CASAN · Level-Up & AI Adoption
From your level to
measurable outcomes
Once you know your CASAN level, the next question is what to do about it. This is how we close the gap — a service package for every level, concrete AI-adoption projects, and a pragmatic engagement journey. CASAN is modular: start small with one assessment or pilot and expand as the value proves out.
Partner with FPT to move from your current level to the next — with measurable outcomes.
Service packages
The right engagement for every level
Each CASAN level has a distinct problem to solve and a matching FPT service package.
Customer problem
AI is used in fragmented ways, with no shared policy.
FPT service package
AI readiness assessment, executive workshop, a starter policy set, and use-case discovery.
Expected outcome
A clear baseline, a minimum policy, and your first prioritized use cases.
Customer problem
Tools are in place, but adoption is uneven.
FPT service package
Copilot rollout, productivity measurement, and prompt / context & safe-use training.
Expected outcome
Higher individual and team productivity, with lower risk.
Customer problem
Many pilots, but they are hard to scale.
FPT service package
Data readiness, AI governance, a use-case portfolio, reference architecture, and an evaluation framework.
Expected outcome
A platform to scale AI safely, with risk under control.
Customer problem
You want to put agents into real workflows.
FPT service package
Agent workflow design, AI delegation, AgentOps, a control plane, and red-teaming.
Expected outcome
Controlled automated workflows with measurable business KPIs.
Customer problem
You want to re-architect the business around AI.
FPT service package
AI-native operating model, outcome-based services, platform strategy, and a transformation roadmap.
Expected outcome
AI as a core operating and competitive capability.
Level-up roadmap
What it takes to reach the next level
The goal, the key actions, and how FPT helps at each step of the journey.
Goal: Move from fragmented individual experiments to official, safe, productivity-focused AI use.
Key actions
- Establish an Acceptable Use Policy — define clear rules on how AI can be used.
- Secure your data — restrict sensitive or proprietary data from public AI tools.
- Provide official enterprise licenses to ensure a secure environment.
- Target 5–10 quick-win use cases that visibly boost daily productivity.
- Train the workforce on AI literacy, safe usage, and basic prompt engineering.
- Measure early impact — time saved, task completion rates, employee satisfaction.
How FPT helps: AI readiness assessment, a starter AI policy set, a training program, use-case discovery workshops, and pilot deployment.
Goal: Move from individual productivity to a repeatable, controllable organizational capability.
Key actions
- Prioritize use cases — build a centralized portfolio ranked by business impact, feasibility, and risk.
- Standardize and protect data — data classification, sensitivity labels, access controls, and DLP.
- Define the AI lifecycle — approval, build, test, deploy, monitor, and retire.
- Build centralized assets — reusable prompt libraries, agent templates, validation suites.
- Establish dedicated governance — form an AI Governance Board with AI Product Owners, Data Owners, and Security Owners.
How FPT helps: AI governance framework, data readiness assessment, enterprise AI architecture, a standard operating model, and deployment services.
Goal: Move from standardized processes to automated workflows run by controlled AI agents.
Key actions
- Select automated workflows with a clear ROI and controllable risks.
- Design an AI Delegation Architecture — define boundaries, tools, and human-approval gates.
- Build a control plane — manage Agent identities, permissions, audit trails, and rollback procedures.
- Establish AgentOps — monitor AI Agent performance, costs, latency, and hallucination rates.
- Shift the human role from checking every output to managing exceptions and reviewing high-risk tasks.
How FPT helps: AI agent deployment, AgentOps operations, Harness Engineering, workflow redesign, security integration, and change management.
Goal: Move from workflow automation to an AI-native enterprise architecture.
Key actions
- Redefine the operating model — redesign core processes to be event-driven and AI-orchestrated.
- Build an enterprise knowledge layer — a real-time memory and knowledge base all AI Agents can securely access.
- Implement Multi-Agent orchestration — deploy specialized AI Agents that collaborate on complex problems.
- Upgrade data pipelines — support real-time inference, continuous feedback loops, and continuous learning.
- Shift KPIs — measure new business models, revenue growth, customer experience, and organizational learning speed.
How FPT helps: Strategic transformation partner, enterprise architect, AI-native platform builder, data architecture partner, governance partner, and co-innovation partner.
AI adoption
Autopilot services you can start with
Concrete, self-running workflows where AI agents do the work under human oversight — each with a reference delegation level.
| Autopilot service | Key outcome | Reference delegation |
|---|---|---|
| Software delivery | User stories completed, tests passing, pull requests merged, documentation updated. | L3–L4 small tasks · L1–L2 large or high-risk |
| Testing & QA | Test cases generated, regression suites maintained, defects found earlier, coverage improved. | L3–L4 |
| IT operations | Tickets handled within SLA, incidents triaged, patches applied, uptime improved. | L3 bounded infra · L4 tier-1 · L1–L2 complex |
| Document processing | Records extracted, classified, validated, and cross-checked. | L3–L4 templated · L1–L2 complex legal |
| Contract review | Contracts reviewed against a playbook, exceptions escalated, redline suggestions generated. | L2 critical · L3 standard NDAs |
| Financial reconciliation | Invoices matched to PO and goods receipt, discrepancies flagged, faster month-end close. | L3 clean matches · L2 discrepancies · L1 large amounts |
| Customer support | Tickets handled within SLA, CSAT maintained or improved, humans handle exceptions. | L4 tier-1 · L2 tier-2 · L1 sensitive |
| Procurement leakage detection | Off-contract spend, duplicate vendors, expired contracts, and price creep detected; savings opportunities created. | L2 · never changes vendors/contracts autonomously |
Reference
The six AI delegation levels
How much authority an AI agent holds — from observe-only to high-risk autonomy under strict control.
Observes, searches, summarizes, classifies — no system changes.
e.g. Meeting summaries, ticket classification
Creates drafts; humans approve 100%.
e.g. Draft emails, proposals, code
Recommends options; humans decide.
e.g. Pricing options, test strategy, shortlists
Executes low-risk tasks within clear limits.
e.g. Generate test cases, update docs, route tickets
Runs a workflow with guardrails and audit.
e.g. Tier-1 support, incident triage, document processing
Automates complex areas under strict control.
e.g. Fraud & compliance monitoring, autonomous maintenance
How we engage
A five-step journey to AI-native
Start with a diagnosis, prove value in a short pilot, then scale with governance.
Diagnose
A workshop with leadership and process owners, plus a structured assessment across eight dimensions. Your organization is scored on the CASAN 1–5 scale overall and per process.
Blueprint
Design your CASAN architecture: a governed data layer, an enterprise Harness, L0–L5 delegation thresholds, the 5-layer reference architecture, governance, a talent plan, KPIs, and dashboards.
Pilot
Run one or two autopilot services with clear ROI and controllable risk as an 8–12 week proof-of-value — with a baseline, measurable KPIs, transparent cost, and a before/after report.
Scale & govern
Roll the autopilot services out to more processes, and bring governance, an agent registry, a validation team, and AgentOps into daily operations.
AI-native operating model
Restructure the unit or process around AI capability — knowledge layer, agent memory, adaptive governance, and new AI-native products and services.
Get started
Build your level-up proposal
Run the level diagnosis, then talk to our team. We'll turn your result into a tailored proposal — the right packages, a starter autopilot pilot, and a 90-day plan with measurable KPIs.
