--- title: "Process Pulse: October 2026" description: "October 2026 in process management: a study puts a number on process context, SAP's Autonomous Enterprise, object-centric mining in Power Automate, BPM 2026." image: https://blog.noreja.com/hubfs/blog/featured-images/2026-11-03-news-roundup-process-pulse-october-1.png --- [Skip to content](https://blog.noreja.com/en/process-pulse-october-2026#main-content) ![](https://blog.noreja.com/hs-fs/hubfs/blog/featured-images/2026-11-03-news-roundup-process-pulse-october-1.png?width=1820&height=1024&name=2026-11-03-news-roundup-process-pulse-october-1.png) GenAI BPM Process Intelligence # Process Pulse: October 2026 ![Lukas Pfahlsberger](https://app.hubspot.com/settings/avatar/6b6ceb34347ba3ac2667e67af4e9f63f) Lukas Pfahlsberger November 3, 2026 Welcome to Process Pulse, our monthly round-up of what moved in process management and process mining. Once a month we pick the announcements, studies and research results that matter for people who run and improve processes, and add the part the press releases leave out: what it means for a mid-sized company. October 2026 had one word running through almost everything: context. Vendors, analysts and researchers arrived from different directions at the same conclusion, which is that AI agents are only as good as the picture they have of how work actually flows. ## A Study Puts a Number on Process Context On 1 October, ARIS and The Hackett Group published the [Process Context Study 2026](https://aris.com/newsroom/organizations-with-strong-process-context-5x-more-likely-to-deliver-successful-ai-outcomes/), a survey of more than 200 senior leaders at Global 2000 companies. The headline figure is that organisations with substantial process-context experience are five times more likely to report very successful AI outcomes than those with little. Behind it sit more sobering numbers: 86 percent of respondents believe agents cannot be deployed reliably without process context, yet only 22 percent say their organisation has comprehensive, real-time visibility into its end-to-end processes. Only 18 percent report a mature, company-wide AI governance framework. The German trade press picked the study up on 12 October; [IT-Business](https://it-business.de/ki-braucht-prozessverstaendnis-nicht-nur-daten-a-243799c29d0f2af4479a4a51402d6cc6) framed it under the headline that AI needs process understanding, not just data. On 20 October ARIS followed up with the product side, [Process Context for Enterprise AI](https://aris.com/newsroom/aris-launches-process-context-for-enterprise-ai/), which bundles object-centric process mining, a Process Lens for locating where AI pays off, and Model Context Protocol access so that agents can read process context straight from the platform; it has been available since 26 October. Why it matters: the sample is large enterprises, but the gap it describes is more acute in the mid-market, where process documentation is thinner and the people who know how things really work are fewer. The study is vendor-sponsored, so the five-times figure deserves the usual caution, but the readiness gap itself matches what we see in projects: most companies that want agents next year have no current model of how their order, claims or procurement process actually runs today. Documentation describes the intended path; the executed path is what an agent will meet. If you want to know how large that gap is in your own organisation, our earlier piece on [processes that exist only on paper](https://blog.noreja.com/en/processes-exist-only-on-paper?hsLang=en) is a practical starting point. ## SAP Declares the Autonomous Enterprise, and Ties It to Process Measurement At SAP Connect in Las Vegas on 6 and 7 October, Christian Klein presented what SAP calls the [Autonomous Enterprise](https://news.sap.com/2026/10/sap-connect-sap-puts-autonomous-enterprise-to-work/). Since Sapphire in May, SAP says it has built more than 20 Joule Assistants and over 200 Joule Agents, with a target of more than 400 by year-end, and Joule Work, a natural-language interface across SAP and third-party data, is rolling out to customers this month. Two details are more relevant for process people than the agent count. First, SAP positions agents as operating inside deterministic application workflows, using the same data, authorisations and audit trail as the applications themselves, with three levels of oversight from human-in-the-loop approval to human-in-command override. Second, in the [keynote](https://news.sap.com/2026/10/sap-connect-keynote-autonomous-enterprise-in-action/) Klein pointed to SAP Signavio as the tool that measures process value before and after a change, and explicitly linked that measurement to justifying