Pryme Intelligence Academy
Pryme Intelligence Academy teaches people to run governed AI agents on the real product. Four tracks, each the key to the next. Lessons, labs and tests are free; lessons can be read or listened to. Certificates are issued through VerifyDoc.ai and can be checked by anyone.
Foundations: Understand governed agents
Anyone: owners, executives, buyers, students. What a governed agent is, how the decision chain works, how depth is earned, and what a business pays for. Six modules, self-paced, with a free workspace to try every idea in.
Time: About 4 hours. Credential: Foundations badge, valid No expiry, stamped with the curriculum version.
- F1 Why agents must be governed (40 min): Intelligence was never the bottleneck, permission was; What an agent is and is not; Why ungoverned agents never get deployed. Lab: Quiz: spot the ungoverned action in ten examples.
- F2 The decision chain (50 min): The five checks: role scope, certificate, sources, your limits, guardrails; The lowest check wins; Allowed, held, refused, withheld; Why the chain does not ask the model. Lab: Walk through three decisions in the sandbox.
- F3 Depth and certification (50 min): Depth 1 (read) to depth 5; Depth is earned in Attest, never bought; The seven gates; 90-day certificates and what a lapse does. Lab: Open Attest and read what a certificate will need.
- F4 The workspace tour (40 min): Workspace, Build and Govern sections; The six Overview numbers: Decisions taken, Refusals, Waiting on you, Depth in use, Certified, Ledger height. Lab: Tour a free workspace.
- F5 What you pay for (30 min): A governed action; What is never charged; Draft-only at the cap; The plans. Lab: Estimate the actions in a small business’s month.
- F6 Ownership, data and trust (30 min): Context Protocol: the memory belongs to the customer; The tamper-evident audit trail; Refusals are logged; Data protection at a glance. Lab: Quiz.
Operator: Run a live workspace
The people who run a live workspace every day: reviewers, team leads, practice and operations managers. Reading decisions, clearing the review queue, teaching agents, keeping knowledge current and handling the day something goes wrong. Eight modules and a written exam, with a recommended two-week practicum on real decisions or a practice set.
Time: About 14 hours, plus a two-week practicum. Credential: Certified Pryme Operator, valid 12 months. Assessment: 40 questions, 60 minutes, pass mark 80%, online. Two weeks of real review decisions: 95% correct and no missed red line. Recommended, not required for the certificate today: customers do it in their own shadow period; public candidates on a practice set the Academy provides. The certificate says exactly what was checked.
- O1 How a governed agent decides (1.5 h): Reading a decision record; Which check decided it; Allowed vs Held vs Withheld in practice; Common misreadings. Lab: Sort 20 sample actions and name the deciding check.
- O2 Human Review (2 h): The review queue; Approve or decline, and say why; Response-time targets; Two-person approval for money (a customer rule, coming); What happens to your answer. Lab: Work through 25 sample held items.
- O3 Teaching the agent (2 h): Corrections and why they are banked; House rules and blind preference; Why your rules survive a model change; Filing a real case. Lab: Bank 10 corrections and promote a house rule after a blind duel.
- O4 Reading the workspace (1.5 h): Overview numbers; Live decisions, Decision flow, Certification map; Audit Trail; Where actions went by desk. Lab: Trace three decisions from Overview to the ledger.
- O5 Keeping knowledge current (2 h): Kinds of knowledge and who may read them; Updating a price or policy the same day; Ask as an agent in the Knowledge Base: preview what an agent would cite, unbilled. Lab: Change a price, then confirm what the agent now cites.
- O6 When something goes wrong (2 h): Stopping an agent acting: switch-off today, draft-only pause coming; The escalation directory; Urgent cases that cannot wait their turn; Complaints and disputes; Data-subject requests. Lab: Tabletop drill with four incidents.
- O7 Usage, limits and billing (1 h): Reading Plan & usage; Pace through the month; Top-ups; What draft-only at the cap means for your team. Lab: Forecast next month’s actions.
- O8 Running the rhythm (1 h): Daily, weekly and monthly routines; Preparing for 90-day recertification; Cover when reviewers are away. Lab: Write your team’s review rota.
Builder: Design and certify agents
Whoever designs, qualifies and certifies agents: customer power users, implementation consultants, partner staff. From a plain-language job to a certified agent: roles, knowledge, guardrails, scenario suites, Attest, adversarial testing, integrations and the road from shadow to production. Ends with a capstone agent certified in Attest.
