Selected engagements
What changed in the running of the business
Six representative engagements, described by the operating change rather than the feature list. Figures are described qualitatively here; we walk clients through the underlying methodology in a reference conversation.
Global banking group
Underwriters stopped retrieving and started judging
The bank's credit analysts spent most of each memo assembling policy, bureau data and precedent rather than exercising judgement. We modelled the credit decision as an enterprise ontology, built a digital employee to assemble the memo against it, and left approval with the underwriter — with a complete record of what the agent used and why it flagged what it flagged.
→Operating result
Preparation time dropped substantially and the turnaround SLA was met consistently. Judgement time per memo rose, because that is what the analysts were being paid for.
→Governance result
Every decision became reproducible on demand — rule version, precedent, inputs — which the bank previously could not produce at all.
→Capability left behind
Internal team owns the ontology and extends it to the next credit product without us.
Agentic businessOntology90-day pilot
Southeast Asian manufacturing group
Twelve sites finally shared one model of their own equipment
Each site had its own workarounds and a great deal of knowledge held by a few people. We built a shared ontology of assets, failure modes and work procedures, grounded a diagnostic digital employee in it, and deployed to the technician tablet. The agent was the visible part; the durable asset was the shared model.
→Operating result
Unplanned downtime and mean time to repair both came down, and the improvement held as it rolled to further sites.
→Organisational result
Site engineers began working from one equipment model instead of twelve local ones, which changed how they trained each other.
→Where it went next
The same model became the basis for operator onboarding on new equipment lines.
Ontology & contextAgent designMulti-site rollout
Regional insurance group
Compliance CPD moved from spreadsheet chasing to evidence on demand
Regulatory training was administered by three people chasing spreadsheets across business units, and every audit began with a scramble. We mapped the obligation register to roles and jurisdictions, automated assignment and credit tracking in wizBank, and replaced passive completion with scenario assessment on the higher-risk material.
→Operating result
Administrative effort fell sharply and an audit pack that previously took days now assembles in minutes.
→Quality result
Assessment evidence became defensible — a regulator can see demonstrated competence, not attendance.
→Boundary held
The learning system stayed the system of record. The AI layer reads from it; it does not duplicate it.
Next-generation learningCompliance & CPDwizBank
Telecom operator, APAC
Tier-1 deflection that customer satisfaction improved despite
The operator had attempted a support bot twice and abandoned both because forced deflection damaged satisfaction scores. We grounded the new agent in the service ontology and policy, and designed the handoff as part of the model: any case the agent was not confident about reached a human immediately with full context. The confidence threshold was the design, not a safeguard added later.
→Operating result
A material share of tier-1 contacts resolved without a handoff, average handling time fell, and satisfaction improved after launch — the reverse of the previous two attempts.
→Design result
Escalation rate is now a monitored operating metric, reviewed like any other, rather than a complaint category.
→Workforce result
Tier-1 staff moved to complex and coaching work, with reskilling sequenced ahead of the operating change rather than after it.
Agentic businessService operationsSkills & training
Private hospital provider
Ward-level competence visible on one screen, per person
Clinical governance wanted assurance that staff were competent for the work they were doing, not trained against a checklist. We built a capability taxonomy mapped to ward roles, assessed through scenario and simulation, and published a live competence picture per ward — visible to managers and executives, and in a controlled form to inspectors.
→Operating result
Clinical incident rate fell, and reliance on agency cover for competency gaps dropped as gaps became visible and addressable.
→Governance result
Competence assurance became something a manager or an inspector can read directly, rather than reconstruct.
→Where we stopped
No autonomous agent was deployed into clinical decision-making. Administrative delegation only.
Next-generation learningCompetence assuranceSimulation
Consumer goods group, APAC
One standard, delivered identically, including to partners
The business had five regional training programmes, five quality standards and no reliable view of who was actually ready. We consolidated onto one curriculum architecture delivered in local languages on mobile, and — the part that mattered most — brought outsourced and channel staff into the same capability system as employees, assessed the same way.
→Operating result
One programme architecture replaced five, time to proficiency shortened across territories, and audit pass rates improved.
→Ecosystem result
External partners stopped being a quality blind spot, because they were inside the system rather than alongside it.
→Management shift
Supervisors got live readiness data, so they could intervene early rather than certify late.
Next-generation learningPartner enablementMobile delivery
On confidentiality and numbers. Client names are withheld under confidentiality agreement. We describe results qualitatively on this page deliberately: baselines, measurement windows and market conditions differ so much between engagements that side-by-side percentages would imply a comparability that does not exist. In a client reference conversation we will give you the actual figures, the methodology behind them, and the caveats — including engagements where the result was less than we hoped.