deployment Means for Every Utility Operations Leader
Duke Energy, National Grid, and Xcel Energy have deployed AI for grid monitoring, predictive maintenance, and outage prediction through GE Vernova and ABB. The vendors claim 20–30% reductions in unplanned outages. The governance questions utilities have not resolved: who owns the operational technology data that trains these AI systems, what NERC CIP compliance actually requires when AI models process OT network telemetry, and what the liability exposure is when an AI misclassifies a grid condition during peak demand. This episode dissects each of those unresolved questions and the governance posture every utility needs before scaling.
The Safety Stock Bet: AI Demand Forecasting and the Inventory Risk Retailers Aren't Modeling
Blue Yonder serves 76 of the Fortune 100. Relex claims 15–30% inventory reductions. Retailers are cutting safety stock based on those numbers. But AI demand forecasting has a dirty secret: it works well on stable SKUs and underperforms confidently on promotions, new product launches, and supply disruptions — exactly the moments that drive disproportionate revenue. This episode dissects the event-type performance gap most retailers haven't measured, what Target's $1.5B write-down reveals about AI forecasting governance, and the audit every retail planning team needs to run before the next safety stock reduction.
The ATO Bottleneck: What Federal Agencies Discover When AI Meets the Authorization Process
Federal agencies are deploying AI across procurement, benefits processing, and compliance monitoring — but the Authorization to Operate process was designed for static systems. FedRAMP authorizes cloud infrastructure, not AI model behavior. Most commercial frontier AI tools have no FedRAMP authorization. This episode dissects what happens when model updates make an existing ATO stale, how agencies are navigating the gap between procurement speed and ATO timelines, and what a defensible AI governance posture actually requires for federal CIOs.
The Algorithmic Underwriting Audit: What NAIC AI Requirements Mean for Every Insurer
38+ states have adopted the NAIC Model Bulletin on AI. Colorado mandates external algorithmic audits for life insurance AI. California has directly challenged AI-generated property risk scores. Lemonade, Tractable, and ZestyAI are handling claims and underwriting at scale — at carriers without the governance documentation regulators are now requiring. This episode dissects what the NAIC and state requirements actually obligate, why a vendor compliance sheet doesn't satisfy them, and how to build the algorithmic audit program before your regulator builds it for you.
The SR 11-7 Blind Spot: What Banks Discover When AI Hits Model Risk Management
Banks are deploying AI in credit scoring, fraud detection, compliance monitoring, and customer operations — but SR 11-7, the model risk framework regulators use, was written in 2011 for statistical models, not LLM APIs. This episode dissects the validation gap every bank running third-party frontier AI in production is carrying, what happens when the model version changes without a governance event, and the examination question your team should be able to answer before the next OCC or Fed visit.
The Shop Floor AI Bet: What Siemens' Industrial Copilot at BMW Means for Every Manufacturing CIO
Siemens Industrial Copilot is live at BMW plants — reading PLC logs and generating real-time maintenance recommendations. Rockwell, Honeywell, and ABB are building the same capability into their next service contracts. This episode dissects the data portability gap every manufacturer with an active OEM service contract should audit now, the EHS governance process most enterprises haven't defined for AI-generated maintenance steps, and what the shadow AI your maintenance engineers are already running tells you about your actual readiness.
The Harvey Partner: What Law Firms Aren't Telling Clients About AI in Legal Review
Harvey AI is deployed at 200+ law firms including A&O Shearman, Paul Weiss, and Cleary Gottlieb — drafting memos, reviewing contracts, running due diligence. Most clients haven't been told. ABA Formal Opinion 512 makes the disclosure decision deliberate but not automatic. This episode dissects what the 'Harvey partner' model means for client privilege, how firms are structuring AI use to stay below disclosure thresholds, and the governance questions your legal department and outside counsel relationships need to address before the next engagement letter is signed.
The AI Clinical Note Your Physician Didn't Write — and Signed Anyway
Microsoft Dragon Copilot, Abridge, and Ambience Healthcare are generating clinical notes across thousands of health systems — notes that physicians review and sign. Most enterprise deployments haven't updated their consent workflow, liability framework, or specialty-specific accuracy requirements to reflect what they've actually deployed. This episode dissects the signature-as-approval pattern that has replaced documentation review, why aggregate accuracy metrics hide the specialty gaps that drive liability events, and exactly how to build the governance workflow your health system hasn't built yet.
