AI Strategy Transforms AP Automation Platform: 90% Touchless Invoice Processing
Strategic roadmap delivers natural language Copilot, intelligent matching with 250+ patterns, and industry-leading automation rates
Executive summary
A leading global Procure-to-Pay (P2P) automation provider serving enterprise clients across energy, renewables, and other industries engaged Zartis to develop a comprehensive AI strategy that would drive product innovation and competitive differentiation.
The problem
With millions of invoices flowing through their platform, the company saw untapped AI potential for intelligent matching, automated coding, and natural language insights. However, internal teams lacked direction on where to invest and how to implement AI capabilities that would deliver measurable ROI.
The solution
Zartis AI consultants conducted strategic assessment workshops to identify high-impact use cases, evaluate technical feasibility, and create a phased implementation roadmap aligned with business priorities.
The outcomes
- AI-powered Copilot launched: Natural language query interface delivering instant AP insights
- Intelligent Smart Matching deployed: 250+ learned patterns achieving 90% straight-through processing for PO invoices
- Smart Coding & Routing automated: 89% faster non-PO invoice processing through AI-driven automation
- Company-wide AI strategy: A detailed transformation roadmap executed from workshop to production
About the client
The client is an enterprise software provider delivering cloud-based Procure-to-Pay automation solutions that handle eProcurement, invoice processing, vendor management, and payments for hundreds of enterprise customers, processing millions of invoices annually
This client’s platform serves enterprises with complex procurement needs, including energy and renewables clients with IoT-enabled smart metre billing and multi-site organisations requiring centralised AP processing. Whilst their platform already automated significant workflows, manual intervention remained necessary for exception handling, non-PO invoice coding, and complex matching scenarios, creating competitive vulnerability.
The problem
The platform processed massive invoice volumes with sophisticated matching logic, but exceptions and edge cases still required human intervention. Finance teams spent significant time manually resolving PO-to-invoice discrepancies, coding non-PO invoices, and searching historical data.
Market pressure was intensifying
- Competitors marketed AI-powered features positioning "intelligent automation"
- Enterprise RFPs explicitly requested AI capabilities
- Product roadmap debates paralysed decision-making across departments
The cost of inaction became unaffordable
- Competitive disadvantage: Losing deals to vendors with proven AI automation
- Customer churn risk: Clients exploring competitors with better performance
- Resource inefficiency: Engineering building AI experiments without strategic alignment
- Missed revenue: Premium AI features could command higher ACVs, but no clarity on what to build
Internal attempts failed
- Engineering team lacked production AI/ML experience
- No framework existed to prioritise competing departmental visions
- Uncertainty on build vs. buy decisions
- No clear path from prototype to production deployment
Why they chose Zartis
We were brought on to the project for our dual expertise: deep AI technical capability combined with enterprise finance software knowledge. Unlike pure strategy consultancies that deliver decks or offshore teams that lack business context, we could translate AI opportunities into an actionable roadmap their product and engineering teams could actually execute.
how we worked together
The Zartis Approach
6-week engagement structure from discovery to full strategy
Discovery & AS-IS assessment (2 weeks)
We began with deep technical workshops to understand their current matching algorithms, data pipelines, and processing workflows. These sessions ran in parallel with product strategy discussions exploring roadmap priorities and competitive positioning. We conducted customer success interviews to identify pain points directly from their enterprise clients, while simultaneously performing a comprehensive data landscape analysis examining invoice volumes, data quality, current match rates, and exception handling patterns.
Use case prioritisation (2 weeks)
Following discovery, we facilitated cross-functional brainstorming sessions to generate AI opportunities across the platform. Each potential use case underwent rigorous technical feasibility analysis—evaluating data availability, model complexity, and integration requirements—alongside business impact assessment measuring efficiency gains, cost reduction potential, and revenue opportunities. We applied a prioritisation framework that categorized opportunities into Quick Wins (immediate value), Strategic Capabilities (competitive differentiation), and Long-term Bets (future innovation).
Strategy & roadmap edvelopment (2 weeks)
The final phase delivered a comprehensive TO-BE architecture design for their AI-enabled platform, supported by a phased 18-month implementation roadmap. We provided technology stack recommendations covering ML frameworks and serving infrastructure, along with build vs. buy analysis for specific capabilities. The deliverable included detailed resource planning outlining required team composition and skill requirements to execute the transformation successfully.
