Unlock radical speed-to-market with agentic execution powered by Celonis Process Intelligence

Summary

The Celonis Process Intelligence Graph acts as a 'Living Digital Twin' by stitching together event data from siloed enterprise systems across R&D, sourcing, manufacturing, sales, and distribution, giving Databricks AI Agents the end-to-end process context they need to make decisions that account for downstream consequences.
A Databricks AI Agent powered by Celonis process data can autonomously identify root causes — such as a 40% sales order rejection rate putting $29 million at risk — execute corrections on master data, and proactively alert teams to risks before a product launch.
A global consumer goods company deployed this Celonis and Databricks combination and accelerated time-to-market by 17 weeks, saved 2,000 hours of manual work per year, and reduced the risk of revenue loss by $50 million.

Unlock radical speed-to-market with agentic execution powered by Celonis Process Intelligence

Companies struggle to keep pace with shifting consumer demand and strict regulatory mandates because their go-to-market process is bogged down by isolated processes, disparate data, and limited visibility. This lack of transparency across R&D, sourcing, manufacturing, sales, and distribution leads to missed seasonal windows and penalties, putting millions in revenue at risk. The Celonis Process Intelligence Graph acts as a "Living Digital Twin" to fuel Databricks AI Agents, providing the company-specific context they need to operate autonomously and orchestrate the end-to-end product development and launch journey. By agentifying the go-to-market process, organizations are replacing fragmented manual coordination with autonomous execution to achieve radical speed-to-market without sacrificing quality or compliance. A global consumer leader leveraged Celonis to accelerate time-to-market by 17 weeks, save 2,000 hours in manual work per year, and reduce the risk of revenue loss by $50M.
Talk By: Lauren Willette, Senior Value Engineer, Celonis ;

Chapters

FAQs

What is the Celonis Process Intelligence Graph?

The Celonis Process Intelligence Graph is a 'Living Digital Twin' of an organization's operations, built by connecting event data from the disparate systems used across R&D, supply chain, manufacturing, sales, and regulatory functions. It gives Databricks AI Agents company-specific context so they understand how a decision in one system cascades into delays or costs in another — for example, knowing that approving a product specification will delay sourcing by six weeks. Without this process intelligence, agents approve changes without visibility into the downstream consequences.

Why is new product innovation so slow in large enterprises?

New product innovation fails to move quickly because each function operates in its own system — your R&D team uses one platform, supply chain uses another, and manufacturing uses a third — and no single view connects them. This video explains that the resulting fragmentation means product attributes can be populated correctly in one system but inconsistently across the ERP, PLM, and supply chain platforms, causing sales order rejections, rework, penalties, and missed seasonal windows. The Celonis Process Intelligence Graph addresses this by providing a unified, stitched view of the end-to-end process.

How does the Databricks AI Agent fix master data problems autonomously?

As demonstrated in this video, a Databricks AI Agent reads the Celonis Process Intelligence Graph to identify a root cause — in the example shown, a 40% sales order rejection rate with $29 million at risk — and then executes corrections directly on master data across the relevant enterprise systems. The agent also sends proactive alerts to responsible teams when it detects a risk condition that could affect an upcoming product launch, replacing the manual coordination that would otherwise require human intervention across multiple departments. This loop of detection, execution, and notification is what the video describes as 'agentifying' the go-to-market process.

What results did a real company achieve using Celonis and Databricks together?

According to this video, a global consumer goods leader that deployed the Celonis and Databricks combination accelerated its time-to-market by 17 weeks, eliminated 2,000 hours of manual work per year, and reduced the risk of revenue loss by $50 million. These outcomes came from replacing fragmented, manual coordination across enterprise systems with autonomous agentic execution driven by process intelligence. The before-and-after results dashboard is shown in the video to illustrate the full impact of the deployment.

