Commercialisation and delivery

Executive summary

Stage 5 is where validated AI opportunities become funded products, signed contracts, and production services that survive real usage—not demo applause. You translate discovery into backlog and PRDs, defend economics with TCO and benefits realisation, engineer latency and unit cost at scale, design trust-aware UX, run programmes with clear gates, drive adoption through change management, shape presales and RFP responses, evaluate vendors honestly, and operate with SLOs. Completing this stage means you can take a solution from approved business case to steady-state operations with measurable adoption and controlled run rate.

Outcome

Primary outcome: You can take a solution from opportunity to production—with product discipline, commercial credibility, performance economics, human-centred design, delivery governance, adoption traction, and operational readiness.

Secondary outcomes:

• Produce Stage 5 core deliverables (PRD, business case, FinOps model, UX flows, programme plan, change plan, proposal pack, vendor scorecard, runbook) linked to prior discovery artefacts.
• Defend unit economics (cost per successful task, TCO, benefits register) in steering and procurement conversations.
• Run pilot-to-scale transitions without silent scope creep, run-rate shock, or adoption collapse after go-live.

Stage 5 does not require you to master every specialty alone. It requires you to orchestrate product, finance, engineering, UX, PMO, change, sales, and ops into one coherent delivery story.

Who Stage 5 is for

Profile	Typical gap	Stage 5 fix
Strong builder, weak product	Ships features; no PRD, metrics, or backlog hygiene	Topics 06, 23
Strong architect, weak commercial	Great diagrams; cannot price or defend ROI	Topics 07, 29, 30
Strong consultant, weak FinOps	Wins phase one; run rate explodes in year two	Topics 07, 22
Delivery PM without AI literacy	Gan charts exist; eval gates and model risk missing	Topics 25, 21, 22
Presales engineer without ops realism	Beautiful demos; no SLOs or support model	Topics 29, 32

Entry criteria

Start Stage 5 when all of the following are true:

• A validated use case exists from Stage 1–2 (problem statement, hypothesis supported, opportunity scored)—see AI Opportunity Discovery.
• Technical feasibility is credible (data access, integration path, eval approach)—see Stage 2–4 topics.
• A sponsor and budget line (or RFP path) is identifiable—even if not yet signed.
• You can name the primary user, success metric, and decision body for phase one.

Do not enter Stage 5 when discovery still produces tool-shaped requirements without outcome metrics, or when compliance has not seen a DPIA/risk summary for the intended pattern.

Exit criteria

Declare Stage 5 complete when all exit checks pass:

Artefacts (minimum set)

Artefact	Source topic	Pass test
PRD + MVP scope + backlog	06	In/out explicit; linked to discovery; adoption metrics defined
Business case / TCO model	07	Year-1 and year-3 TCO; sensitivity; cashable vs non-cashable benefits
FinOps / performance budget	22	Cost per successful task; p95 latency; optimisation backlog
UX flows + trust patterns	23	Override, uncertainty, escalation documented
Programme plan + RAID	25	Gates align with 8D; dependencies named
Change and adoption plan	26	Champions, training, comms, resistance map
Proposal / SOW or RFP response	29–30	Assumptions, exclusions, SLAs, exit clauses
Vendor scorecard (if buy)	31	Weighted criteria; not demo-led
Runbook + support model	32	On-call, incident severity, model rollback

Knowledge checks

• Answer ≥12/15 questions from topics 06, 07, and 22 without notes.
• Explain one negative case from Stage 5 using a sector-relevant example (run-rate shock, adoption failure, or procurement trap).
• Walk a 60-second executive story: problem, MVP, economics, risks, next gate.

Behavioural checks

• You have not committed to fixed-price delivery without sensitivity on token volume and human review load.
• A peer confirms artefacts are filed in your pattern library with sector tag and date.

Capability map

Stage 5 spans eleven numbered topics across product, economics, performance, experience, delivery, change, commercial, and operations. Use this map to see dependencies—not as a waterfall reading order only.

