AI-based Finance Transformation Initiative

This paper is the AI transformation plan of Finance processes for one of leading European companies engaged in manufacturing, sales and service of heating, ventilation and air conditioning systems (HVAC) with its application and development centers across major geographic regions.

Since 2020 the organization has worked to bring all sites and offices on single ERP platforms. It has now provided a suitable platform and business maturity on knowledge of systems to be able to appreciate and use immense data available to the next level of

business drivers. Single ERP has helped the company to automate routine processing of invoices, payment, and managing delinquencies; however current automation has not

helped in reduction of invoice errors, collection improvement and customer churn. It has initiated system automation to the next level using AI based tools across its value streams of order to cash, source to pay, HR and financial reporting processes.

A conservative estimation of the profit realized because of implementing AI technologies and solutions is ~ 15%-25% over current plus soft returns in terms of employee retention,

less customer churn and innovation. Caution has been taken to discount the uncertainty of benefits, computing ROI based on point of time and taking all value stream projects

individually with a room of adjustment

Strategic plan

Strategic alignment

The organization of customer and supplier base is spread across geographies. To gauge the right market, the organization is leveraging Gartner AI adoption strategy framework (source Gartner 730S70_c) which consists of Vision, Value, risk and adoption and tailor this to suit our organization vision and customer centric value system

1. Transformation vision

• Process, policy and framework
  • Optimize operating model
  • Continuous improvement and lean six sigma projects
  • Process consolidation and standardization
• Success metrics
  • Accelerated cash flow of past dues – improve from ~75% to <10%
• Improved cash conversion cycle from 181.04 to 45.31
  • Days sales outstandings: from 90.55 to 22.15
  • Days inventory outstanding: from 86.12 to 29.11
  • Days payable outstanding: from 25.23 to 5.95
• People and skills
  • Hiring policy rationalization
  • Upskilling and training in new tools
  • Rebadging as necessary

2. Value: value delivery levels

• Order to cash
  • Reduce invoice to collection lifecycle
  • Standardize billing framework
  • Actionable insights into disputes
• Source to Pay
  • Supplier onboarding based on intelligent data inputs
  • Catalog penetration across globe
  • Invoice to Pay automation (AP flow)

3. Risks

• Third party /cyber risk assessment for AI driven technologies and adopt control /mitigation
• Privacy impact assessment cross systems and regions where these systems will be used and secure approvals from local regulatory bodies
• Work with regional teams to engage with employee representative bodies to secure approvals for planned AI usage and eliminate any legal barrier and adverse impact
• Draw business continuity and disaster recovery framework/plan

4. Adoption

• Identify use cases: Propose to conduct a series of workshops to assess problem hotspots and validate what are most valuable /feasible use cases with the quickest time to value across value streams.
  • Feasibility criteria include technical, internal and external factors.
  • Skills: Assess the skills needed to use AI techniques.
  • Data: Readiness of data to address select use cases and clear data management and governance requirements to power AI
  • Technology: needed in the current stage of data/skill and use cases
• AI decision framework
  • Transform into value stream level decision making by tower leads to using AI to generate data driven insights
  • Adopt trained ML models into decision and role models to allow maximum transparency, control and agility over a period
• Decision governance
  • Establish a mechanism for effective management
  • Integrate this with operating and transformation governance to enable seamless delivery and continuous process improvement
  • The Governance model will be a formal mechanism for operational and strategic alignment to ensure ongoing innovation and improvement

Assessment and Preparation

• Senior management went through a rigorous process to select the right partner for assessment tasks
• This followed a series of workshops, site visits to understand current challenges
• The partner team identified high-stake customers and suppliers and underwent a process of interviewing them
• team of business owners and partners engaged in eight weeks of workshop to prepare a blueprint of business roadmap and technology selection to achieve the end goal.

AI and Technology Selection

We are not sharing the cost of implementing this due to business confidentiality. However it is fair statement that the solutions are pre built with advanced AI features . The license costs are taken with a baseline of transactions and user base

Development and Deployment

Phased approach to Develop and deploy the new application platform across sites/ plants and offices organization

Phase 1: Platform development environment readiness to be completed by Q1 2025

• Complete all due diligence and contractual activities (Statement of work, cyber, PIA, security, enterprise architecture team clearance and approvals to onboard the systems)
• Complete the blueprinting of applications
  • To be process and system mapping
  • Integration /interface requirements
  • Data discovery /cleansing /migration and access to labeled data

Phase 2: Integration components and readiness to be completed by Q3 2025

Organization has shortlisted leading development frameworks and libraries for AI development. The Development team will be using TensorFlow and PyTorch ML model development. Additionally, AI libraries such as Scikit-Learn, Keras, and Caffe, which provide the developers with API sets to accelerate application development and deploy with flexibility to innovate as much as possible. Tools like Scikit have an easy to learn approach and low code giving them an advantage for our organization which is low on AI learning curve and looking for mature as we implement and grow. Kera offers a more intuitive interface to build complex neural networks

For managing development Organization is keen to leveraging its knowledge workforce on AzureDevops AI capabilities for requirement generation and analysis, data driven decision making, code analysis, automated testing enhancement, monitoring and feedback.

The organization will be leveraging Planview copilot for portfolio management. The internal userbase is trained to use Copilot.

Phase 3: Value stream Phase wise deployment s change management Q4 2025 – Q3 2026

The organization has developed a cluster roll-out approach supported by ADKAR model of change management. It will be using a comprehensive set of tools to provide E2E

deployment and support. Key features include

• Version control and change control to build, train and deploy
• Upgrade with each cluster if needed to take care of local features
• Ability to monitor performance and ensure we are getting reliable AI driven results

Monitoring and Evaluation

Throughout the building and deployment of these tools, addressing model drift identified by monitoring is critical component for successful deployment.

The organization is working with its implementation partner to establish a mechanism to address the change in context of data and concepts.

Conclusion

The integration of new technologies leveraging AI and ML with our core DT systems supporting business is a strategic initiative to help redesign three core E2E business processes and drive maturing organization’s performance across functions and geographies. We have strong endorsement and support from our executive leaders to navigate challenges through prioritization, communication and change management.