Methodology and global patterns
The board-level framing, five questions, taker/shaper/maker, the five-layer assessment, the transformation journey, a banking example and the global patterns the agents mirror.
Updated: September 2026Board-level framing: work and judgment
AI is the first technology that automates judgment as well as work. Earlier waves accelerated data entry, calculation and communication and always left the decision to people. Agents classify, prioritise, recommend and, within defined limits, decide.
That is why this is an operating-model redesign: which decisions are delegated to agents, which stay with people, how the limits are measured and audited. The answers belong at board level.
Five questions boards must answer
- Value pool: where and how large is the value AI can create in the enterprise?
- Processes: which processes will be redesigned with agents, which merely accelerated?
- Organisation and talent: how do roles, skills and structure change over the next 24 months?
- Adoption strategy and platform: taker, shaper or maker; which platform choices follow?
- Governance: how are decisions bounded, audited and stopped when necessary?
Adoption strategy: taker, shaper, maker
Every enterprise chooses what to build itself. The right question is which role to take in which area. Most enterprises mix the three strategies scenario by scenario.
| Strategy | What it means | When it fits |
|---|---|---|
| Taker | Use ready-made AI capabilities as they are | Standard, non-differentiating work; quick benefit |
| Shaper | Build agents on a platform with the enterprise’s data and processes | Core processes that create advantage; sovereignty requirements |
| Maker | Develop own models and infrastructure | Very large scale, unique data and strong engineering capacity |
Aginies is the platform for the shaper strategy: the enterprise builds its agents with its own data, processes and limits, and makes model and infrastructure choices without lock-in.
The five-layer assessment
We decide with numbers which AI ideas are worth building and when to start. Every scenario passes through five layers, and each layer ends in a clear decision: proceed, fix or stop.
| Layer | Question | Output |
|---|---|---|
| 1 · Value | Why should this be done at all? | A business case in numbers and a prioritised scenario portfolio |
| 2 · Readiness | Do we have the inputs we need? | A readiness scorecard with gaps closed before the project |
| 3 · Feasibility | Can we actually build this? | A realistic MVP plan with clear scope and identified risks |
| 4 · Operations | How will it live after go-live? | A living, measured, continuously improving AI product |
| 5 · Governance | Is it secure, controlled and auditable? | An audit-ready, controlled, trustworthy AI operation |
Value layer checks:
- Problem: which business problem, how is it solved today, what does it cost?
- Benefit: cost reduction, revenue, risk reduction, experience; which and how much?
- Measurement: target KPIs, expected improvement, ROI and payback period
- Strategic fit: aligned with company strategy, a management priority?
Readiness layer checks:
- Data: sources available and digital; quality, freshness and ownership clear?
- Integration: systems and APIs ready; real-time access possible?
- Organisation: does the business unit own it; sponsor and experts identified?
- Technology: cloud/security infrastructure, AI platform and licences ready?
Feasibility layer checks:
- Technical fit: should AI solve this at all; classic software, classic ML, agents?
- Complexity: how many systems, data sources and users are affected?
- Model performance: expected accuracy, impact of a wrong output, human approval needed?
