What's Next For Process Mining; Conversational Platforms Probe Integrating LLMs
Process mining will thrive in the AI era, aiding process optimization and automation. Convo platforms should pivot into agentic systems, integrating LLMs for dynamic industry solutions.
Beyond Spreadsheets: Why Process Mining Will Thrive In an AI-powered Age
Understanding how businesses operate has relied on manual process mapping, interviews, and educated guesswork for years. Then came process mining, a game-changer that promised to shed light on the messy reality of real-world workflows. And now… even your grandma is excited about generative AI.
So, What Exactly Is Process Mining?
Simply put, process mining is extracting knowledge from event logs generated by IT systems. Think of it as an analyst using digital footprints of a process to reconstruct a series of events. By analyzing these logs (which record every user action, timestamp, and associated data), process mining tools create a visual map of how processes are executed in real life. They reveal bottlenecks, deviations, and areas for improvement that would be invisible using traditional methods.
Why Process Mining is Back In The Spotlight
While process mining has always been valuable, the rise of AI is creating a perfect storm for its resurgence. The CEO of Celonis, the second-largest startup in Europe, explains it well in a recent Prof G podcast.
Data Explosion and Context Clarity: AI thrives on data, and organizations are drowning in it. Process mining provides the link between this raw data and insights. Process mining tools become the eyes of AI agents. They provide the crucial context and understanding of current processes, feeding data and insights into AI algorithms.
New Frontier of Automation: AI-powered automation, mainly Robotic Process Automation (RPA), is now being topped by Agentic Process Automation (APA). However, automating inefficient processes makes things worse faster. Process mining helps identify the right processes to automate, ensuring that AI is deployed strategically.
Hyper-Personalization: AI enables hyper-personalized customer experiences, but delivering on this promise requires optimized, frictionless processes. While process mining reveals the friction of customer journeys, we do not see our clients relying on related tools expensively. Because marketers and loyalty managers don't think in the same categories, process mapping and optimization are not on their agenda. We still don't know “which 50% of the marketing investment brings the return”.
The Future: a Powerful Duo
So, where do process mining tools and platforms fit into this AI-driven landscape? They're not being replaced; they're being augmented. Think of it this way: if the process mining tools are the “eyes,” AI agents become the “brains” behind process improvement. They can analyze process mining outputs, identify complex patterns, predict future bottlenecks, and even autonomously suggest and implement process improvements at some point.
Imagine an AI agent using process mining data to not only identify a process bottleneck but also to:
Predict when and how that bottleneck impacts other processes,
Recommend the best course of action to resolve the issue,
Autonomously fix implementing the best solution, like adjusting resource allocation or modifying a workflow.
Here's our prediction: as AI agents evolve, agentic platforms will offer packaged process mining and mapping capabilities combined with RPA and workflow automation.
In 2025, enterprises will be even more interested in understanding how they operate.
In the end, AI is “intelligent.” Like an intelligent human who tries to understand what's happening before solving a problem, AI systems will be more effective if they obtain additional context before passing tasks to agents to execute.
Conversational AI Platforms Are Pivoting Into an Agentic Space 🌟
With advancements in GenAI, conversational platforms are a rapidly changing area, similar to the Robotic Process Automation (RPA) we have covered a number of times. In those editions, we have expressed our views on the future of RPA companies and their chances in the agentic systems race. Well, a similar transformation is underway for conversational AI platforms.
Conversational AI platforms initially aimed to solve a fundamental problem: classifying messages. During the early days, primarily, it was a net of “if” statements, later working to understand user intent leveraging Natural Language Processing (NLP). However, genuine desire has always been an engine that perfectly grasps what the user needs as a real human being.
When ChatGPT appeared, many platform businesses rushed to adopt it as a potential replacement or an improvement for their existing conversational interfaces and chatbots. The initial implementation often involved feeding company data into prompts or utilizing features like OpenAI's Assistance file search. However, the initial "wow" effect from successfully handling 80% of cases was quickly overshadowed by an inability to address the remaining 20%.
This exact realization has been a lifeline for many incumbent players in the conversational AI space. They recognized the importance of leveraging their decade-long expertise in client-specific workflows, industry nuances, and regulations to address the final mile of complex, edge-case queries. However, these platforms often lacked the enterprise-level functionality that RPA solutions usually have.
Some of the most forward-thinking companies are trying to add agentic capabilities to their offerings. For instance, Kore.ai, Avaamo, Aisera, Moveworks, and Cognigy are all actively expanding their teams and clearly focusing on this new priority. Each of these players is fighting with the non-deterministic nature of LLMs, trying to make conversations more predictable and stateful. Some use LLMs primarily for understanding and dialogue, which is quickly becoming a commodity. Others are focusing on building bespoke solutions for specific industries. For example, Alltius.ai is crafted for financial services, and Hyro.ai is for healthcare. Meanwhile, some platforms from the previous generation, such as Leena, Haptik, or Yellow, seem struggling to find their future strategy in the new landscape and are now shrinking. 📉
Rasa's co-founder, Alan Nichol, offers a practical perspective on integrating LLMs into the conversational customer journey. He suggests using LLM/AI-based solutions primarily for dynamic, multi-variable, unstructured cases. Existing systems often provide a more reliable and preferable solution for more straightforward, deterministic scenarios.
As we saw with orthodox RPA platforms, conversational AI platforms are shifting toward agentic systems. In 2025, we will likely see multiple SaaS markets, including RPA and conversational AI, converging into a single, highly competitive market of agentic platforms. Initially, these platforms may focus on upselling to their existing customer base. However, the competition will become incredibly fierce once that initial wave disappears. 🏆
No laundry list today; apologies, we’ve been too busy building AI agents last week.