Unlock the full value of your data with IBM watsonx and Celonis

Historically, extracting the valuable business understanding bound into unstructured data at scale – let alone folding it into existing analytical models – was either technically impossible or simply impractical. Too manual, too time consuming, too expensive.

Now, a combination of process mining and generative artificial intelligence (GenAI) can hand businesses the keys to the vault. Organizations have the opportunity to leverage all of their data – including the unstructured 80% that was previously beyond their reach – to drive value creation.

That was a principal takeaway from a recent webinar at which Celonis titanium partner IBM showcased practical applications of unstructured data delivered into the Celonis platform by its own GenAI foundational model – watsonx™.

https://delivery-p141552-e1488202.adobeaemcloud.com/adobe/assets/urn:aaid:aem:b1906160-474b-4aec-a575-1ddca3d8f1cd/as/ibm_watsonx_and_celonis_2024.png

The IBM team demonstrated how GenAI enables organizations to access, analyze and activate customer, colleague and supplier interactions, compliance and operational frameworks, and qualitative performance data points held within unstructured data.

"Most business transactions include some kind of unstructured communications," said Elias Hagemann, IBM Celonis Lead for the DACH region. "These messages and documents are crucial for business. However, traditionally, they're not part of process mining initiatives because unstructured data is so hard to work with. But now, with GenAI it has become possible."

How the new foundational models changed the game

The IBM watsonx platform (watsonx.ai™, watsonx.data™, and watsonx.governance™) acts as a copilot for the enterprise, providing both the tools to deploy GenAI models and the security and data privacy required by enterprise AI.

The IBM team prefaced the use case discussion with a review of the advantages and appeal of newer foundational GenAI models like watsonx when it comes to conducting cognitive tasks like processing unstructured data.

"Traditional AI models were siloed, they required a lot of task-specific training – often human supervised training," explained Hagemann. "But the new foundation models, on the other hand, come pre-trained with vast amounts of different external data. So with more out-of-the-box intelligence, very little training afterwards is necessary. They are more adaptable, they're better at multitasking, they're just exponentially more capable. This means lower upfront costs because we don’t need to train the model beyond minor prompt engineering. We have faster deployment and we have better accuracy."

GenAI models like watsonx have many capabilities that help process unstructured data and make it available for use in the Celonis platform, in particular:

• Classification: Reading and classifying written inputs, such as using sentiment analysis to sort and segment the severity of customer complaints.
• Insight extraction: Analyzing unstructured text content to identify specialist insights according to set criteria (such as user research findings).
• Named entity recognition: Pinpointing and extracting the pertinent, essential information from unstructured text.
• Retrieval-augmented generation: Creating a chatbot or user Q&A feature based on documentation or dynamic content.
• Summarization: Creating personalized overviews of text data, summarizing and highlighting key points from multiple text sources.

Watsonx uses these capabilities to extract the required information from multiple unstructured data sources, analyze it based on customer parameters, then translate it into a format that Celonis can consume. The previously unreachable insights become part of an organization’s overall Process Intelligence data model – able to be interrogated and dashboarded, to help uncover value opportunities.

• For more details, watch the webinar.

Celonis + watsonx use cases

The IBM webinar walks the audience through a number of practical use case examples of Celonis and watsonx creating real business value from unstructured data. This includes live demonstrations of a customer touchpoint analysis and a purchase to pay (P2P) supplier assessment.

Customer touchpoint analysis

With quantitative system log data it’s straightforward to calculate KPIs like the resolution time, number of customer messages or call volumes. "But from that," said IBM Process Intelligence Consultant Johanna Kaiser, "we don't really know why the customer is contacting us or if they're having a good customer experience. But now, with GenAI, it becomes possible to automatically extract information such as customer sentiment or support ticket criticality data, and to generate summarizations or categorizations."

