01 · Evidence before seats
Discover
Name owners, freeze a value case, and map Steady panes before you scale. Wave 0 starts here: Security, GRC, Cost and Value, Licence, Adoption.
Panes · Wave 0 Steady · SEE awareness
Managed AI Services
Discover. Build. Run. Improve.
A managed path through the window — Wave 0 foundations, scoped build, operating cadence, and a closed Adoption Loop. Govern sits across all four. Guide prices. No invented ROI. DIGITAL SKILL WORKS LTD / 12 PANES.
Phases
Four phases
Mapped to panes and Wave 0
Govern runs across Discover, Build, Run, and Improve: risk decisions, control evidence, exceptions, and Counsel dials. Unmeasured stays red. RAG is computed from evidence — never asserted by an Action.
02 · Scoped delivery on foundations
Build
Architecture, configuration, and evaluation with Counsel dials held. Build only what Steady can hold — agents and workflows with kill criteria.
Panes · Security · Frontier · Cost & Value
03 · Operate the estate
Run
Production operations, monitoring, and service management on a cadence. Reclaim, register hygiene, and value reconcile — not usage theatre.
Panes · Licence · Adoption · Narrative
04 · Close the loop
Improve
Continuous improvement from measured evidence. Issues become Actions and Waves; Tasks land with owners. Roadmap is a supply list, not a wish plan.
Panes · Horizon · Narrative · Counsel
Capabilities
Fourteen deliverables on the hub
Each capability lists deliverables and example measures. Phase tags show where the work usually sits; Govern still spans the loop.
01 · discover
Strategy and discovery
Deliverables
- Wave 0 owner map and risk appetite sketch
- Frozen value case outline
- Prioritised discovery findings against four pains
Example measures
- Named risk and value owners on Steady panes
- Value case frozen before seat spray
02 · discover
Data foundation
Deliverables
- Source register sketch (permissions, freshness, action rights)
- Overshare hotspot list for top estates
- Data readiness note for first cohort
Example measures
- Hotspots with severity and owner
- Sources recertified on a named rhythm
03 · build
Architecture
Deliverables
- Reference architecture for scoped build
- Integration and identity boundaries
- Bounded frontier rules for agents
Example measures
- Architecture decisions logged with owners
- Kill criteria dated before release
04 · build
Build and configuration
Deliverables
- Configured services for the agreed scope
- Environment and release checklist
- Configuration runbook for handover
Example measures
- Scoped items delivered versus SoW
- Handover checklist signed by owners
05 · build
Evaluation and responsible AI
Deliverables
- Evaluation plan with pass / fail gates
- Responsible AI checklist tied to GRC
- Counsel dial draft for release risk
Example measures
- Gates passed before production
- Exceptions logged with expiry
06 · build
MLOps and LLMOps
Deliverables
- Release and retirement records for models and agents
- Prompt and version hygiene sketch
- Rollback and kill-switch path
Example measures
- Inventory with named business owners
- Retirement dates on the register
07 · run
Production operations
Deliverables
- Operating cadence for Steady and Move panes
- Incident-AI runbook prompts
- Service health checklist
Example measures
- Cadence held (weekly / monthly as agreed)
- Incidents with AI context closed with owners
08 · run
Model and agent monitoring
Deliverables
- Monitoring sketch for quality, cost, and drift signals
- Alert owners and thresholds
- Monthly monitoring digest template
Example measures
- Alerts with named responders
- Unmeasured signals treated as red
09 · run
Security and privacy
Deliverables
- Overshare remediation programme list
- Privacy and DPIA trigger fields on the register
- Access review rhythm
Example measures
- Hotspots remediated or accepted with expiry
- DPIA triggers fired before material change
10 · run
Governance and compliance
Deliverables
- AI register with control evidence hooks
- Exception and acceptance log
- Board / risk narrative draft across Steady panes
Example measures
- Register coverage versus estate inventory
- Exceptions with expiry and owner
11 · run
Adoption and change
Deliverables
- Cohort plan beyond seat volume
- Literacy plan and clinic design
- Adoption pulse capture for clinics
Example measures
- Eligible → Beneficial funnel for cohort 1
- Weekly-active habit, not licensed headcount alone
12 · run
FinOps and license optimization
Deliverables
- Cost stack lines with owners
- Idle-seat reclaim runbook (30/60/90)
- Inform → Optimize → Operate reporting sketch
Example measures
- Licences versus active use reconciled
- Reclaim waves dated before renewal
13 · run
Service management
Deliverables
- Service catalogue sketch for AI capabilities
- Request and change path for scoped work
- Handover and escalation matrix
Example measures
- Requests with SLA owners
- Changes logged against the register
14 · improve
Continuous improvement
Deliverables
- Improvement backlog as supply list (Waves)
- Issue → Action → Task loop with evidence
- Quarterly improve pack for board and ops
Example measures
- Issues closed from evidence, not assertion
- Actions that do not recolour RAG without new evidence
Adoption Loop
Evidence to Waves
Three levels, three lenses, four objects. Evidence → measures → computed RAG → Issue → decision (fund / accept with expiry / stop / escalate) → Actions → Wave → Tasks. Ties to the board view and operating framework.
