Take one AI workflow from Jira backlog to a tested staging demo in four weeks.
4–29 January 2027 · Remote · 20 live one-hour sessions · £849 per person
Live sessions run at a UK time agreed with your cohort team before the start. Expect about 3–5 hours a week of practice on top of the live sessions.
Experienced product managers, product owners, Scrum Masters, delivery leads, Agile coaches, developers and technical leads.
You will need professional-level English and experience working in a team or project setting. A readiness call confirms your fit. The January 2027 cohort runs remotely from 4 to 29 January 2027.
Evidence you can show
A working increment
A live demo of an AI agent or AI-enabled feature, built on a real team project.
A linked evidence trail
A Jira backlog, GitHub review and test evidence, and a staging record that connect back to your goals.
A release decision
A reasoned call to release, hold or run a limited pilot, with the remaining risks stated.
A 30-day improvement plan
A backlog with owners that carries the work on after the programme.
Tools you will use:
From business problem to release decision
Each week moves your team along the journey: plan in week 1, build in week 2, validate and deploy in week 3, then review the release and present in week 4.
One team, one project, 20 sessions
Spend 30 minutes on teaching or a live walkthrough, then 30 minutes on guided practice, discussion or a team review. Complete the practical task before the next session. Friday tasks are due on Monday, except the final submission on Day 20. Everyone contributes to the same project:
Product roles
Lead discovery, prioritisation and acceptance.
Delivery roles
Facilitate planning, track flow and help resolve blockers.
Engineering roles
Build, test and deploy the increment.
Four weeks, 20 sessions
Open any day to see the teaching, the starter questions for the discussion, and the practical task.
Plan an AI delivery project
4 to 8 January 2027Agree what the team will deliver, how it will measure success, and which risks and dependencies could affect delivery.
By the end of the week: A solution brief, prioritised Jira backlog, PI objectives, dependency map, ROAM risk log and initial architecture.
Day 1Mon 4 JanDiscover the business problem
Teaching
Map the current workflow, identify users and bottlenecks, and establish a baseline. Compare agent, conventional automation and manual options.
Starter questions
What measurable change would justify building this? What is the smallest useful workflow?
Practical task
Write a one-page brief in Confluence with the user, baseline, target outcome and one workflow to improve. Use Penta to structure discovery where available.
Day 2Tue 5 JanDefine the AI agent and its boundaries
Teaching
Specify inputs, outputs, data sources, tools, actions and human approval points. Define how the agent should handle uncertainty or unavailable data.
Starter questions
Which actions require a person’s approval? When should the agent decline or hand off?
Practical task
Write an agent specification with allowed actions, data limits and three examples of acceptable behaviour.
Day 3Wed 6 JanTurn the brief into a Jira backlog
Teaching
Break an epic into features and small user stories. Write testable acceptance criteria and agree how the team will use its Kanban board and limit work in progress.
Starter questions
Which story gives the earliest evidence of value? What must be true before it moves to Done?
Practical task
Create an epic, one feature and three to five linked stories. Add acceptance criteria and prioritise the first deliverable.
Day 4Thu 7 JanRun a PI Planning simulation
Teaching
Practise SAFe Planning Interval (PI) Planning. Agree objectives and team capacity, map dependencies, and classify risks as Resolved, Owned, Accepted or Mitigated (ROAM).
Starter questions
What should move if capacity drops? Who owns the dependency that could block delivery?
Practical task
Record two measurable PI objectives, a dependency map and a ROAM risk log. Assign an owner to every unresolved dependency.
Day 5Fri 8 JanReview architecture and delivery readiness
Teaching
Sketch the agent, integrations and trust boundaries. Agree repository ownership, review rules, environments and the evidence needed for release.
Starter questions
Which integration could the team mock first? Where could sensitive data leave the approved environment?
Practical task
Publish a simple architecture diagram and one decision record in Confluence. Link the highest-risk assumption to a Jira story.
Build the first working increment
11 to 15 January 2027Use Claude Code in a supervised engineering workflow and demonstrate one end-to-end agent task using approved test data.
By the end of the week: A working prototype, a reviewed pull request, a basic evaluation set and documented permissions.
Day 6Mon 11 JanSet up GitHub and Claude Code
Teaching
Connect the approved repository and agree how to use branches, pull requests, required reviews and Claude Code permissions. Keep credentials out of prompts and source control.
Starter questions
Which commands can Claude Code run automatically? Which require explicit approval?
Practical task
Create a branch linked to a Jira story, document the project’s AI usage rules and open a draft pull request.
Day 7Tue 12 JanBuild one end-to-end agent workflow
Teaching
Implement one workflow with structured inputs and outputs. Start with a mock or read-only tool and validate the agent’s response before any downstream action.
Starter questions
What happens if the tool returns invalid data? When should the workflow stop?
Practical task
Build one working path with approved sample data and demonstrate both a successful run and a safe fallback.
