Reconstruct actual processes
Link work-order, equipment, procurement and settlement events and establish relationships among business objects. Check log completeness and time definitions before analysing waiting, rework and exceptions.
PRACTICE / MANAGEMENT
Processes, performance and human-AI collaboration.
Anchor improvement in business events, so performance measures explain results and guide action.
Methodology: The ZHICE Value-led Operating MethodTHE REAL CHALLENGE
Define the problem first.
Process documents describe intended work; business records show how work actually happens. Energy operations involve assets, work orders, contracts and settlements, so a departmental view rarely reveals all waiting and rework. AI also changes task allocation and review burden. ZHICE considers process diagnosis, performance design and human-AI responsibilities together, examining the overall cycle, quality and risk. Event evidence is cross-checked with frontline interviews, avoiding the mistake of treating a faster local step as end-to-end operating improvement.
BEYOND THE SYMPTOMSLook beyond symptoms.
Look beyond symptoms.
Understand the business logic.
Process analysis must consider business objects and events together. One maintenance case may involve multiple assets, work orders, purchases and settlements; following one document can miss their relationships. ZHICE first checks whether data reflect actual work, then combines event chains and frontline interviews to identify waiting, returns, duplicate entry and exceptions. Efficiency is assessed with outcome quality. A faster step that increases downstream review or rework does not establish overall improvement. Measures should explain results and point to actions people can take.
Human-AI collaboration brings task boundaries and validation ownership into the process together. ZHICE separates information preparation, recommendation generation, professional judgement and final execution, defining where assistance is appropriate and which evidence and reviews its outputs require. Use patterns and workloads are examined by staff experience rather than by tool-use volume alone. Pilots assess complete business processes, including cycle time, quality, exceptions and review costs, before determining expansion or changes to roles, training and controls.
A TOOL FOR THE WORKING SESSION
Turn diagnosis into verifiable questions.
Joint diagnosis of processes and human-AI work
Swipe to explore the complete table
| Diagnostic focus | Business evidence required | Analytical emphasis | Improvement direction |
|---|---|---|---|
| Business-object relationships | Mappings and event times across equipment, work orders, contracts, procurement and settlement. | Relationships omitted by a single-process view and the distinction between missing logs and actual waiting. | Establish object and event relationships, prioritising data and responsibility gaps that affect judgement. |
| Handovers and exceptions | Reasons for returns, supplementary information, waiting times, exception handling and manual re-entry records. | Whether delivery standards are clear and whether problems arise from missing inputs, authority conflicts or capability gaps. | Define acceptance conditions and exception paths; resolve handover problems before automating. |
| Performance relationships | Outcome and process measures, definitions, evaluation arrangements and observed behaviours. | Whether measures encourage local optimisation or shift costs, workload or risk downstream. | Connect outcomes and drivers, add quality and risk constraints, and assign responsibility for variance. |
| AI-assisted tasks | Task inputs and outputs, evidence references, human reviews, corrections and adoption records. | Differences across tasks and people, including review burden, misuse and exceptions. | Define assistance boundaries, evidence requirements and human responsibility, with risk-based evaluation and training. |
Draws on object-centric process analysis and human-AI research. Findings require business verification; tool logs or external studies do not directly establish operating benefits.
Four checks before the engagement begins
01Are business-object relationships complete, and could missing records distort the process assessment?
02Does any local acceleration add review, rework or re-entry for another role?
03Can measure definitions and responsibilities explain variance and support specific corrective action?
04Are AI quality, errors and review costs examined separately by task and user group?
HOW WE WORK
From diagnosis to implementation.
Three connected levels form the working approach. Select a stage to explore the detail.
Trace performance conflicts
Examine whether departmental measures encourage opposing behaviours, such as lower purchase prices increasing downtime or shorter handling times increasing rework and review workload.
Observe human-AI responsibilities
Distinguish tasks suitable for assistance, those requiring review and those retaining human judgement. Examine differences by staff experience and effects on downstream quality.