- Project feasibility: MVP duration, team, external dependencies, technical risks
Operations layer checks:
- Operating model: product owner, process owner, operations and support defined
- User management: training, adoption tracking, feedback loops
- Monitoring: performance, metrics and error management in real time
- Financial operations: monthly run cost, growth projection, budget ownership
Governance layer checks:
- Security: sensitive data map, data residency, control of data flowing to models, masking
- Compliance: KVKK and GDPR impact, sector regulation, internal policy requirements
- Risk: assessment of hallucination, wrong decisions, data leakage and prompt injection
- Control: human approval points, automation limits, kill-switch, decision history
Duration, participants, tools and red flags of each layer:
| Layer | Duration | Participants | Tools and outputs | Red flags |
|---|---|---|---|---|
| Value | 1–2 weeks | Business owner, finance, sponsor | Scenario portfolio workshop and value card (problem, volume, cost, KPI) · Comparison with the closest scenario in the agent library | Benefit without numbers: “efficiency will rise” but no volume or cost · No sponsor; the request comes from IT and the business watches |
| Readiness | 1–2 weeks | Data owner, integration team, process owner | Readiness scorecard: data, integration, process, organisation (each axis 0–2) · API and data-access test; sample data set | The target system has no API, only a screen · Process rules live only in people’s experience |
| Feasibility | 2–3 weeks | Aginies engineering, enterprise architecture, information security | Approach decision: classic software, classic ML, agents, generative AI · Prototype and draft test corpus; wrong-output impact analysis | A wrong output is irreversible and no approval point is defined · Scope grows at every meeting |
| Operations | 1 week (plan) + continuous after go-live | Product owner, process owner, operations and support | Operating-model template: roles, thresholds, review sample, budget owner · Dashboards and the monthly review calendar | Nobody owns operation after go-live · No budget owner for the monthly run cost |
| Governance | 2 weeks, in parallel with the other layers | Information security, compliance, legal, internal audit | Data map and masking rule set; approval matrix · Control set and evidence list; sector regulation mapping | Personal data involved, no map · Governance postponed as “we will add it later” |
Example: how the dialogue banking assistant passed the five layers:
- Value: Simple transactions reaching the call centre and cost per transaction put in numbers; target 40% reduction. Sponsor: digital banking.
- Readiness: Transaction APIs and a test environment for core banking and cards existed; OTP stack ready; policy thresholds written with information security.
- Feasibility: Approach: supervisor + four specialist agents; per-transaction OTP and approval band against wrong-output risk; test corpus of 212 scenarios.
- Operations: Product owner digital channels, process owner operations; monthly review and threshold updates on the calendar.
- Governance: Personal-data map and masking rules; limit increase > 60%, dispute > 5,000 and new payee to approval; all systems in Türkiye.
The transformation journey
The assessment repeats inside a four-stage journey. Each stage’s output is the next stage’s input, and every stage progresses on the same platform.
- Discover and portfolio: value pools, candidate scenarios, a prioritised portfolio and sponsors.
- Assess: each candidate passes the five layers; a proceed, fix or stop decision.
- MVP and validate: the agent is built, the test corpus run, KPIs validated on real data.
- Production and scale: staged rights, monitoring, cost attribution; move to the next scenario in the portfolio.
Banking example: the core stays in place
In a bank the core system stays in place; the AI orchestration layer sits on top of it. Core services are agentised through MCP and REST; agents see account, card, credit and payment services as tools. Rights open in stages.
- Read: balance, statement, limit and transaction queries; measurement starts with zero write risk.
- Controlled write: limit increases, transfers and instructions written under threshold and approval rules.
- Autonomous: proven scenarios run without approval within defined limits; the kill-switch stays ready.
- Conversational banking: an assistant that transacts in the app and by voice
- Core-banking AI transformation: staged agentisation of core services
- Real-time task tracking in operations: request, SLA and bottleneck visibility
- Customer satisfaction management: conversation analysis, complaint and vulnerability signals
Every scenario is observable and costed: run-level traces, cost attribution by unit, channel and branch, quality and error timelines, capacity projection and audit reporting. The compliance frame: banking IT regulation (BDDK), primary and secondary systems in-country, kill-switch and mandatory human approval on critical decisions.
Global patterns
The agents in the library mirror practices proven around the world. The list below summarises those patterns; volume and cost figures are simulation definitions.
- FNOL automation: large carriers opening claims at first contact without call-centre handling load.
- AP automation: shared-service centres running three-way matching from invoice inbox to ERP posting.
- KYB onboarding: payment institutions completing self-serve merchant onboarding in minutes.
- Agentic SDLC pipelines: large software organisations delivering from request to production with agents.
- WISMO deflection: e-commerce converting “where is my order” contacts to self-service through proactive notification.
- AML alert triage: alerts enriched and narratives drafted, with the decision left to the analyst.
- Outage communications: operators and distribution companies cutting inbound call volume with proactive updates.
- Conversational banking: in-app assistants that complete the transaction rather than only inform.
- 100% QA coverage: BPOs and bank contact centres scoring every call instead of a sample.
- Prior-authorisation automation: health systems and payers automating document collection and rule checks.
- Digital front door: patient triage and symptom assessment automated with clinical decisions kept with clinicians.
- Zero-touch onboarding: large enterprises coordinating accounts, equipment and training automatically.