Day 8Wed 13 JanAdd permissions and human approval
Teaching
Apply least-privilege access, separate secrets from code and require approval before consequential writes. Capture the approval and action in the audit trail.
Starter questions
What evidence should an approver see? What happens if approval is denied?
Practical task
Add one approval gate. Show that the workflow cannot perform the protected action when approval is denied or missing.
Day 9Thu 14 JanWrite tests for code and agent behaviour
Teaching
Use test-first development for predictable code paths and example-based evaluations for variable model outputs. Cover malformed inputs, unsupported requests and uncertain answers.
Starter questions
What can you assert exactly? Where does an output need a scoring rubric?
Practical task
Add a deterministic test and build a small evaluation set with normal, edge and failure cases. Record the baseline results.
Day 10Fri 15 JanReview the increment and improve flow
Teaching
Review AI-generated changes for correctness, maintainability and acceptance criteria. Demo the increment, inspect blocked work and agree a change to improve flow.
Starter questions
Which part of the pull request needs a closer look? What held the team up this week?
Practical task
Update the pull request with review and test evidence. Record one retrospective action in Jira with an owner and a due date.
Validate and deploy to staging
18 to 22 January 2027Build confidence in the release with repeatable checks and a staged deployment. Practise diagnosing a failure and recovering safely.
By the end of the week: A release candidate with test results, CI/CD evidence, a staging record and a tested recovery procedure.
Day 11Mon 18 JanRun integration and end-to-end tests
Teaching
Test tool contracts and the full user workflow. Cover timeouts, unavailable dependencies, duplicate requests and the path from agent output to human approval.
Starter questions
What happens if the same request arrives twice? Which failures should trigger a retry?
Practical task
Run an end-to-end scenario in a test environment. Record one failure case and link the result to its Jira acceptance criteria.
Day 12Tue 19 JanTest security and misuse scenarios
Teaching
Review threat scenarios such as prompt injection, unauthorised tool use and sensitive data in outputs or logs. Test safeguards with approved test data.
Starter questions
Can untrusted content cause the agent to call a protected tool? What should the logs reveal after an incident?
Practical task
Run three misuse scenarios and record the expected and observed behaviour. Assign owners to any unresolved findings.
Day 13Wed 20 JanAutomate quality checks with CI/CD
Teaching
Configure GitHub Actions to run tests and security checks on pull requests. Review the build artefact, approval rules and the link between a commit and a deployment.
Starter questions
Which failure must block a merge? What evidence should stay with the build?
Practical task
Run the pipeline on a pull request. Demonstrate one failing check and save the link to the passing run after correction.
Day 14Thu 21 JanDeploy the release candidate to cloud staging
Teaching
Deploy from a reviewed build to a cloud staging environment. Run smoke tests and verify configuration, access controls and rollback readiness.
Starter questions
Who can approve the deployment? How would you restore the previous version?
Practical task
Deploy to an approved staging environment or cohort sandbox, record the build and smoke-test result, and rehearse rollback.
Day 15Fri 22 JanMonitor the agent and rehearse recovery
Teaching
Review logs and metrics for failed runs, latency, cost and approval outcomes. Walk through an incident and decide when to pause or disable the workflow.
Starter questions
What signals show the agent is causing trouble? Who acts on them?
Practical task
Create a short runbook with one alert threshold, an owner, a recovery step and evidence from the rehearsal.
Review the release and demonstrate the work
25 to 29 January 2027Show what the team delivered, explain the remaining risks and decide whether to release, hold or run a limited pilot.
By the end of the week: A final demo, evidence pack, release decision and a 30-day improvement plan.
Day 16Mon 25 JanMeasure delivery flow and product outcomes
Teaching
Use Jira and, where available, Magnesia to inspect cycle time, blocked work and progress against objectives. Compare the agent’s results with the Week 1 baseline.
Starter questions
What do the data tell you about user value? What is still too early to judge?
Practical task
Produce a one-page report with one outcome measure, one flow measure, the supporting evidence and one improvement to test.
Day 17Tue 26 JanRun a release and governance review
Teaching
Check access, data handling, test results, approvals and operational ownership. Revisit unresolved risks and decide whether to release, hold or run a limited pilot.
Starter questions
Which open risk would stop the release? Who owns it after handover?
Practical task
Record the release decision, rationale, sign-offs, open risks and the conditions for rolling back or disabling the agent.
Day 18Wed 27 JanPractise your SAFe certification pathway
Teaching
Work through role-based scenarios on facilitation and flow (SSM), value and backlogs (POPM), team delivery (SAFe for Teams), and test-first engineering (ASE).
Starter questions
Which project decision best illustrates your pathway? Where do you need more practice?
Practical task
Complete a short practice set for your pathway and use the official Scaled Agile study guide to plan your revision.
Day 19Thu 28 JanRehearse the demo and assemble the evidence
Teaching
Practise the story from business problem to release decision. Review the trace from PI objective to Jira story, pull request, test results and staging build.