what companies spend on AI tokens. SAP LeanIX gains a central inventory of enterprise agents. Why it matters: the interesting move is not the number of agents, which Klein himself said matters less than outcomes, but the framing. SAP is saying that agents belong inside governed workflows, that their value has to be measured at the process level, and that you need an inventory of which agents exist and what they touch. That is a process management agenda, stated by an ERP vendor. For companies on SAP it raises a practical question: who owns the before-and-after measurement, and does the process owner have the data to do it? Our quick tips on [mistakes with process frontier agents](https://blog.noreja.com/en/mistakes-with-process-frontier-agents?hsLang=en) cover the ownership side of this in more depth. ## Microsoft Brings Object-Centric Analysis Into the Power Platform On 1 October Microsoft announced in its message centre that [Process Intelligence Studio](https://mc.merill.net/message/MC1484591) will be available for object-centric process mining in Power Automate Process Mining, with general availability on 31 October. The studio is a single workspace for exploring event data: tabbed analyses, grid layouts that mix process maps, charts and metrics, filters that apply across the whole workspace, and cross-object metrics; configurations can be saved and shared. Why it matters: object-centric process mining, the ability to analyse a process as the interplay of orders, deliveries, invoices and other objects rather than as a single case ID, was a research topic five years ago and a premium feature two years ago. It is now shipping inside a platform that many mid-sized companies already license through Microsoft 365. That lowers the barrier to seeing how objects actually relate across systems, the view that case-based mining tends to flatten. It does not lower the barrier to good event data: the hard part is still extracting clean, linked objects from the source systems, and no workspace design solves that. ## BPM 2026 Rewards Process Analytics That Act The 24th International Conference on Business Process Management took place at York University in Toronto from 27 September to 2 October, the conference's second North American edition. The [Best Paper Award](https://www.yorku.ca/events/bpm2026/awards/) went to Gyunam Park for work on compiling constraint knowledge into prescriptive process analytics through neuro-symbolic outcome distillation, which in plain terms means teaching predictive models to respect business rules when they recommend actions. The Industry Forum Best Paper, from Dominik Voigt, Juergen Mangler and colleagues around Stefanie Rinderle-Ma, was on process-powered business intelligence, meaning contextualised data for real-time analysis and decision support. The Process Technology Forum honoured work on representing BPMN choreographies in OCEL 2.0. Why it matters: the research community is moving from describing processes to prescribing actions, and the award-winning work is about doing that safely, within constraints a business actually has. Read alongside the vendor announcements above, that is the academic version of the same conclusion: recommendations and agents need process context, including the rules, to be trustworthy. The OCEL 2.0 work is also a signal that object-centric event data is becoming the common format researchers and vendors build on, which is good news for anyone who wants to avoid being locked into one tool's data model. ## Celonis Ships Routing Rules, and Wins a Round Against SAP in Court Celonis published its platform update on [5 October](https://community.celonis.com/product-updates/what-s-new-across-the-celonis-platform-6803). The new routing rules assign responsibility across a team by distributing business objects to owners on logic-based conditions, Task Mining now integrates natively with object-centric data pools so that desktop-level event data flows into the same Process Intelligence Graph as system data, and data integration gains zero-copy connections to Microsoft Fabric lakehouses. In the same week, Celonis won a first round in its patent dispute with SAP. The Munich Regional Court ([case 7 O 12849/25](https://www.juve-patent.com/cases/celonis-lands-first-win-in-process-mining-patent-battle-against-sap-in-germany/)) found that a feature of SAP Signavio Process Intelligence infringes Celonis patent EP 3 139 274, a method for calculating net throughput times between process steps that excludes non-business hours. SAP may no longer offer or use the feature in Germany, must disclose the extent of the use, and is liable for damages. The judgment is not final: SAP has appealed, is challenging the patent before the Federal Patent Court, and