Full transcript

[00:08] So before I dive in, I want to tell you briefly about Celonis. We are the pioneer and leader in process intelligence. We work with thousands of customers worldwide and we're supported by a vast network of partners. Some of the largest and most complex
[00:20] companies in the world trust us to deliver process intelligence to their humans and agents. And here's why process intelligence matters. If we think about an example in new
[00:32] product innovation, we can think about deploying an agent that helps you quickly approve product specifications. So it automatically approves that product product spec. But what it can't see is how approving that product spec actually impacts you
[00:44] downstream and delays sourcing by 6 weeks. Now let that same agent run with process intelligence and the understanding of what happens when you actually change that product spec and that ends up
[00:57] delaying that sourcing. It can make a smarter, more informed decision and truly outweigh the different costs and benefits that happen with making a change. And that's the difference that Celonis makes.
[01:10] So what's the core problem? Why is the new product innovation process so challenging? New product innovation is complex and enterprise-wide touching R&D, sourcing, manufacturing, sales, distribution,
[01:22] marketing, and regulatory. But each of these functions operates in silos. Your R&D team uses one system, your supply chain uses another, manufacturing has its own tools, and so does supply
[01:34] chain. And across all these systems, your CFO, COO, CIO all have different objectives and different data. The result? A fragmented process, data locked in disparate systems, neither you
[01:46] or your AI agent truly understand what's happening in the end-to-end process to identify where to improve that process and where to deploy AI.
[02:00] So what happens when we try to execute complex enterprise-wide processes in a competitive market at the speed we need to? Everything has to align perfectly. You need three critical things to happen flawlessly. Accuracy, the first one.
[02:13] You need product attributes to be populated in systems correctly. Consistency, the second one. Once you have those product attributes populated in one system, are they populated in the rest of your systems? Across your ERP, your PLM
[02:27] system, your supply chain system. What happens when sourcing pulls one product number and manufacturing pulls another? Delays, rework, and penalties. The third one is speed. Can you move
[02:39] through this process fast enough to capture that seasonal window or to hit that regulatory deadline? When you fail in any one of these, the consequences compound. Engineering changes create delays in a production,
[02:51] material lead times get out of sync. You end up writing off inventory because you missed the window or you're paying penalties for regulatory non-compliance.
[03:03] So, what's the solution? You need what we call living digital twin of new product innovation. Not a static model, not a reporting dashboard. A living, breathing representation of your actual process, updated in real time, connected to all of your systems,
[03:16] and smart enough to guide autonomous agents through execution. And that's where Celonis comes in. We create a process intelligence graph. This is fundamentally different than
[03:29] traditional data warehousing or BI. It's not just a place to store your data, it's not just a place where you can make charts, graphs, and reports. We map the actual flow of your process across all of your systems. We see which
[03:41] steps matter, how they're connected, where the bottlenecks are. And any system that creates event data can be pulled in. So, as you see at the bottom of the screen, we can take unstructured data, structured data, and semi-structured
[03:54] data. We upload that data to our process intelligence graph, and we stitch together a semantic, event-driven representation of how your processes run. We open this to agents and humans alike
[04:07] through our MCP server. With that ability, we then become the foundation for intelligence across your enterprise. We can analyze it to find efficiencies, bottlenecks, and opportunities.
[04:20] We can identify where manual work is happening, where delays are occurring, and crucially, we can feed this intelligence to AI and humans.
[04:32] Now, here's where Databricks comes in. Databricks gives you the infrastructure, the lakehouse, where all your data lives. You have MLflow for model management. You have AI gateway for governance. You have Unity Catalog for security. And that's the foundation.
[04:45] But, traditional data platforms optimize for insight. They're great at answering what happens. But, they struggle with what should we do next? And actually doing that.
[04:57] And that's where Celonis changes the game. Databricks AI agents can tap into the process intelligence graph. They don't have to guess how your process works. They see it. They understand the dependencies. They know which steps are
[05:09] critical, how they impact each other, and which steps need to be optimized. They can make autonomous decisions with business context, and not just statistical models.
[05:24] So, in a world where consumer demand shifts overnight, and regulatory mandates are constantly changing, speed is the key to survival. The companies that will win in this market are the ones who can orchestrate their entire go-to-market process autonomously
[05:37] with the power of AI. So, what does Celonis and Databricks actually deliver? I'll break down three ways that we create value. We first eliminate manual coordination.
[05:49] We know that your go-to-market process is bogged down by manual hand-offs. R&D sends specs to manufacturing, manufacturing sends forecasts to supply chain, to the supply chain sends materials to the plant, and every one of
