- Spend classification & category intelligence
- Supplier risk radar
- Contract leakage detection
- Guided buying
Spend reduction · cycle time · risk
Artificial intelligence
AI is a tool, not an objective
AI is not an objective nor an initiative — it is merely a tool to achieve one. The question is never “where can we use AI?”, but “which operational result do we want, and where does AI help us get there faster?”
Value-first
Process & governance
Adoption
Measured impact
The point of view
The AI gap is no longer adoption. It is value capture.
A large share of organisations run AI pilots; only a minority report any bottom-line impact. The differentiator is rarely the model — it is whether the surrounding process, data, governance and behaviours were redesigned to let AI change a decision that matters.
Why value stalls
The signals worth taking seriously
A few market signals that frame how I approach AI in operations — pragmatically, and with the P&L in mind.
~4 in 10organisations report any EBIT impact from AI — adoption is high, value capture is not.
Workflowredesign is consistently the lever most linked to bottom-line impact — not the tool itself.
Risk & rulesgovernance, risk and regulation (incl. the EU AI Act) are now top deployment barriers.
Sources: McKinsey, The State of AI: Global Survey (2025); Deloitte, State of Generative AI in the Enterprise; European Commission, EU AI Act timeline.
My point of view
AI creates value when it becomes part of the operating model
Not a side tool sitting next to the process — a redesigned way of deciding and working. Value is the product of four factors; if any is missing, impact collapses.
Process redesign
Decisions, exceptions, handovers and SOPs reshaped around the AI.
×
Data readiness
Sources, quality, ownership and refresh rate good enough to trust.
×
Governance
Human validation, controls, risk and compliance built in.
×
Adoption
Training, rituals, behaviours and clear KPI ownership.
= measurable operations impact
Where AI helps
Where AI changes operations
The highest-value use cases sit where decisions, exceptions and data meet. My focus is the operating routines, not the tools — procurement, supply chain, industrial performance and the transformation office.
- Demand forecasting
- Inventory optimisation
- S&OP simulation
- Control-tower alerts & flow recommendations
Cash · service level · lead time
- Predictive maintenance
- Quality analytics
- Scheduling & energy optimisation
- Supervisor copilots
Productivity · OEE · scrap · energy
- Initiative tracking & benefits cockpit
- Risk detection
- SteerCo synthesis
- Knowledge management
EBITDA delivery · transparency · pace
Why it is hard
Why clients struggle to implement
The barriers are usually operational, not only technical. The common symptom: pilots stay disconnected from the processes, roles, KPIs and routines that actually generate value.
The offering
An offer ladder, from first scan to embedded delivery
A productised journey: a low-friction entry point, then as much embedded transformation leadership as the value justifies. Each step is fixed-scope, with senior involvement throughout.
1
AI Opportunity Scan
Map 2–4 processes, surface 5–10 use cases with value hypotheses and feasibility, and leave you a prioritised next step.
2
AI Operations Value Sprint
Use-case heatmap, prioritised backlog, top 3–5 business cases, a governance model and a 3–6 month roadmap.
3
AI Use-Case Design Sprint
Turn one use case into a buildable pilot: as-is/to-be process, user stories, data sheet, risk controls and a pilot plan with KPIs.
4
AI Pilot-to-Scale PMO
Stabilise the pilot, run the delivery cadence, track adoption and benefits, and produce a clear scale / stop / redesign decision.
5
Fractional AI Operations Lead
Embedded leadership of the AI-in-operations portfolio: governance, value reviews, adoption and capability transfer to your teams.
Each step is fixed-scope with direct senior involvement — adapted to your company size, complexity, data access and the functions involved.
A practical sequence
An example 12-week journey
From ambition to a clear scale decision — with a real decision point at week 12: scale, stop, redesign, or hand over to internal teams.
Week 1
Opportunity Scan — align priorities, map first opportunities.
Weeks 2–4
Value Sprint — prioritise use cases, size value, define the roadmap.
Weeks 5–8
Use-Case Design — process redesign, requirements, data and controls.
Weeks 9–12
Pilot-to-Scale PMO — adoption, benefits tracking, scale recommendation.
Deliverables
What clients get
Decision-ready artefacts that bridge business, IT, data and operations. The objective is not another slide deck — it is a common operating language for sponsors, users, IT, data teams and finance.
Use-case heatmap
Where AI can create value.
Prioritised backlog
What to pursue first.
Business case
Why it matters financially.
Future-state process
How the work actually changes.
Pilot cockpit
How delivery is governed.
Scale playbook
How to industrialise what works.
Why me
Why Chen Advisory for AI in operations
I am not an AI engineer — I am an operations transformation consultant who makes AI deliver value. The hard part is rarely the model; it is the process, governance, adoption and value tracking around it. That is exactly my craft.
Operations strategy
Kearney experience in industrial transformation, EBITDA improvement and operating-model roadmaps.
Execution excellence
Orphoz / McKinsey delivery heritage: PMO, adoption, productivity and working-capital results.
Procurement & supply chain
A pharma procurement AI workstream, plus segmentation, inventory, flow and planning experience.
Risk & controls
EY background in process controls, risk mitigation and auditability — essential as AI meets regulation.
How to start
A low-risk first step
Identify the highest-value opportunities and build alignment — before committing to any build.
1
30-min alignment
Confirm the business priorities and the scope worth scanning.
2
5-day Opportunity Scan
Identify the high-potential AI use cases and size the prize.
3
Decision on next step
Value Sprint, use-case design, or pilot PMO — only if the value justifies it.