Institutional readiness
Map leadership objectives, teacher and student needs, existing systems, data categories, internal IT capability and risk tolerance.
AI advisory
Start with institutional requirements, evidence and governance — not a shopping list.
ZYK's content-v2 advisory architecture combines strategic readiness and roadmap work with technical model evaluation, procurement and private-AI migration concepts. The aim is to help institutions decide what should be built, governed and tested before committing to hardware, models or platform choices.
Strategy & readiness
Decide what the institution actually needs before deciding what to buy.
Use-case prioritisation
Separate high-value, lower-risk early opportunities from later use cases that require stronger validation, approval or technical controls.
Deployment choices
Compare on-premise private infrastructure, controlled hosting and hybrid approaches against privacy, performance, cost and operational capability.
Governance design
Bring identity, access, logging, retention, consent, content controls, human oversight and responsibility into the architecture before deployment.
Technical advisory
Turn the strategy into defensible technical decisions.
Model Evaluation Service
Compare candidate models against institution-specific tasks such as Chinese and English performance, curriculum reasoning, coding, latency, safety behaviour and hardware efficiency.
AI Procurement Advisory
Translate educational, governance and capacity requirements into a defensible hardware, deployment and supplier specification rather than beginning with a preferred vendor.
Private AI Migration
Map current public AI use, identify suitable workflows for institutional migration, classify data sensitivity and design a staged route into governed private infrastructure.
From strategy to pilot
Break “AI transformation” into stages that can be tested.
1 · Discover
Interview key stakeholders and establish requirements, constraints, current AI use and risk areas.
2 · Design
Define the target architecture, governance boundaries, success measures and pilot population.
3 · Pilot
Test bounded educational or operational workflows and record adoption, quality, governance, capacity and support burden.
4 · Decide
Use evidence to expand, modify, pause or replace the proposed technical path.
5 · Scale
Expand only when controls, training, integration, capacity and support are ready.
Evaluation
A benchmark should resemble the work the institution actually cares about.
Education tasks
Lesson planning, rubric interpretation, bilingual tasks, document analysis, coding support and subject-specific questions.
Operational tasks
Admissions FAQs, policy retrieval, internal document workflows and other bounded institutional use cases.
Technical measures
Latency, throughput, memory requirements, concurrency and infrastructure cost can be considered alongside output quality.
Governance measures
Institutions may also need to compare controllability, deployment location, logging options, licensing and model lifecycle considerations.
Procurement & model neutrality
Preserve institutional choice.
Capacity & cost
Translate user numbers, concurrency, context length, media workloads and service expectations into testable capacity assumptions rather than buying by GPU count alone.
Acceptance criteria
Define supplier evaluation dimensions, deployment assumptions and technical acceptance tests before procurement. Final procurement, legal and financial decisions remain with the institution and its advisers.
Exit & migration
Consider model replacement, data export, knowledge migration and supplier exit at architecture stage to reduce future lock-in.
Private-AI migration
Move the right workflows — not everything at once.
Discover
Identify how staff and students currently use public AI tools, APIs and informal workarounds.
Classify
Separate low-risk experimentation from workflows involving institutional, student or sensitive information.
Prioritise
Select use cases that offer clear educational or operational value and are practical to govern.
Pilot & evaluate
Move a bounded set of workflows into the private environment and evaluate quality, usability, governance and cost before expansion.
These advisory services are strategic content-v2 concepts. Final methodologies, deliverables, pricing, professional responsibilities and jurisdiction-specific requirements should be defined before commercial launch. ZYK does not present this page as regulated legal, audit or other professional advice.