[06:01] those hand-offs is a potential failure point. Every step takes time, and with autonomous execution, these steps happen instantly and intelligently. Second, the process intelligence graph
[06:14] stitches together where your fragmented data into one unified layer, giving you end-to-end visibility into where your process breaks down, and that's exactly where you deploy AI for maximum impact.
[06:26] Third, we compress time to market. Every month of delay costs you. Every regulatory penalty, every regulatory miss creates penalties, and every write-off inventory write-off hits the bottom line.
[06:39] When you replace fragmentation with orchestration, time compresses. One of our customers quantified this by accelerating their process and reducing their exposure to revenue loss by 50
[06:51] million. So, what I want to do now is take you through a demo of how this actually looks in Celonis. I'll show you how we ingest data from multiple sources, how we create the process intelligence graph, what
[07:05] insights emerge and what you can do with those insights, how Data Bricks agents leverage these insights to operate the process, and how we can monitor the impact that these AI agents have on your process to ensure that they're providing you that ROI.
[07:18] So, now we're in the Celonis platform. What I want to start by is showing you the process intelligence graph that we can create. We start by pulling in data from any of your source systems, SAP, Oracle, Salesforce, PLM systems,
[07:32] Data Bricks, and we upload that event data to show how your objects run throughout your processes. So, an object can be a sales order, a purchase order, a production order, a formula, a delivery. And we can connect
[07:46] all of those to show you how they actually interact in your real process. The next thing we do is layer layer in business context. We can start by adding in KPIs. So,
[07:58] what's your re- what's your rework cost? How do you define on-time delivery? These are things the AI agent needs to know to make impactful autonomous decisions. We then can layer in any AI capabilities
[08:10] you already have. Maybe you already have a material delay predictor in place. The AI agent can work alongside that and pull insights from that AI that you already have in place today. The next thing we do is bring in
[08:22] different business rules. So, what do you need to do in order to actually approve the artwork that you're making for this new product? And next, we can bring in your process flows. We can bring in how you have
[08:34] designed your new product innovation process to actually work. So, what you're left with is a true representation of how your business should be running and how it's actually running.
[08:46] Everything you see from here on out, every analysis, every chatbot, every agent is fueled by this process intelligence graph. So, now we're in a good position to actually understand the opportunities within new product innovation
[09:00] and deploy AI in an impactful way. And this is what it looks like on the front end to your end users. What you're seeing now is a front-facing view of your enterprise.
[09:13] We're going to do new product innovation as the example, but this can be done across all industries and all different processes. So, we'll start upstream with new product innovation, and we're pulling in data from the PLM system to capture the
[09:26] actual process that's running. So, we start with a product specification activity, and we can see that there is 50 objects that it's considering. So, 50 times that this process has happened.
[09:38] And we can also change that to throughput time. So, we can see how long it's taking to actually create a product specification and then approve it. And what I can do is start pulling in
[09:51] the other subsequent processes that happen after this product specification is approved. I can pull in the formula process, the packaging artwork process, the packaging specification process. And
[10:03] now we have a complete view of what's happening upstream to actually tie all these components into a bill of materials, into a product that we can actually sell. So, what I'll do is now start to look
[10:15] downstream. I'll add in the bill of materials and move downstream so we can see what happens next. So, once the bill of materials is
[10:27] approved, what we can do is now start setting this up for our retailers or our customers to order. So, we send the master data over to our retailers for them to them to set up in their systems and start ordering our new product.
[10:40] And lastly, I can see this all the way through to sales orders. So, all the way from R&D creating specifications to customers actually purchasing the new product.
[10:52] And what I see here is that 40% of our sales orders are being rejected. So again, this is the what. This is what a lot of times we're left with. 40% of sales orders are rejected, but
[11:04] what we need to know is why. So, we can start looking upstream in the process to say what's causing that 40% rejection rate. And if I scroll upstream in the process, I can see that only 70% of times the old
[11:17] product code is being deactivated. That means that if we think back to the presentation, we're not being accurate or consistent across our new product innovation process, and that's costing us in speed, and that's costing us in
[11:29] revenue. And while you might be thinking, "This is really great insight." you might also be thinking, "What do I do with this insight now?" And that's where Celonis becomes really powerful. It's continuously monitoring what it's seeing
[11:42] in your process intelligence graph to pull out the top opportunities for you to go after. So, if you look over on the other hand other side of the screen, you can see that Celonis is identifying these top opportunities. And