Starter questions
Could a reviewer reproduce the evidence from your links? What will you show if the live demo fails?
Practical task
Assemble the evidence pack, record a backup demo and use peer feedback to sharpen the presentation.
Day 20Fri 29 JanPresent the capstone and hand over
Session format
Teams demo the agent, explain a key trade-off and show how they tested a safeguard. Each participant describes their contribution. Coaches give feedback on evidence and next steps.
Starter questions
What outcome changed? What would you improve before expanding the rollout?
Final submission
Submit the demo, evidence pack, contribution summary and a 30-day improvement backlog with owners.
One story, traced to a release decision
Sample, demo data. This is an illustrative example, not a real client project. It shows the evidence trail every team builds.
- Jira storyKIAN-142 As a support lead, I want new tickets triaged by an AI agent so urgent issues reach a person within 10 minutes.
- GitHub pull requestPR #37 Add triage agent and escalation rule. Reviewed by two team members; 3 requested changes resolved.
- Test result24 of 24 automated tests passed, including 6 cases for escalation and 2 for refusing unsafe requests.
- Staging buildBuild 2027.01.27-3 deployed to staging; demo run with 40 sample tickets.
- Release decisionLimited pilot: release to one support queue for two weeks. Open risk: unusual ticket wording. Owner: delivery lead.
Sponsor your team
Send a product, delivery and engineering group through Pro together, on one of your own workflows. Here is what you are buying.
Product roles
Lead discovery, prioritisation and acceptance.
Delivery roles
Facilitate planning, track flow and help resolve blockers.
Engineering roles
Build, test and deploy the increment.
What your team brings
- A named product, delivery and engineering lead (one person can cover more than one role in a small team).
- One real workflow or backlog item that AI could improve, agreed before the start.
- Access to a Jira project, a GitHub repository and a staging or test environment, or the sandbox option below.
- Time for each person to join the live sessions and complete about 3–5 hours of practice a week.
Sandbox option. No repository or staging environment ready? Sepenta can set up a sandbox with demo data, so the team practises the full backlog-to-staging path without touching production systems or confidential data. Ask us about the sandbox when you enquire.
What your team takes away
A tested staging demo
The AI workflow running in staging, shown live to your stakeholders.
Linked delivery evidence
A trail from Jira story to GitHub pull request, test result and staging build.
A documented release decision
A reasoned decision to release, hold or run a limited pilot, with remaining risks stated.
SAFe pathways and what the £849 covers
Pro connects your project work to four SAFe pathways. The £849 covers Sepenta coaching and the programme only. It does not include official SAFe training or exam access, and completing Pro does not award a SAFe credential. Scaled Agile sets eligibility and awards certification after a passing exam.
| Pathway | Typical fit | Training and exam access |
|---|---|---|
| SAFe Scrum Master (SSM) Facilitation and flow | Scrum Masters, delivery leads and Agile coaches | Not included in the £849 |
| SAFe Product Owner/Product Manager (POPM) Value and backlogs | Product owners and product managers | Not included in the £849 |
| SAFe for Teams (SAFe Practitioner, SP) Team delivery | Members of Agile teams | Not included in the £849 |
| SAFe Agile Software Engineering (ASE) Test-first engineering | Developers and technical leads | Not included in the £849 |
Additional fees and booking steps: book the official course and exam for your chosen pathway directly with Scaled Agile or an authorised training partner, and check their current fees and eligibility rules. We can talk you through the options on your readiness call.
Sepenta Pro is a Sepenta programme and is not run by Scaled Agile.
Format, price and scholarships
Format
Remote, live. One hour of live coaching on each teaching day at a UK time agreed with your cohort, plus about 3–5 hours of practice a week.
Price
£849 per person for the 4-week programme. This covers Sepenta coaching only; SAFe training and exams are separate. Team and organisational rates are available, and scholarships may apply to high-potential candidates, subject to availability each cohort.
Readiness call
A short call with the Sepenta team confirms you are ready for the pace of the programme before you enrol.
Start date
The next cohort starts on Monday 4 January 2027 and finishes on Friday 29 January 2027.
Not sure which level fits? Take the free Talent Challenge (about 15 minutes), or see the LaunchPad syllabus.
Agile and AI, taught by people who deliver it
Sepenta is an Agile and AI consultancy, training and apprenticeship company based in Cambridge, UK and Kuşadası, Türkiye, led by SPC 6 certified practitioners. We build AI-ready technology talent through real project delivery, and we help organisations adopt Agile and AI-enabled ways of working. Pro is the third step of the Sepenta pathway, after LaunchPad and FastTrack.
Apply for the January 2027 cohort
Sepenta Pro, 4–29 January 2027. A short readiness call confirms your fit. There is no commitment until you enrol.
Prefer to talk first? Book a 30-minute call or email info@sepenta.io. Sponsoring a team? Sponsor your team.