told [Computerwoche](https://www.computerwoche.de/article/4234212/celonis-erzielt-etappensieg-im-patentstreit-mit-sap.html) that Signavio remains available to customers. [Celosphere 2026](https://www.celonis.com/celosphere) is confirmed for 2 and 3 December in Munich. Why it matters: routing rules are a small feature with a large implication. Once a process mining platform does not only show where work piles up but also decides who gets it, the platform has become part of the process, and the rules inside it need the same ownership and review as any other process rule. The task-mining integration closes the gap between system logs and what people actually do on screen, where the workarounds from our [piece on shadow processes](https://blog.noreja.com/en/why-workarounds-become-the-real-process?hsLang=en) live. The court case is a different kind of signal. When two of the largest vendors in the market litigate over how a throughput time is calculated, process context has become contested intellectual property. For a customer of either platform, the practical question is less who wins on appeal than how much of their own analysis logic sits inside one vendor's patented method, and whether the event data underneath it would travel if the tool changed. ## What We Read This Month Two pieces are worth your time beyond the headlines. ActuIA published a [status of the EU AI Act as of 6 October](https://www.actuia.com/en/news/ai-act-obligations-in-force-and-delays-under-omnibus-2026-1744/): the omnibus regulation that entered into force in July moves the high-risk obligations for Annex III systems to 2 December 2027, the transparency rules under Article 50 have applied since 2 August 2026, and the next deadline is 2 December 2026, when generative systems already on the market must add machine-readable marking to their outputs. In Germany, the Bundesnetzagentur has been the market surveillance authority since the end of July. For process teams the pressure on high-risk classification has eased by a year, but documentation and logging obligations have not gone away, and companies that treat them as a process design task now will have less to retrofit. One concrete building block arrived on 14 October, when the first harmonised standard under the AI Act, [EN 18286 on quality management systems](https://www.actuia.com/en/news/eu-ai-act-first-harmonised-standard-covers-only-quality-system/), was cited in the Official Journal; it covers only the requirements of Article 17, and standards for risk management, data governance and human oversight are still outstanding. The second piece is UiPath's [October release notes](https://docs.uipath.com/automation-cloud/automation-cloud/latest/release-notes/october-2026), which are less glamorous than a keynote but show where agent governance is heading in practice: organisations can now bring their own guardrail providers to the AI Trust Layer, every guardrail evaluation is recorded in the agent trace, and administrators get a policy for automatic usage top-ups. That is what governance looks like once agents leave the pilot. ## Our Take If October had a thesis, it was that the market has stopped arguing about whether AI agents need process context and started arguing about who provides it. ARIS says it is the process model, SAP says it is the application workflow plus a measurement layer, Microsoft and Celonis say it is the event data, and the BPM research community says it is the constraints. They are all partly right, and none of it can simply be bought. Context is not a product feature; it is the current, evidence-based understanding of how your work actually flows, including the rules it is supposed to obey and the exceptions it tolerates in practice. A licence does not create that understanding. For a mid-sized company the practical consequence is unglamorous. Before the agent conversation, there is a measurement conversation. Which processes are candidates, what does their executed path look like today, where do the variants come from, and who owns the rules an agent would have to respect? Those questions can be answered with process mining in weeks rather than quarters, and platforms built for the mid-market, [noreja among them](https://topai.tools/t/noreja), exist precisely so that the answer does not require an enterprise programme. The companies that do this work first will be the ones for whom the five-times figure in the ARIS study turns out to be true. The ones that skip it will have agents that are fast, confident and wrong in the same places their people currently improvise. ## Food for Thought If you had to hand one of your processes to an agent next quarter, which one would it be, and could you show its executed path from the last three