[11:55] revenue leakage is one of them. It's saying that what you're seeing in the process explorer is true and we've quantified this revenue leakage. So, let's look at the impact that that missed master data is having on our
[12:07] process. So, what we'll look at once it loads is leading up to a product being launched, what revenue is at risk and why. We can see in this graph that
[12:19] leading up to 2 weeks of the product being launched, we have some supply delays, incomplete data, missed approvals, and we're quantifying the amount of revenue that is at risk. So, we see that misaligned
[12:31] system go live, which we saw in the process explorer, is actually putting 29 million at risk leading up to our product launch. We can look even further and say what product lines are falling into this trap
[12:44] most often and where do we need to improve? So, at this point we've gone from looking at a 10,000-ft view of the process to a very clear insight with actual value tied to it. And this helps us realize where we should be deploying
[12:57] AI for maximum impact. We've looked at an extremely complex enterprise-wide process, one that requires deep knowledge, and we've identified that there's a 29 million in revenue impact we can have if we apply
[13:11] AI in an autonomous way to reduce these inconsistencies and inaccuracies we see in master data. So, what we can do is open up this process intelligence to your teams into
[13:23] data bricks to then take this process intelligence and run the process autonomously. So, what we can do today is look at a master data integrity agent that will act as an autonomous data steward across
[13:35] all of your systems, making sure that master data is entered correctly, master data is carried across all systems, and any approvals and urgent work that needs to be done is automatically surfaced to
[13:47] your team. So, where it gets powerful is that anybody in your organization can have access to this data as you prefer it. The team is no longer sifting through
[13:59] emails and systems and spreadsheets to identify where you're at in a new product launch. You can simply type that into Databricks, it calls on the process intelligence, and it tells you exactly where you are within that process,
[14:12] and what's left and what needs to be actioned on. But, where it gets even more powerful is that a Databricks agent is continuously monitoring that data for you. So, actually don't have to use that
[14:24] chatbot, the Databricks agent will proactively reach out to you when it sees a risk emerging in the process. So, for the example we're looking at today, it's reaching out to me and telling me that a product code has not
[14:36] been successfully communicated to our customers. We need to take action and communicate that so that way we don't end up with rejected sales orders, and we don't miss our revenue.
[14:50] And once that agent is deployed, we want to make sure it's successful. So, what we can do there is actually start to look at how that agent is interacting in our process. So, now we're brought back to the original view where we saw the bill of
[15:02] materials being created. And what I can do now is actually show the agent and the action it's taking in the process. So, we can see that once the bill of materials is approved, the agent is coming in and notifying the planners
[15:15] that there's new planning master data. It's creating that form for retailers to actually send off the submission of new master data. And then it's taking that crucial step
[15:27] of verifying that the master data has been updated by the retailers. And if not, it'll ping you that autonomous autonomous- and tell you that action needs to be taken. And while it's running, we can look at
[15:40] how it's impacting our key objectives in the new product innovation process. We can see before and after the agent goes live to see if we're actually meeting our goals. So here we see that after goes live, we actually are increasing
[15:53] our revenue to where we thought it would be with this product launch. We are low- lowering the cost and taking out that rework. The agent is making a smoother process for us. And for cycle time, we see that go-to-market cycle time drastically
[16:06] increase. And what I just showed you is one use case, one way that you can identify and accelerate your new product innovation process. But if you're familiar with new product
[16:18] innovation, you know that there's so many handoffs and bottlenecks that can be solved. So there's so many areas that AI can be strategically deployed through Databricks and Celonis to take impactful action across your
[16:31] processes. And this isn't theoretical. Leading companies in CPG, pharmaceuticals, manufacturing, and food and beverage are already using Celonis process intelligence to fuel AI agents and
[16:44] accelerate their processes. One of our global consumer leaders accelerated time to market by 17 weeks, and they're still going. They eliminated 2,000 hours of manual
[16:56] work per year that it took to bring new products to market. And they reduced revenue loss risk by $50 million. The outcome is that they didn't just improve on inefficiency, they transformed how they competed in the
[17:08] market. So in a world where consumer demand shifts overnight and regulatory mandates are constantly changing, speed of survival. The companies that win are the ones that
[17:20] can orchestrate their entire go-to-market process with the intelligence of AI. Celonis provides the process intelligence and Databricks provides the autonomous execution.
[17:32] Together, they give you radical speed to market without sacrificing quality or cost. Thank you.

Learn more about the Databricks Data and AI platform.

The information provided herein is for general informational purposes only and may not reflect the most current product capabilities or configurations.