Compass AI: five phases
for successful AI deployment

Most AI deployments end at the pilot stage — not from lack of technology, but from skipping the basics. Compass AI works differently: it secures the foundations first, then effectively builds innovations on top.

Team during a Compass AI implementation workshop

Five phases of AI transformation

Each phase ends with a decision gate — the next stage is only unlocked once the defined exit criteria have been met.

Processes47 mappedReady
Shadow AI7 toolsRisk
Data quality••/100Needs work
IT infraOn-premise + CloudReady
ERP integr.JDE / NetSuiteActive
ClassificationNo policy yetMissing
GDPRPartialIn progress
AI Score62 / 100Moderate
Processes47 mappedReady
Shadow AI7 toolsRisk
Data quality••/100Needs work
IT infraOn-premise + CloudReady
ERP integr.JDE / NetSuiteActive
ClassificationNo policy yetMissing
GDPRPartialIn progress
AI Score62 / 100Moderate
Phase 01

Zero workshop

An audit of business processes, IT infrastructure and data quality. We map the scale of Shadow AI — unofficial use of tools like ChatGPT — and turn it into the starting point for secure deployments. Deliverable: AI Readiness Report.

AI Readiness Report accepted by the sponsor; minimum 3 processes identified for automation.
Anna Wiśniewska — AI Champion, Head of Finance Department
Anna Wiśniewska
Head of Finance Department
AI Champion
HR / Operations
Department being recruited
In progress
Phase 02

Leadership training

We build a shared language and business awareness among board members and senior management. We identify an AI Champions network — change leaders who will be technology ambassadors in their departments. Deliverable: Certified AI Champions Network + ROI Calculator.

Min. one AI Champion per key department; ROI Calculator approved by the board.
0% AI Compliance
12Approved
4In progress
3Pending
Phase 03

AI Governance

Before the first solution is built — we set the rules of the game. We design the AI Usage Policy, data classification matrix and the approval path for new tools. Compliance becomes an operational process, not a one-off declaration. Deliverable: AI Usage Policy.

AI Usage Policy signed by IT, Legal and Compliance; data classification matrix operationally deployed.
Email automation
SalesCustomer serviceAI Agent
Deployed
Invoice analysis (POC)
FinanceOCRAI Validation
In progress
BI Reporting
HRDashboardAnalytics
Planned
Phase 04

Quick Wins & first POC

We move from strategy to action. We identify high-value, low-complexity initiatives, build a prototype in an isolated sandbox environment and measure real results. The POC delivers hard data for the go/no-go decision. Deliverable: Working POC + Evaluation Report.

POC demonstrates measurable business value; go/no-go decision made by the project sponsor.
ERP / CRM Integration Deployed
AI Agent Deployment Deployed
Human-in-the-loop Active
Documentation & SLA Planned
Phase 05

Production deployment

The validated POC moves to the production environment. We integrate the solution with ERP, CRM and email, implement Human-in-the-loop mechanisms and a monitoring dashboard. We hand over documentation and a maintenance plan — the organisation becomes self-sufficient. Deliverable: MVP / V1.0.

System running in production for min. 30 days without incidents; documentation and maintenance plan handed over to the Client's IT team.

Six principles
that govern every phase

Compass AI combines business ambition with security discipline — these principles apply regardless of the project stage.

Security first Data governance and classification — before the first deployment.
Value over technology We measure by tangible outcomes: ROI, savings, error reduction.
Human-in-the-loop AI handles repetitive work — humans control the key decisions.
Eliminating Shadow AI We replace unauthorised tools with secure ones — we don't ban them.
Stage-gate approach We unlock the next phase only after defined outcomes are achieved.
Client-side competence We transfer knowledge — the organisation becomes self-sufficient.

Decision gate model

Each phase ends with a review where the project sponsor and the Xelto team verify the exit criteria and make an informed decision. No solution reaches production without first proving its value and compliance.

Proceed
Exit criteria met — the next phase is unlocked as planned.
Revise
Results need refinement — the phase is repeated or supplemented before moving on.
Pause
Change in business context or priorities — the project is stopped without losing the investment made.

Key roles

Compass AI engages both sides — Xelto guides through the phases, the Client builds internal competence and makes the gate decisions.

The decision-maker on the management side. Owner of the business objectives, budget and decision gates — they approve the move to the next phase, a correction or a project pause.
Change leaders identified in Phase 2, one per key department. They bridge the Xelto implementation team and end users, translate technical decisions into business language and build an internal AI culture.
AI strategy and oversight consultants, AI/ML engineers, solution architects and business analysts. They guide the organisation through all five phases — from diagnosis to the handover of the production solution.
Guardians of compliance, security and integration with the organisation's systems. Involved from Phase 3 — they co-create the AI usage policy, the data classification matrix and approve every new tool through the AI Approval Workflow.

4 processes your AI agent
handles instead of your team

Manually re-keying orders from unstructured PDF files sent by distributors wastes enormous time and creates costly logistics errors. The situation is further complicated by sudden stock shortages and orders that don't align with packaging units (UoM/MOQ).

100%
orders aligned to logistics units
2 min
stock shortage resolution time
0%
ERP index errors
Read moreRead more

Claims arrive chaotically — descriptions in emails, photos from production floors, scanned delivery notes. Customer Service loses hours to manual triage and checking batch histories.

45 sec
claim registration time (was 30 min)
24/7
claims handling coverage
100%
traceability — defective batch detection
Read moreRead more

Bulk payments from large retail chains without invoice numbers are an accounting nightmare. AR teams lose days manually opening emails, analysing Excel specifications and cleaning up penny-level payment discrepancies.

92%
AR lines closed without touching the keyboard
−4 days
DSO reduction
−75%
finance team workload
Read moreRead more

Traditional B2B debt collection ties up language resources and risks damaging relationships with long-standing trading partners if actions are taken too aggressively.

−22%
DSO reduction
+65%
soft-collection success rate
Global
collections in any language
Read moreRead more

Ready for a safe
AI deployment in your company?

Schedule a consultation with a Xelto AI expert. We'll show you how to guide your organisation through Compass AI — step by step, risk-free.