Core thread: Product (06) defines what and how we measure; Commercial (07) defines whether we fund; FinOps (22) defines whether we can afford to run; UX (23) defines whether humans trust it; Delivery (25) defines how we ship on time; Change (26) defines whether anyone uses it; Presales/RFP/Vendor (29–31) define how we sell and buy; Ops (32) defines how it stays alive.

Topics in this stage

Product and economics (build the right thing, fund it honestly)

#	Topic	You leave knowing
6	Product Management	Discovery → backlog, PRDs, prioritisation, MVP vs platform, adoption metrics
7	Commercial and Financial Modelling	Business cases, TCO, ROI, sensitivity, pricing models, benefits realisation

Pair with Business case and prioritisation and AI Product Management roadmap when moving from learning to client work.

Performance and experience (run it affordably; humans stay in control)

#	Topic	You leave knowing
22	Performance Engineering and AI FinOps	TTFT, caching, batching, cost per successful task, capacity planning
23	User Experience and Human Factors	Trust, uncertainty, override, accessibility, error recovery

Pair with Model FinOps roadmap and AI FinOps and Commercial Design.

Delivery and adoption (ship with gates; usage sticks)

#	Topic	You leave knowing
25	Delivery and Programme Management	Plans, RAID, dependencies, quality gates—see also Project Management (PMBOK® study map)
26	Change Management and Adoption	Stakeholders, training, resistance, benefits tracking

Pair with Adoption guide and Delivery guide.

Commercial motion and operations (win work, run production)

#	Topic	You leave knowing
29	Presales and Solution Shaping	Discovery in sales cycle, demos, POC boundaries, win themes
30	RFP, Procurement and Contracting	Compliance, SLAs, liability, IP, model change clauses
31	Vendor and Technology Evaluation	Scorecards, PoC discipline, exit strategy
32	Operations and Production Support	SLOs, incidents, drift, cost alerts, retirement

Recommended study paths

Path A: Product-led delivery (internal platform team)

1. Product Management → UX → FinOps
2. Delivery → Change
3. Ops

Focus: Backlog hygiene, adoption metrics, unit economics before feature breadth.

Path B: Consulting / SI delivery (client-funded programme)

1. Commercial → Presales
2. Product → Delivery
3. RFP → Vendor → FinOps

Focus: Business case defensibility, contract assumptions, vendor-neutral evaluation.

Path C: FinOps and scale rescue (post-pilot run-rate crisis)

1. Performance FinOps → Commercial
2. Product (scope reduction) → Ops

Focus: Cost per successful task, caching, routing, capacity—then re-baseline business case.

Stage 5 gates (align with 8D)

Map Stage 5 artefacts to 8D Framework gates:

Gate	Stage 5 evidence	Common failure
Fund / commit	Business case (07), pricing model, sensitivity	Pilot cost extrapolated linearly to production volume
Design sign-off	PRD (06), UX trust flows (23), FinOps budget (22)	No human override path; no latency budget
Build complete	Backlog done for MVP; eval green on golden set (21)	Feature creep; demo dataset only
Pilot exit	Adoption metrics (06, 26), cost per task (22), benefits baseline (07)	"Go live" without usage or run-rate sign-off
Operate	Runbook (32), alerts, chargeback/showback (22, 07)	No owner for model version drift

Real-world scenario: pilot to production (cross-topic)

A European insurer completes a 12-week claims summarisation pilot: 340 adjusters use the tool weekly; 62% active usage; average 4.2 minutes saved per complex claim on sample QA. Steering asks for EU-wide rollout to 2,400 adjusters.

Product (06): PRD v2 adds multilingual corpus, batch upload for legacy scans, and explicit "not for coverage decision" scope. MVP vs platform: phase one stays summarisation + citation; phase two structured extraction only after eval ≥90% on top 15 loss types.