months today, with the variants and their frequency, rather than the documented one? Who in your organisation owns the rules that agent would have to follow, and when were they last checked against what the systems actually show? If a platform starts routing work to people on rules you configured, who reviews those rules, and how would you notice when they are wrong? And when your AI governance framework is written, will it describe decisions at the level of processes, where the risk actually sits, or at the level of models, where it is easiest to describe? [Contact us](https://outlook.office365.com/book/Kennenlernen@noreja.com/?ismsaljsauthenabled=true) ## FAQ #### What were the most important process mining announcements in October 2026? Microsoft announced Process Intelligence Studio for object-centric process mining in Power Automate, generally available from 31 October, and Celonis shipped routing rules, native task-mining integration with object-centric data pools and Machine Learning Workbench 2.0 on 5 October. Both moves push object-centric analysis further into mainstream platforms. #### What did the ARIS and Hackett Group Process Context Study 2026 find? Based on more than 200 Global 2000 leaders, the study reports that organisations with strong process context are five times more likely to achieve very successful AI outcomes. At the same time, 86 percent say agents cannot be deployed reliably without process context, while only 22 percent have comprehensive real-time visibility into their end-to-end processes. #### What did SAP announce at SAP Connect 2026? SAP presented the Autonomous Enterprise with more than 200 Joule Agents and a target of over 400 by year-end, the rollout of Joule Work as a natural-language interface, and three levels of human oversight for agents. SAP Signavio was positioned as the tool that measures process value before and after a change, and SAP LeanIX gains a central agent inventory. #### What did the Munich court decide in the patent dispute between Celonis and SAP? In a first-instance ruling in October 2026, the Munich Regional Court found that a function of SAP Signavio Process Intelligence infringes Celonis patent EP 3 139 274 on calculating net throughput times between process steps. SAP may not offer or use the function in Germany and is liable for damages. The judgment is not final: SAP has appealed and says the product remains available to customers. #### What changed in the EU AI Act timeline in 2026? Regulation (EU) 2026/1744, in force since 27 July 2026, moved the high-risk obligations for Annex III systems to 2 December 2027 and for Annex I products to 2 August 2028. Transparency rules under Article 50 have applied since 2 August 2026, and from 2 December 2026 generative systems already on the market must add machine-readable marking to their outputs. ## Share this post [mailto:https%3A%2F%2Fblog.noreja.com%2Fen%2Fprocess-pulse-october-2026](mailto:https%3A%2F%2Fblog.noreja.com%2Fen%2Fprocess-pulse-october-2026) ## Keep reading ### [![](https://blog.noreja.com/hs-fs/hubfs/blog/featured-images/2026-10-27-two-for-one-workarounds.png?width=1820&height=1024&name=2026-10-27-two-for-one-workarounds.png) Two-For-One GenAI BPM Two for One: Why Workarounds Become the Real Process](https://blog.noreja.com/en/why-workarounds-become-the-real-process?hsLang=en) ### [![](https://blog.noreja.com/hs-fs/hubfs/blog/featured-images/2026-10-20-quick-tips-simplifying-processes.png?width=1820&height=1024&name=2026-10-20-quick-tips-simplifying-processes.png) Quick Tips GenAI BPM Quick Tips: Why Simplifying Is Harder Than Adding](https://blog.noreja.com/en/why-simplifying-is-harder-than-adding?hsLang=en) [![noreja\_logo\_white](https://blog.noreja.com/hs-fs/hubfs/noreja_logo_white.png?width=300&height=78&name=noreja_logo_white.png "noreja_logo_white")](https://blog.noreja.com/?hsLang=en) [Privacy Policy](https://noreja.com/en/privacy) · [Legal](https://noreja.com/en/imprint) · © 2026 Noreja. All rights reserved. ```json { "@context" : "https://schema.org", "@type" : "BlogPosting", "author" : { "@type" : "Person", "name" : "Lukas Pfahlsberger", "url" : "https://blog.noreja.com/en/author/lukas-pfahlsberger" }, "dateModified" : "2026-11-03T11:28:56.678Z", "datePublished" : "2026-11-03T08:00:00.000Z", "headline" : "Process Pulse: October 2026", "image" : [ "https://blog.noreja.com/hubfs/blog/featured-images/2026-11-03-news-roundup-process-pulse-october-1.png" ], "mainEntityOfPage" : { "@id" : "https://blog.noreja.com/en/process-pulse-october-2026", "@type" : "WebPage" }, "publisher" : { "@type" : "Organization", "logo" : { "@type" : "ImageObject", "url" : "https://blog.noreja.com/hubfs/noreja_logo_white-1.png" }, "name" : "Noreja Intelligecne GmbH" } } ```