Our approach
What we delivered
Current State Assessment Report
Data landscape mapping: Invoice volumes, formats, quality; PO data structures
Technical architecture: Current matching logic, rules engines, integration points
Performance baseline: Automation rates, processing times, accuracy metrics
Gap analysis: Infrastructure, skills, data, and processes needed for AI
Prioritised AI Use Case Portfolio
8 high-impact opportunities ranked by business value and feasibility:
Tier 1: Quick wins
- Natural language query interface (Copilot)
- ML-enhanced PO-to-invoice matching
- Automated GL account and cost centre assignment
Tier 2: Strategic capabilities
- Predictive exception detection
- Intelligent routing optimisation
- Anomaly detection for unusual patterns
Tier 3: Advanced transformation
- Conversational AI support
- Supplier risk scoring
Future AI Strategy
Product positioning: Transformation from "workflow automation" to "intelligent automation"
Competitive differentiation through AI capabilities
Data strategy: Infrastructure investments needed
Team enablement: Hiring plan, upskilling roadmap
Success metrics: Measuring AI impact on automation rates, speed, satisfaction
Technology Stack & Build vs. Buy Recommendations
Buy
LLM APIs for natural language (OpenAI/Anthropic) - commoditised capability
Build
Matching algorithms with proprietary ML - core competitive differentiator
Enhance
Existing rule-based logic with ML augmentation
Resource Planning
Team composition: ML engineers, MLOps engineer, data engineer
Infrastructure costs: Cloud ML platform, LLM APIs, training compute
Timeline with clear milestones and dependencies
Our approach
Key strategic decisions
Quick wins built momentum
We identified 3 use cases delivering measurable value in 3-6 months with existing data: Copilot, enhanced matching, and automated coding. This created organisational buy-in before requesting larger investments.
Pragmatic technology choices
Cloud-based ML platforms and managed services over custom infrastructure accelerated time-to-value, allowing focus on use cases rather than platform engineering.
Cross-functional alignment
Balanced prioritisation framework
Use cases scored across business impact, technical feasibility, and strategic alignment—preventing “shiny object syndrome.”
Honest readiness assessment
The AS-IS report identified specific data quality issues, infrastructure gaps, and skill shortages that would block success—addressing these became part of the roadmap.
Scalable adoption strategy
Phased implementation roadmap and proper resource planning from technological investment to team composition, and skill requirements for full adoption.
the results
From strategy to production:
Real AI capabilities launched
Following the strategic engagement, the client executed the roadmap and delivered measurable AI capabilities:
AI-Powered Copilot
Natural language query interface embedded in the platform, allowing users to interact with invoice data conversationally. Users can identify suppliers with highest volumes, retrieve invoice histories, calculate trends without manual filtering or complex reports.
Impact:
Instant insights, reduced training burden, faster decision-making for AP managers.
Intelligent Smart Matching
AI-powered engine learning from millions of real-world invoices with 250+ unique match patterns, intelligent tolerance logic, and confidence-based matching that routes only true exceptions to humans.
Impact:
Straight-through processing for PO invoices (up from ~70% baseline), reduced manual exception handling, faster month-end close.
Smart Coding & Routing
AI-driven automation coding non-PO invoices to correct GL accounts and routing to appropriate approvers based on learned patterns, with continuous improvement from user corrections.
Impact:
Faster processing of non-PO invoices, reduced coding errors, PO-level automation extended to historically manual workflows.
The results
Measurable business outcomes
Increased processing efficiency
90% touchless processing
for PO invoices—industry-leading automation rate.
89% faster processing
Faster non-PO invoice processing. Dramatic reduction in manual effort.
Better competitive positioning
New AI capabilities
AI is now central to sales messaging and competitive differentiation.
Higher win rates
in enterprise RFPs where AI automation is evaluation criteria.
Increased revenue and ROI
Premium pricing
justified through demonstrable automation performance.
Upselling
Revenue expansion from existing customers adopting new AI features.
Higher operational efficiency
Reduced support burden
through Copilot answering common questions.
Faster onboarding
Seamless onboarding process for new users.
Higher productivity
Improved AP team productivity at client organisations.