Private and Verified Agentic AI for Leaders
James Gray
Berkeley AI Instructor | ex- CIO & CPO
Your agent works. Now make it safe to trust with sensitive data.
The workflows that would help your business most carry sensitive data: customer records, employee files, IP. Sending it to an AI service is a risk, so they stay manual, or run with no way to see what the AI did.
Here you practice safely: run agents on fictional data, see where it can leak, and inspect evidence of which model ran and what it did.
Then take one workflow you run today from useful to defensible:
✅ Know what data your agent handles, what it can do, and who is affected
✅ Find the obligations and threats you must answer for, and when to call in experts
✅ Pick the right protection and rate the evidence behind every claim
✅ Judge what hardware-backed proof does and does not establish
✅ Test a live agent workflow run against the standard you set
✅ Decide go, pilot, remediate, or no-go, with a 90-day plan
The first half builds judgment for any vendor and domain. The second half goes deep with Open Intelligence Network on confidential computing (hardware that protects data in use) and verified inference (proof of which model ran).
You'll leave with a Verified AI Deployment Record security, legal, and compliance can approve, and the judgment to assess the next one.
What you’ll learn
Move from building agents that work to deploying agents you can defend: private by design, secured by controls, verified by evidence.
Map what data your agent handles, what it can do, and who could be affected
Follow one sensitive document through the workflow to find every point where it is copied, stored, sent, or exposed.
Leave with a one-page map of the workflow's data, actions, and permissions, with an owner named for each risk.
Run a leader's eight-question checklist to find what must be investigated and when to bring in privacy, legal, or security.
Threat-model your agent: the ways it could be tricked, overstep its authority, misuse a tool, or leak what it remember.
Leave with a list of the ways it could go wrong, tested in a walkthrough, and a control that stops, catches, or recovers from each.
Choose controls that fit the risk, from limiting what data gets in to who can approve what gets out, and record what each leaves open.
Compare consumer AI, enterprise SaaS, model APIs, self-hosting, and confidential routes on one scale, from marketing claim to real proof.
Leave with an architecture decision and a record of every claim: did you see it, calculate it, get told it, or get promised it?
Read a plain-language attestation summary, the hardware's own report of where your workflow ran, and sort proven from promised.
Assess what confidential computing and OIN's three proofs (model, inference, privacy) establish, and what stays unresolved.
Leave with a Workflow Trust Profile: the written standard a run must meet before you accept its results.
Run fictional workloads through an ordinary route and a protected route, and inspect the evidence each one returns.
Tell a clean result from a stale, mismatched, missing, or delayed one, and decide what each means for approval.
Leave with documented test results that separate what you observed from what a provider told you.
Decide go, limited pilot, remediate, or no-go, with the remaining risk, the approval conditions, and an accountable owner on record.
Plan the first 90 days: what to monitor, when to re-check the evidence, what change forces a review, and how to switch it off.
Leave with a Verified AI Deployment Record you can take to security, legal, and compliance for approval.
Learn directly from James
James Gray
UC Berkeley AI Instructor | Former CIO/CPO | Trained 5,000+ Leaders
Who this course is for
Operational and Business Leader: Leaders whose highest-value workflows stall because the data is too sensitive for today's AI services.
Product and Technology Leader: Leaders responsible for the architecture, controls, and evidence behind agents that touch sensitive data.
Risk, Security, and Compliance Partner: Partners who review agentic workflows and need a shared method with the leaders who build them.
What's included
Lifetime access
Go back to course content and recordings whenever you need to.
Community of peers
Stay accountable and share insights with like-minded professionals.
Certificate of completion
Share your new skills with your employer or on LinkedIn.
Maven Guarantee
Your purchase is backed by the Maven Guarantee.
Frequently asked questions
Maven for Teams
Reimbursement
Get your company to pay
Everything L&D needs: email template, receipts, and certificate of completion.
Get reimbursedPrivate cohort
Run a cohort for your org
A dedicated cohort with a custom schedule and curriculum, tailored to your team.
Book a private cohort