This enables the end user to track customer sentiment changes throughout the journey using the Celonis platform. The support ticket scenario in the webinar demonstrates how watsonx can detect the urgency and negativity from the customers' language ("I can’t work without this being fixed" for example). The GenAI algorithm can categorize the calls and trigger prioritized responses to higher urgency inputs.

This type of touchpoint analysis can create value in three key areas:

1. Identify and eliminate drivers of negative customer experience: By understanding why your customers are contacting you it’s possible to review and eliminate the root causes of unwanted requests. IBM estimates this could drive a 10–20% reduction in customer service requests.
2. Discover automation opportunities: With greater transparency on the types, timings and volumes of customer requests you can establish which automation initiatives (such as RPA) will improve customer journeys the most. IBM estimates this could lead to 25% higher ROI on automation initiatives.
3. Increase customer satisfaction: Being able to measure sentiment from all customer interactions means you understand the levers to increase their satisfaction and the processes to optimize. IBM estimates this could increase Net Promoter Score (NPS) by 20–30% and decrease customer churn by 10–20%.

P2P supplier assessment

The webinar’s second use case involves a live demo of supplier assessments within a purchase to pay function. Structured data captured performance metrics in Celonis including on-time delivery, partial delivery rate, lead times, delivery date changes, product price increases, plus product quality and returns.

"So from that I could try to choose a vendor," said Kaiser, "but usually I'd go back to my contracts and find out more about the vendors like inflation-bound pricing, special costs, clauses or payment terms. And now our GenAI superpower means I really don't need to do that. Watsonx automatically extracts the contractual information for easy access within the Celonis platform."

And by combining structured and unstructured data insights within a single platform, vendor selection is not only simpler but also based upon a far broader data driven rationale – saving time and money in the long and short term.

Other transformative use cases

Other transformative use cases are highlighted during the webinar that combine structured data and unstructured data using the watsonx and Celonis interaction, including:

Four-way mismatch in accounts payable / procurement: By extracting unstructured data from commercial contracts, watsonx enables Celonis to proactively spot and flag discrepancies in POs, invoices, and goods receipts, potentially creating a 15-25% reduction in errors and up to 40% time savings in contract comparison.

Inquiry Processing & Response Generation: Connecting processes across ServiceNow and SAP, watsonx can scan, analyze and categorize incoming unstructured inquiries (like emails) and generate automated responses grounded in Celonis Process Intelligence – the integration of detailed process data, gained from process mining, with standardized process knowledge. Estimated benefits include a 10-20% reduction in errors and a 20% productivity improvement.

Claims Management: Celonis and watsonx can improve operational efficiency by automating areas such as loan processing, document management, and compliance. The customer experience can also be improved by streamlining the process for opening an account and reducing document review times. Improvement opportunities include, claims prioritization, first contact resolution, claims reopenings, SLA adherence rate and high non-indemnity risk.

Hagemann was keen to stress that for each of these watsonx deployments there’s a fast time to value. "Each of these use cases came to life not in a matter of several months, but in a matter of weeks," he confirmed. For full details of these use cases, watch the webinar.

No limit to future applications

The combination of GenAI with Celonis Process Intelligence – delivered through its Process Intelligence Graph – can transform qualitative data to quantitative, inaccessible insights to value-creating processes. As Alex Rinke, co-founder and Co-CEO of Celonis put it: "The Process Intelligence Graph and watsonx are a great combination. Together, they enable customers to harness the power of AI and their process data to improve efficiency, enhance business performance and drive meaningful results in weeks not years."

What’s more, the sample watsonx use cases cited above can be applied to new Celonis implementations or be used to enrich and enhance existing deployments. Either way, they represent just a drop in a colossal data lake. Given the volume and variety of unstructured business data, there are practically no limits to the ways you can tailor use cases to individual process pain points or business needs. "There’s probably hundreds of thousands of use cases out there from which to create value," concluded Hagemann.

• For a step-by-step guide to unlocking the full value of your data with GenAI and Celonis, watch the webinar now.