The AI Agent Security Audit You Haven't Done
Enterprise AI agents with tool use — web browsing, email, database access, code execution — are in production across organizations that reviewed them as chat models. This episode dissects why prompt injection is the primary unsolved threat, how tool permission creep happens silently across development cycles, and what a structured AI agent security program actually looks like to build.
The Open-Source Sovereign AI Decision
Cohere released Command A+ — a 218B sparse MoE model under full Apache 2.0 license — on two H100s, with benchmark parity to managed frontier APIs. This episode dissects why paying per-token for AI inference is now a choice your organization is making, not a technical constraint, and exactly how to build the 90-day parallel pilot that lets you decide from evidence rather than vendor positioning.
Salesforce Agentforce: The $2-Per-Conversation Bet
Salesforce Agentforce reached GA in October 2024 with a headline number — $2 per conversation — that makes autonomous contact center AI look like an obvious financial win. This episode dissects why that number is meaningless without your actual deflection rate, query mix, and escalation cost profile, and exactly what to put in place before your contact center commits at scale.
The EY-Microsoft Alliance: When Your Auditor Becomes Your AI Vendor
On May 21, 2026, EY and Microsoft announced a $1B+ joint initiative. EY Canvas — the platform running 160,000 audit engagements — now runs on Azure AI. EY is simultaneously your auditor and a $1B+ Microsoft commercial partner. This episode dissects what that means for your governance framework, why standard conflict disclosure doesn't close the gap, and what your audit committee needs to know before the next board meeting.
Microsoft Copilot for M365: The Adoption Numbers Your Vendor Doesn't Want You to Read
Microsoft defines a Copilot 'active user' as one interaction in 28 days — a metric no enterprise would accept for any other software investment. Independent research shows real weekly utilization at 30–50% of license count. With renewal windows now arriving, this episode breaks down the measurement gap, the auto-renewal architecture that compresses your window to act, and the 60-day independent sprint that changes the negotiation entirely.
The EU AI Act Compliance Deadline: What Your Enterprise Is Not Doing
August 2, 2026 is ten weeks away. Most enterprises haven't completed their Annex III inventory, haven't assigned human oversight owners, and have no 72-hour incident reporting workflow. This episode dissects the deployer gap — the obligations your vendor's compliance documentation cannot satisfy — and the four-week sprint your team needs to run right now.
The OpenAI Deployment Company: What Your Consulting Firm Didn't Disclose
OpenAI just raised $14B to build a deployment company — with Bain and McKinsey as founding partners. This episode breaks down what that structure actually means for your AI procurement decisions, how consulting conflicts operate below formal disclosure thresholds, and the four governance moves you need before your next vendor conversation.
Klarna's AI Customer Service Reversal
March 2024: Klarna announced its AI was handling two-thirds of all customer service. Late 2024: they were quietly re-hiring humans. This episode dissects what broke, why NPS failed as a signal, and exactly how to build tiered routing before it happens to you.
The Ad Machine: What Enterprise Marketing Teams Haven't Governed When AI Is Generating Brand Creative at Scale
Adobe Firefly has generated 9 billion+ images since launch. Meta Advantage+ AI creates autonomous ad creative variations for 4M+ advertisers. Google Performance Max gives AI simultaneous control over bidding, audience, and creative. AI-generated creative may lack copyright protection under current USPTO guidance, platform agreements may allow training on your brand assets, EU AI Act Article 5 prohibitions on manipulative techniques apply to emotion-optimization tools, and FTC enforcement on AI-generated advertising claims is an active posture. This episode dissects the four governance gaps enterprise CMOs have not closed and the governance framework needed before the next campaign cycle.
The Personal AI Subscription Problem: What Your Consultants, Lawyers, and Auditors Are Doing With Your Confidential Data
73% of knowledge workers use AI tools their employers have not sanctioned. Your external consultants, lawyers, and auditors are using personal ChatGPT Plus, Claude Pro, and Copilot subscriptions on your confidential files — and consumer AI accounts are not covered by firm-level data processing agreements. OpenAI's consumer terms allow training on conversations unless users actively opt out; most do not. This episode dissects why standard NDA language doesn't close this gap, what GDPR Article 28 requires when personal AI subscriptions touch EU personal data, and the three contract postures every enterprise must add before the next professional services engagement.
New episodes every Tuesday.
Subscribe to get each episode in your inbox at 7 AM CT — with the full memo and 10-slide PDF deck attached.
Free during beta. No credit card.
Issue #22 is live · Free during beta