Level
Board
Risk appetite, fund / stop / escalate, and narrative across Steady panes. Links to the board view.
Level
Management
Owners, Counsel dials, and operating cadence. Accept with expiry when evidence is incomplete.
Level
Operations
Tasks on Waves, reclaim rhythms, clinics, and register hygiene. Evidence in; theatre out.
Lens
Risk
Exposure, control evidence, exceptions. Unmeasured risk is red until evidence lands.
Lens
Cost
Spend, seats, consumption, reclaim. FinOps Inform before Optimize.
Lens
Value
Cohort outcomes and benefit ledger. Each measure belongs to exactly one lens.
Object
Issue
A computed RAG gap or decision need. Raised from measures, not from opinion.
Object
Action
A funded response. Does not recolour RAG — only new evidence does.
Object
Task
Work on a Wave with an owner and date. Executable, not aspirational.
Object
Wave
A supply package of Tasks. Roadmap is a supply list of Waves, not a multi-year wish plan.
Design rules
- RAG is computed from evidence, not asserted by sponsors or vendors.
- Unmeasured is red — no silent green.
- An Action does not recolour RAG; only fresh evidence moves the colour.
- Roadmap is a supply list of Waves, not a plan that pretends certainty.
- Each measure belongs to exactly one lens: Risk, Cost, or Value.
- Decisions are fund, accept with expiry, stop, or escalate — then Actions land on Waves as Tasks.
Packages
Managed packages
Guide prices. Confirm in writing.
Discover from £5,000 · Build from £12,000 · Run from £8,000 / quarter · Improve from £6,000 / quarter. Excluding VAT if applicable. Govern included across. Same honesty as Foundation / Operate / Scale: named week-1 outputs, inclusions, and exclusions.
01 · discover
Discover
From £5,000
Who: Organisations that need a clear Wave 0 base and honest discovery before build or seat spray.
Focus: Strategy and discovery, data foundation sketch, and Govern hooks across Steady panes. Aligns with Foundation packaging honesty.
Duration: Typically 3–6 weeks of focused discovery, shaped to your calendar.
What you get in week 1
- Named risk and value owners for Steady panes
- Discovery findings against four pains (short pack)
- Frozen value case outline and cost baseline sketch
- Prioritised Issue list with first Wave candidates
Included
- Wave 0 discovery pack (owners, baselines, pains)
- Thin AI register template and DPIA trigger fields
- Overshare hotspot list (top severity)
- Counsel dial draft matched to risk appetite
- Govern note: exceptions, evidence, and RAG rules
- Working sessions with Digital Skill Works Ltd (scoped hours in SoW)
Excluded
- Full estate tool build or custom software
- Legal opinions, DPIA sign-off, or regulator submissions
- Production build, agents, or model hosting
- Ongoing retainer unless added in writing
Measure plane
Before: Demos, seat counts, and unowned risks as the default story.
After: Named owners, frozen value case, Issue list, and a Steady baseline you can reconcile.
