Celonis Labs plays a huge role in that mission. This “collective of innovators, researchers and concept designers” is constantly imagining and inventing new ways for the world to work.
Beyond, the recently released Celonis Labs journal, is our chance to shine a light on their efforts. It’s also a way we highlight the latest analysis, research and Process Intelligence projects from our community to gamechangers - those who know processes are our greatest lever for value and fastest lever for change.
In this two-part series, I take a look at some of the most important findings in the Journal.
And how could we not start with AI! As our Applied AI Lead Alexander Hill explains in the Journal there is no secret that AI has the potential to transform how enterprises run their processes enabling better decision-making, leading to improvements in revenue, cost efficiency, and sustainability.
However, despite the rapid adoption of AI tools such as ChatGPT among consumers, many enterprises struggle to see a meaningful return on their AI investments.
To unlock AI’s full potential, companies must overcome critical challenges such as:
1. Prioritising the right AI use cases
Alexander says that many businesses start their AI journey with broad ideas from brainstorming sessions. This approach can result in investing in areas with little to no return while neglecting those that could yield significant benefits. He said companies need a structured, data-driven method to effectively prioritise AI initiatives, ensuring resources are allocated to projects with the highest potential business impact.
2. Providing AI with the relevant business context
For an AI to be fully reliable and operational, says Alexander, it needs to understand a process as humans do. However, most AI systems have limited access to the often scattered and unstructured data and lack insights into unique knowledge, such as company-specific KPI definitions or industry benchmarks. This limits their ability to make informed decisions and thus generate value. Businesses must provide AI with both a unified data model and relevant business knowledge.
3. Measuring AI’s impact with outcome-oriented metrics
AI solutions are often evaluated using execution-based metrics such as frequency of runs, execution time, and successful completion rates. However, says Alexander, these metrics don’t tell the whole story. Instead, AI should also be measured using outcome-based KPIs such as its impact on on-time delivery or customer satisfaction.
Overcoming Enterprise AI Challenges with Celonis
“Overcoming these challenges and harnessing AI’s full potential in the enterprise requires more than just access to the latest AI models. It requires a strong Process Intelligence foundation,” says Alexander.
The Celonis Process Intelligence Graph provides both the necessary data (e.g. order management process data from SAP & Salesforce) as well as the necessary context (e.g. a company’s unique on-time delivery definition and related industry benchmarks) to provide AI with deep contextual understanding.
A global manufacturing company used the Process Intelligence Graph to tackle these challenges and built an AI agent to handle credit blocks. After implementing their order management process in Celonis, the company analysed several improvement opportunities. They discovered that credit blocks were a significant issue, requiring considerable manual effort and causing delivery delays due to lengthy resolution times.
Using data driven insights, they prioritised automating credit block handling, as it offered the highest return on investment. The manufacturing company wanted to build a Large Language Model-powered agent to emulate human credit management experts, enabling nuanced decision-making beyond simple rule-based automation.
The agent was built on the Process Intelligence Graph, leveraging needed business context, such as detailed order information, and process mining-related metrics, such as a customer’s on-time payment history. This allowed the Agent to closely mirror human experts, deciding to maintain or automatically release credit blocks.
The main goals for the AI agent were to reduce the time required to handle credit blocks and improve the on-time delivery rate through quicker resolutions. The company created a monitoring dashboard to actively track the agent’s actions and its impact on these outcome-based metrics. This approach allowed them to measure the agent’s effectiveness based on tangible business results.
The implementation of this AI agent significantly reduced the effort and time required to review and release credit blocks and positively impacted on-time deliveries. With the initial success of this agent, the company has begun working on additional AI agents in other areas. All these initiatives are supported and accelerated by the Process Intelligence Graph, providing a consistent foundation for AI development across the organisation.
By overcoming common AI adoption challenges through Process Intelligence, companies can unlock AI’s full potential. Providing comprehensive data and context not only enhances AI reliability but also ensures that investments lead to meaningful business outcomes.
Discover the latest insights, innovations and expert perspectives from Celonis and leading process professionals and researchers as they explore how Process Intelligence is driving AI and creating a world where processes work. Download the Beyond Journal here.
Look for the second instalment of Pascal’s take on the Beyond Journal: How intelligent will the future be?
Pascal Coubard is VP Sales APAC at Celonis, the world’s leader in process mining and process intelligence.