Redesign handovers
Specify completeness, acceptance conditions and exception paths for each delivery. Reduce unnecessary back-and-forth before introducing automation or intelligent assistance.
Connect metrics
Link outcome measures to actionable drivers, with quality and risk constraints. Specify definitions, owners, update frequency and responses to variance.
Design collaboration controls
Require evidence references, human confirmation and escalation for AI-assisted tasks. Tailor training, sampling and evaluation to staff experience and task risk.
Pilot a representative process spanning several functions and establish a comparable baseline. Observe cycle time, quality, workload and risk together, recording external changes.
Review effects by business object and exception type, looking for displacement of problems. If pilot evidence is insufficient, continue collecting it rather than claiming enterprise-wide benefits early.
Feed review findings back into processes, roles and data rules, with ongoing ownership. Introduce system and management changes together to avoid creating further manual re-entry.
TANGIBLE OUTPUTS
Deliverables designed for
real work.
Agree depth, data conditions and acceptance criteria at initiation. The engagement scope follows the tasks confirmed by both parties.
End-to-end business and event relationship map
Process breaks and control-point assessment
Operating metric tree and definition dictionary
Human-AI task allocation and pilot evaluation plan
ILLUSTRATIVE SCENARIO
From equipment defect to maintenance settlement
Illustrative scenarioAn energy business plans to improve maintenance across production, procurement and finance, while trialling a knowledge assistant. This example illustrates diagnosis and validation only.
- 01
Link equipment, defect, work-order and procurement records to locate specific waiting and rework.
- 02
Define handover standards for spare-part requirements, maintenance acceptance and settlement documentation.
- 03
Trial AI assistance in appropriate steps such as knowledge retrieval, retaining specialist review and safety controls.
- 04
Assess cycle completion, repeat repairs, review burden and actual adoption together.
BOARDROOM QUESTIONS
Focus the discussion on these choices.
01Has local acceleration improved the full business cycle and outcome quality?
02Do measures create foreseeable conflicts or inappropriate incentives?
03Are AI-related review, correction and training costs included in the assessment?
RESEARCH FOUNDATIONS
Research brought into
consulting practice.
We translate management research and technological advances into working methods for energy businesses. The frameworks below are ZHICE's synthesis of the research and industry questions.
Wil M. P. van der AalstNo AI Without PI! Object-Centric Process Mining as the Enabler for Generative, Predictive, and Prescriptive Artificial Intelligence · 2025-07-31
Application in the ZHICE approachUses research combining object-centric process mining and AI to support process diagnosis and intelligent-application design through business objects and event relationships.
Scope and limitationsThis source is a preprint of a conference keynote paper offering a research perspective. It does not establish universally effective product capabilities or verified enterprise benefits.
Read the original researchErik Brynjolfsson、Danielle Li、Lindsey Raymond / The Quarterly Journal of EconomicsGenerative AI at Work · 2025-02-04
Application in the ZHICE approachUses empirical differences in AI-assisted performance by worker experience to incorporate individual variation, quality and learning into task and evaluation design.
Scope and limitationsThe study concerns customer-support work in one firm. Its results cannot be directly extrapolated to expected improvements in energy production, trading or management roles.
Read the original researchOUR PRACTICE
Explore the critical questions.
Each practice examines diagnosis, solution design, implementation and delivery requirements in detail.
01 / MANAGEMENT
Organisation, control and decision mechanisms
Allocate authority along the value chain, balancing central control with local responsiveness.
Diagnosis · Design · ImplementationExplore this practice02 / MANAGEMENT
Processes, performance and human-AI collaboration
Anchor improvement in business events, so performance measures explain results and guide action.
Diagnosis · Design · ImplementationExplore this practice03 / MANAGEMENT
Asset lifecycle and value management
Connect investment, operation, maintenance and retirement through evidence-based choices on cost, risk and performance.
Diagnosis · Design · ImplementationExplore this practiceSTART WITH A DECISION
Start with a decision.
Define the work around it.
Discuss the choices, constraints and evidence needed, then agree the engagement scope and work programme together.