02 · build
Build
From £12,000
Who: Teams ready to configure and evaluate scoped AI on foundations that already hold — or will hold in the same SoW.
Focus: Architecture, build and configuration, evaluation and responsible AI, MLOps / LLMOps hygiene. Govern and Counsel dials held through release.
Duration: Typically 6–12 weeks for a scoped build wave. Further waves scoped separately.
What you get in week 1
- Scope freeze: in / out, kill criteria, and release gates
- Architecture sketch and identity / data boundaries
- Evaluation plan with pass / fail criteria
- Register entries for models and agents under build
Included
- Scoped architecture and configuration deliverables
- Evaluation and responsible AI checklist tied to GRC
- Release / retirement records and rollback path
- Counsel dial tuning for release risk (Guardian / Steward / Pathfinder)
- Handover runbook to Run owners
- Govern evidence hooks on the register
Excluded
- Unbounded multi-year build programmes
- Guaranteed model performance or ROI percentages
- Certification audits as auditor of record
- Unlimited agent sprawl without kill criteria
Measure plane
Before: Build pressure without gates, owners, or retirement dates.
After: Scoped delivery with evaluation gates, register coverage, and kill criteria dated.
03 · run
Run
From £8,000 / quarter
Who: Organisations that need production operations, monitoring, and service management on a retainer cadence — not a one-off workshop.
Focus: Production operations, model and agent monitoring, security and privacy loops, governance, adoption, FinOps and licence optimisation, service management. Govern included across.
Duration: Quarterly retainer. Scope and hours confirmed in writing each quarter.
What you get in week 1
- Operating cadence calendar (Steady and Move panes)
- Monitoring and alert owner map
- Register hygiene and reclaim rhythm check
- First quarterly Issue board from computed RAG
Included
- Scheduled operating rhythm and advisory hours (scoped)
- Monitoring digest template and alert owner map
- Idle-seat reclaim and register hygiene loops
- Adoption pulse and clinic design support for agreed cohorts
- FinOps Inform reporting sketch and cost owner check
- Govern: exceptions, evidence, Counsel dial reviews
Excluded
- Outsourced CISO / DPO / 24×7 NOC
- Unlimited build or new product development
- Vendor contract negotiation as counsel of record
- Guaranteed uptime or invented board ROI
Measure plane
Before: Usage theatre, shelfware at renewal, fuzzy ops ownership.
After: Cadence held, reclaim dated, monitoring owned, Issues from evidence.
04 · improve
Improve
From £6,000 / quarter
Who: Teams with a Run baseline who want the Adoption Loop closed: Issues → decisions → Actions → Waves → Tasks.
Focus: Continuous improvement from measured evidence. Board / Management / Operations levels; Risk / Cost / Value lenses. Govern included across.
Duration: Quarterly improvement cycle. Waves funded from evidence, not aspiration.
What you get in week 1
- Measure set mapped to Risk / Cost / Value (one lens each)
- Computed RAG snapshot — unmeasured marked red
- Issue board with decision options
- First Wave supply list (Tasks with owners)
Included
- Quarterly improve pack (measures, RAG, Issues, decisions)
- Action and Wave facilitation (scoped hours)
- Design-rule coaching: RAG computed, Actions do not recolour
- Links into board narrative and operating framework cadence
- Counsel dial review against improvement appetite
- Govern: accept-with-expiry discipline and evidence refresh
Excluded
- Roadmaps that pretend multi-year certainty
- Recolouring RAG without new evidence
- Unlimited Waves beyond the funded supply list
- Invented benefit claims for the board pack
Measure plane
Before: Backlogs without evidence; green asserted by sponsors.
After: Issues from measures, decisions with expiry, Waves as supply, Tasks owned.
Next
Same window. Different depth.
Core packages (Foundation / Operate / Scale) remain the method brands. Managed AI adds Discover → Build → Run → Improve with Govern across. Enterprise AI Management is the operating service story. Products are draft / hidden rehearsal pages — not marketplace offers.
DIGITAL SKILL WORKS LTD / 12 PANES · Prices are guides. See disclaimer.