Commercial (07): Year-1 TCO €2.8M (API €420K, human QA review €1.1M, integration €680K, change €380K, contingency €220K). Benefits: €3.6M productivity (cashable with union agreement); sensitivity: breaks if active usage <45% or review rate >35%.

FinOps (22): Pilot averaged €0.18/request; production projection €0.11 after semantic cache on repeat clauses, batching nightly bulk jobs, and routing simple FNOL to 7B regional model. p95 TTFT target 2.8s for interactive; batch SLA 4 hours.

Change (26): Union requires opt-in teams first; champion network in 6 regional hubs; training tied to QA scorecard not "AI magic."

Outcome: Phased rollout approved with FinOps dashboard and monthly benefits review—not big-bang on press release timeline.

Negative cases at stage level

Demo-to-contract gap. Presales demo uses curated data; RFP assumes general availability. Fix: POC acceptance criteria tied to client goldens (29, 31).

Run-rate surprise. Year-one business case used pilot token counts; production adds agent loops and logging. Fix: FinOps model with percentiles and step caps (22, 07).

Adoption cliff. Go-live without change plan; usage drops after week four. Fix: adoption metrics in PRD (06, 26).

UX trust failure. Users disable tool after one confident wrong answer. Fix: uncertainty UX and mandatory citations (23).

Vendor lock-in without exit. Contract silent on model change and data export. Fix: RFP clauses and scorecard exit criteria (30, 31).

Programme without gates. Agile velocity with no eval or compliance gate. Fix: 25 + 21 + 19 aligned milestones.

Related playbook content

• Business case and prioritisation — structure benefits, scenarios, approval narrative
• AI FinOps and Commercial Design — chargeback, pricing, run-rate governance
• Model FinOps roadmap — cost engineering career path
• Adoption guide — change, training, usage measurement
• Delivery guide — programme patterns for AI engagements
• 8D Framework — stage gates from Define through Operate
• How to use this Learning Map — study loop and artefact standards
• Trust and control — prerequisite governance and eval
• Leadership — next stage after delivery mastery

Output

Expected outputs for Stage 5 (portfolio level):

• Integrated delivery pack — PRD, business case, FinOps budget, programme plan, change plan linked by version ID
• MVP definition — in/out scope, acceptance criteria, adoption metrics with baselines
• Commercial pack — pricing recommendation, sensitivity, benefits register with owners
• Operational readiness — runbook, SLOs, support tiers, cost alerts
• Gate evidence folder — eval results, UX review, risk sign-offs per 8D gate

Stage 5 exit checklist

• I can explain the thread from backlog → business case → FinOps budget → ops runbook
• PRD and MVP scope are signed; platform ambitions are explicitly phased
• TCO includes human review, eval, and change—not API tokens alone
• Cost per successful task and p95 latency have targets and owners
• Change plan names champions, resistance, and adoption metrics
• Contract or SOW assumptions match FinOps and eval reality
• I completed deep study on topics 06, 07, and 22 at minimum
• Artefacts filed in pattern library with sector tag

Key takeaways

• Stage 5 turns validated opportunities into funded, shippable, operable products—not longer discovery.
• Product, commercial, and FinOps form a triangle: scope, funding, and run-rate must move together.
• UX trust and change adoption determine whether economics materialise; build metrics accordingly.
• Presales, RFP, and vendor choices should trace to the same PRD and business case—no parallel stories.
• Use 8D gates to prevent demo-driven delivery and run-rate shocks at scale.

What to do next

1. If entering Stage 5 fresh, confirm Stage 1–4 exit artefacts exist—especially problem statement, eval approach, and risk summary.
2. Study Product Management, Commercial and Financial Modelling, and Performance Engineering and AI FinOps as the core triangle.
3. Open Business case and prioritisation and AI FinOps and Commercial Design when drafting client packs.
4. Continue through topics 23, 25, 26, 29–32 based on your role path above.
5. When Stage 5 exit criteria pass, continue to Leadership.