Time: About 24 hours, plus a capstone. Credential: Certified Pryme Builder, valid 12 months. Assessment: 50 questions, 90 minutes, pass mark 75%, online. A capstone agent certified in Attest to depth 2 or higher. Attest seals the certificate with a hash of every marked case and the suite version, and signs it, so it cannot be faked. A Pryme assessor reviews the capstone.
- B1 Designing a role (2 h): From a plain-language job description to a role; Role scope and connector classes; What a role may never do; Pre-trained roles and role requests. Lab: Write a role for a front-desk agent.
- B2 Knowledge architecture (3 h): What a source is and how it gets in: connect, upload, read from an address, write; Which agents may read what; Owners and version dates; Keeping sources current. Lab: Build a knowledge base with four sources and read permissions.
- B3 Live context and Context Protocol (2 h): What an agent knows now vs what it will remember; Provenance; The customer owns the memory. Lab: Bring in a live source and trace a citation.
- B4 Guardrails and ceilings (3 h): Rules on content; Working limits the customer sets; The platform floor; Writing red lines that can be tested. Lab: Write five red lines and prove each with a case.
- B5 Building a scenario suite (3 h): Cases that matter; Gold benchmarks; Coverage and authoritative runs; Filing real cases from a mailbox. Lab: Build a 50-case suite for one role.
- B6 Attest in depth (3 h): The seven gates and their floors; Readiness and the binding gate; Evidence conditions; Replay; Signing a certificate. Lab: Raise a failing gate above its floor.
- B7 Adversarial testing (2 h): Prompt injection, impersonation, pressure; The adversarial room; Blind duels. Lab: Break your own agent three ways, then fix it.
- B8 Integrations (2 h): Read before write; One write action at a time; Channel set-up; Testing below depth 3. Lab: Connect one channel and one system while the agent works below depth 3.
- B9 From shadow to production (2 h): Designing a shadow; Go/no-go criteria; Raising depth one step at a time; Recertification. Lab: Write a go/no-go pack.
- B10 Capstone (2–4 h of work, over at least seven days): Build, qualify and certify one agent to depth 2 or higher in a free workspace. Lab: Submit the Attest certificate, with its suite version and hash.
Architect: Lead regulated deployments
People who lead regulated, multi-agent deployments: partner leads, enterprise AI and risk owners, solution architects. Turning obligations into checks, written policy packs, data protection engineering, governance through the API, multi-agent operations, audit, the deployment method and sizing. Ends with a written design and a board review.
Time: About 30 hours, plus a capstone and board review. Credential: Certified Pryme Architect, valid 24 months. Assessment: A written capstone: a full deployment design for a case Pryme provides. A 60-minute board review with a Pryme assessor. Pass needs a sound design and correct answers on governance and data protection.
- A1 Governance for regulated sectors (4 h): Mapping an obligation to a check in the decision chain; Banking, health and public-sector patterns; Design depth by sector. Lab: Map ten obligations for one sector.
- A2 Policy packs and escalation design (4 h): Writing a sector policy pack; Escalation matrices and response targets; Two-person rules. Lab: Write a policy pack for a clinic or a bank.
- A3 Data protection engineering (4 h): Personal and sensitive data; Concepts across regimes: GDPR and UK GDPR, Nigeria’s Data Protection Act and others; Residency, isolation, sub-processors; What goes in a data-processing agreement. Lab: Draft the data section of a deployment pack.
- A4 Governance through the API (4 h): Sandbox and API keys; Calling the decision chain from your own code (/v1/evaluate); Logging back to the ledger. Lab: Govern an action from a small script.
- A5 Multi-agent operations (3 h): Desks and hand-offs between agents; Command Centre; Avoiding conflicting authority. Lab: Design a four-agent desk.
- A6 Audit and assurance (3 h): Evidence for auditors and regulators; Reading and exporting the ledger; What tamper-evident means. Lab: Prepare an evidence pack for one month.
- A7 The deployment method (4 h): Discovery, training, shadow, go-live, hypercare, handover; Gates and sign-offs; Running a customer’s operator training. Lab: Plan an eight-week deployment.
- A8 Scoping and sizing (4 h): Estimating governed actions; Sizing a deployment; Pricing an Enterprise agreement. Lab: Size and price a sample deployment.
Teams: a free readiness assessment for up to 25 people shows scores by concept, the gaps and the first agent to start with.