Frontier Certified — Trial Program Brief
Prepared by: Frontier Engineering · For: Microsoft · Status: Draft for discussion · Date: 3 November 2026
1. Purpose
This brief proposes a trial of the Frontier Certified program with one consulting firm. It sets out what the trial would test and which decisions we need from Microsoft to start. The trial would prove the format, the learning outcomes and the certificate process before scaling toward 10,000 learners.
2. Program in one paragraph
Frontier Certified is a Microsoft-branded, in-person program that equips consulting firms to deliver frontier AI engineering on GitHub Copilot, Microsoft Foundry and Microsoft Fabric, using Microsoft’s open-source HVE (Hypervelocity Engineering) Core as the working method. Frontier Engineering builds the content and teaches the workshops. Learners earn a Microsoft-issued certificate by passing a hands-on practical, not an exam. Firms gain people who can lead AI delivery for their clients; Microsoft gains a partner ecosystem that drives adoption of its platform.
3. Trial objectives
| Objective | How we’ll measure it |
|---|---|
| Learners can apply HVE on the Microsoft stack | ≥ 80% pass the graded practical |
| The format works for a mixed audience | NPS ≥ 50 overall and for each role path |
| The environment is plug-and-play | ≥ 90% working before day 1; < 30 min lost to setup per learner |
| Skills improve measurably | Pre/post hands-on task: time to working result and quality score |
| Certificate process runs end to end | Results delivered to Microsoft’s issuing process within 10 business days |
| The model can scale | Capacity and cost per learner validated against the 10,000 target |
4. Proposed trial design
Participants. One consulting firm, one cohort of 24–30 people, nominated by the firm. Suggested mix: about two-thirds technical (software, data, architects) and one-third delivery (project managers, engagement leads).
Format. 3 days in person at a Microsoft or firm office. A lead instructor and one assistant from Frontier. About 80% of the time is hands-on.
Teams. Learners work in mixed teams of 4–5 that mirror a real engagement team: builders, a data engineer, an architect and a delivery lead.
Sample project. A fictional insurer wants an AI agent to triage incoming claims. Teams use Fabric for the claims data, Foundry for the agent, and GitHub Copilot with HVE to build, review and hand it over. All data is synthetic.
Environment. One click into a browser-based development environment with Copilot, HVE Core, Foundry and Fabric already configured. Learners don’t install anything on their laptops. A setup check runs the week before.
Agenda (draft)
| Technical path | Delivery path | Together | |
|---|---|---|---|
| Day 1 — Method | HVE Research → Plan → Implement → Review cycle on the starter repo with Copilot | HVE project lifecycle: scoping, decision points, review gates | Kickoff, agentic engineering principles, team formation |
| Day 2 — Build | Fabric data foundation; Foundry agent built through HVE | Estimating and planning with agents, governance, risk, client adoption | Midday and end-of-day team demos |
| Day 3 — Ship & certify | Evaluations, security, responsible AI | Handover plan, client communication | Graded practical (half day) and final demo |
How the certificate is earned
- Individual assessment within team work. Each learner submits their own work: HVE plans, reviews and code (technical path), or engagement plan and governance decisions (delivery path). A short walkthrough with an instructor confirms it.
- Rubric covers five areas: method, context engineering, product use, quality and safety, delivery readiness.
- Frontier sends the results to Microsoft, which issues the certificate.
Optional: deployment phase pilot
A subset of 2–3 teams applies the method to a real or internal project for 8 weeks. They join fortnightly check-ins and submit a final evidence pack. This tests the deployment phase on a small scale before it becomes part of the program.
5. Decisions we’re asking Microsoft to make
For each decision, we give the options and our recommendation.
A. Workshop length
| Option | Instructor need for 10,000 learners (over 2 years) | Note |
|---|---|---|
| 2 days, 40 per cohort | ~5 full-time instructors | Lowest cost; little depth for technical roles |
| 3 days, 25 per cohort ✅ | ~8 full-time instructors | Balanced; recommended for the trial |
| 4 days, 20 per cohort | ~13–14 full-time instructors | Deepest; hardest to scale |
| Role-based (3 days technical, 1.5 days delivery) | ~6 full-time instructors | Strong option once the trial has validated the role paths |
B. Deployment phase
- None. The certificate is earned in the workshop.
- Self-reported evidence pack, with light review.
- ✅ Team-based deployment with a reviewed evidence pack. It’s credible and keeps reviews manageable: about 2,500 team reviews instead of 10,000 individual ones.
C. What firms may claim
- ✅ Individual certificates only. Firms say “N Frontier Certified professionals.” This is simple and hard to misuse, so we recommend it for launch.
- A firm designation once a firm reaches a set number of certified people and completed deployments.
- Firm tiers based on certified headcount, deployments and client references.
D. Delivery at scale
- Frontier instructors deliver every cohort.
- ✅ Frontier delivers the trial and early cohorts, then trains and licenses additional instructors under Frontier quality assurance. This is the only way we see to reach 10,000 learners on a reasonable timeline.
6. What we need from Microsoft
- Program owner and content approval contact, with an agreed review turnaround
- Certificate issuance process and the data format it needs from us
- Brand kit and usage guidelines for workshop materials
- Lab resources: Azure, Fabric and GitHub Copilot tenants or credits for learners
- Funding model for the trial and rollout
- Marketing terms, including how Frontier Engineering is credited
- Introduction to the trial firm, and agreement on nomination criteria
- Decisions A–D above
7. Timeline (weeks from approval)
| Week | Milestone |
|---|---|
| 0 | Brief approved; trial firm confirmed |
| 2 | Curriculum outline and assessment rubric to Microsoft for review |
| 5 | Environment built and tested; content complete |
| 6 | Internal dry run; fixes |
| 7 | Learner setup checks |
| 8 | Trial workshop |
| 10 | Results report and certificates issued |
| 18 | Deployment phase pilot readout (if included) |
8. Key risks
| Risk | Mitigation |
|---|---|
| HVE Core and the products change quickly | Pin versions per cohort; recheck before each delivery |
| Setup problems eat workshop time | Browser-based environment, pre-checks, backup accounts |
| Firm restrictions (data, models, network) | Synthetic data; confirm firm constraints at nomination |
| Mixed skill levels in one room | Placement pre-assessment, role paths, mixed teams, stretch labs |
| Cloud costs per learner | Spending caps, auto-pause, shared capacity where possible |
| Instructor capacity at scale | Decision D: licensed instructors with Frontier quality assurance |
Internal — remove before sending
Sources and open items behind this brief:
- Charter: Whitelabel Training Program · Open items: Questions (A ↔ ^q-005, B ↔ ^q-018, C ↔ ^q-017, D ↔ ^q-024)
- Capacity figures: Scale and Capacity Model · Assessment: Certification and Assessment Design
- Format & teams: Learner Personas · HVE Teaching Plan · Environment: Environment
- Benchmark used for B and D: Claude Frontier Academy
- Numbers that are our assumptions, not agreed: cohort size, 3 days, 80/50/90% targets, 8-week timeline, review turnaround