Lesson 6 of 6
The 24-month roadmap
16 min
Listen to this lesson
The overview course (sap-tst-401) sketched the TCoE's first 24 months: foundation, stabilize, automate, scale, optimize. Contoso has lived through the first year of that. This lesson builds the next 24-month roadmap from what the TCoE now knows: the maturity assessment, the KPI gaps in its scorecard, the AI shortlist and guardrails, the skills plan and the calendar of change. It is a plan the quality council can fund, track and stop if needed.
Inputs
| Input | What it says | Roadmap consequence |
|---|---|---|
| Maturity assessment (lesson 1) | Test Environment partially achieved; non-functional testing not achieved; lifecycle integration and training partial | Fix Test Environment first; plan non-functional and lifecycle work in year 2 |
| Scorecard (sap-tcoe-406) | Escapes 14 a quarter (target under 8); DDP 94%; coverage 80%; key-user hours 64 a year | Procure-to-pay coverage and data; escape root causes; close-week protection |
| AI shortlist (lesson 2) | Adopt failure clustering; pilot test case drafts and data identification; reject autonomous scope | Two pilots with golden sets, then scale what passes |
| Calendar | Two feature pack stack updates a year; Wave 2 for company code 1910; Solution Manager 7.2 mainstream maintenance ends in 2027; TMMi 1.3 assessments end after August 1, 2027 | Land capabilities before stacks; formal assessment on TMMi v2.0 |
| Funding (sap-tcoe-401) | Investment released in three tranches by milestone | Tranche 3 tied to the month-12 checkpoint |
The inputs to Contoso's next roadmap and what each changes.
Four phases
| Phase | Months | Main deliverables |
|---|---|---|
| Fix the foundation | 0 to 6 | Data recipes for all four value streams; masking automated in the runbook; procure-to-pay automation and data (tranche 2); AI use policy and guardrails; AI literacy for all |
| Pilot and integrate | 6 to 12 | Two AI pilots measured on golden sets; TCoE engaged from Explore for every project above the threshold; KPI history complete for Test Measurement |
| Scale what works | 12 to 18 | Passed pilots extended to all value streams; formal TMMi v2.0 level 2 assessment; outcome component in the partner contract starts |
| Reach for level 3 | 18 to 24 | Non-functional testing standard for performance and roles; role-based training program; AI quality specialists evaluate the first Joule feature on the release calendar |
The four phases and their deliverables.
Checkpoints with teeth
A roadmap without checkpoints is a wish list. Contoso's has four, one per phase, each reviewed by the quality council with a decision: continue, adjust or stop. Each checkpoint names KPIs and their thresholds, so the decision is about evidence, not enthusiasm.
| Checkpoint | Evidence required | Decision |
|---|---|---|
| Month 6 | Test Environment process area satisfied in an informal reassessment; data-caused failures under 10%; guardrails approved; AI literacy completed | Release the pilots; continue |
| Month 12 | Weighted escapes under 10 a quarter; procure-to-pay coverage 100% of critical processes; each pilot at or above its golden-set acceptance criteria | Release tranche 3; scale passed pilots; stop failed ones |
| Month 18 | Formal TMMi v2.0 level 2 achieved; enhancement testing at or below 22 hours per change; key-user hours in close weeks at zero | Start level 3 work; reset targets |
| Month 24 | Weighted escapes under 8 a quarter; DDP above 95%; level 3 process areas at least largely achieved in an informal assessment | Next roadmap; consider a formal level 3 assessment |
Roadmap checkpoints. The KPIs come from the scorecard in sap-tcoe-406 and the levers in sap-tcoe-401.
- Did the phase deliver its evidence?Yes → continue to the next phase · No → next question
- Is the gap due to a dependency outside the TCoE?Yes → escalate the dependency and adjust dates · No → next question
- Is the item still worth its cost on current evidence?Yes → adjust scope or approach and recheck in three months · No → stop it and record why
Where an AI-capable automation platform fits
AI in the automation layer sits mostly in phase 2 and 3: failure clustering and flaky-test flags from the start, self-healing with review in every regression, AI-generated data variations under the same review rules as test cases. These features are judged like everything else on the roadmap: a pilot, a golden set or a measured baseline, and a KPI that should move, here regression cycle time, automation stability and enhancement testing hours.
What would you do?
At the month-12 checkpoint, the AI test case drafting pilot scores 55% correct or correctable on the golden set against a 70% threshold, but the analysts like it. What should the quality council decide?
- Build the roadmap from the assessment, the scorecard, the AI shortlist, the skills plan, the calendar and the funding model.
- Phase it: fix the foundation, pilot and integrate, scale what works, reach for level 3.
- Give each checkpoint KPIs and thresholds, and decide continue, adjust or stop on evidence.
- Judge AI in automation like any roadmap item: a pilot, a baseline and a KPI that should move.
Check your understanding · not graded
Lesson check
1. Why does Contoso's roadmap fix Test Environment before scaling AI pilots?
2. What makes a roadmap checkpoint useful to the quality council?
3. Why is the formal TMMi assessment planned on v2.0 at month 15?
4. At month 12, the AI drafting pilot scores 55% against a 70% threshold, but analysts like it. What does the checkpoint rule allow?