We’re at the tipping point, says UiPath CEO Daniel Dines. It’s the fulcrum between our previous iterations and incantations of robotic process automation (RPA), the subsequent progression through predictive, generative and now agentic AI… and the application of software “robots” (his preferred term for automation intelligence) with the ability for enterprises to really start reaping return on investment (ROI) from their agentic AI implementations.
To be clear, Dines and team are not suggesting that RPA over the last decade didn’t deliver ROI – far from it, even forms-based automations had a huge impact on business process execution, as did the latter years of predictive and generartive AI as we moved forward to the agentic age – but the progression now is a more tangible, holistic and far-reaching step forward into how to orchestrate agents, robots, people and systems, with governance and security embedded from the start.
Held from Sept 29-Oct 2 in Las Vegas, UiPath Fusion was described as a conference and partner symposium for software application developers, other technology practitioners and business leaders who are ready to turn AI and automation into measurable value. The event’s core focus was agentic automation, by which we mean the convergence of autonomous AI agents, RPA and human workflows to drive measurable ROI in enterprise settings.
Delivering the opening keynote, UiPath CEO Dines’ address was entitled the ‘UiPath Vision of an Agentic World’ and he used the session to unveil what he claims is a “bold vision” for an agentic future where autonomous AI agents, robots and people are orchestrated with governance and control to transform how work gets done and deliver enterprise-scale ROI.
Process is the hero (not the tool)
The company says that orchestration is the secret to unlocking ROI from rules-based deterministic AI, from goal-based AI that works to pick the best path… and from all forms of agentic AI as we now move forward to automation that works to actually impact business workflows. Addressing the MIT study that has suggested that some 95% of generative AI pilot projects fail, Dines says that we can now focus and really see what the real ingredients are that comprise successful automation projects.
“It is my belief that the most powerful AI services today can only be built on the most performant automation foundations with essential orchestration services. Many companies have started out with ‘small ambitions’ to apply AI to individual processes and see how things work out. The problem with this is, non-deterministic AI has an ability to work on really complex work processes where headaches can really be impacted,” said Dines, suggesting that companies need to view a wider and larger attack surface for successful agentic AI implementations.
Dines says that solid automation works not by agents having access to data in the first instance, it depends on access to tools from the start i.e. this is where solid automations are built, for any size process. We need to combine people, robots and business logic if we really want to be able to create industry solutions that work in really complex places.
Put more skin in the game, now
In terms of skills, Dines recognises that now is a time of “early adopters” and we are seeing firms that pilot projects in this space progress as fast as they can.
“But really, I would like organizations to put more skin in the game now and really look to create the archetypes [in terms of typical examples and use cases that can be applied to real world every day business] that can help entire industries to move forward with automation today and tomorrow – while also embracing the context of open ecosystems,” said Dines.
One of a number of users invited to speak on the stage, Rajat Kalia SVP & CTO at Voya explained how his automation deployment process went from around 60% accuracy in its first iteration, to then roll out to over 90% after a couple of weeks. Voya Financial provides financial solutions focusing on retirement plans, investment management and employee benefits.
Be leader-led, first & foremost
“We moved forward because our CEO knew it was the right agenda to oversee in terms of how the company should execute on an AI-driven digital-first strategy. We don’t talk about being business-like with AI, we talk about being leader-led and making sure that teams work with AI that is applied on the right (high value & high friction) use cases that really help reduce risks, deliver better on service level agreements (SLAs) and apply other positive effects which will be felt across our subject matter experts as they test use cases and help educate the wider organisation on AI literacy in general,” said Kalia.
UiPath’s Dines spoke between user use cases to explain how his management philosophy is always about “being immersed” today. In the company today, UiPath is a 4000-person company and teams have identified and progressed some 500 agentic application ideas which will help (in Dines’ view) the firm’s ability to “grow without growing the headcount” and move from days to hours in terms of so many of the enterprise systems that UiPath itself operates on a day-to-day basis.
The audience also welcomed Oliver Pfeil, CEO for business services at Capgemini who reflected the other speakers’ themes by saying that process-based development is the real key to embracing orchestration as a means of delivering AI.
Grasping orchestration, harmoniously
Capgemini itself decided to work with UiPath on the next stage of its AI deployments primarily because the firms have already worked together for more than a decade and because Capgemini is essentially a services company not a product company. This meant that when it needed to grasp an orchestration “product” (which in reality is really a UiPath service layer) it knew what it wanted to do. The company has used AI to drive business processes across the whole Capgemini group, but query resolution has been a critical use case and (because it can take human operative days to respond at this level), handing over 90% of these processes over to agents made a lot of sense.
To this exact point, Dines straddled into the product keynote sections of this event by talking about how the company has now announced UiPath Screenplay and API Workflows for users to get on with accessible and accelerated building of automations using natural language and end-to-end API workflows. Blending UiPath’s leading RPA and API with large language models and large action models, these tools lower the barriers and simplify development for every employee and every organisation, reducing the time (and cost) of agent and automation building.
Formerly known as UI Agent, Screenplay functions as an intelligent agent that can understand the user’s intent and autonomously control the mouse and keyboard to execute tasks on a desktop. This approach is argued to lower the development entry point and barrier (mainly because it doesn’t require intricate knowledge of selectors or complex coding) and open the door to more adaptability; it can dynamically adjust to changes in the user interface, such as shifting elements or stylistic updates, which makes the resulting automations more resilient and less prone to breaking.
UiPath chief product officer Graham Sheldon hosted the second section of this event’s keynote session and spoke about how leaders in the AI implementation space are going after high-impact complex processes first, but doing so with the UiPath orchestration engine at the heart of that progression.
Determining deterministic vs non-deterministic
But what do we mean when we talk about “agentic orchestration” he asked? It’s about understanding that some work is deterministic, some is non-deterministic and working out how – via a total process of orchestration – a business can connect them all inside a seamless workflow with full visibility. When long complex processes are transformed into a single system for business-critical process, orchestration is the key. Being able to run these systems at scale and also to be able to notice bottlenecks (where an IT team can plug in the right tool to make things happen) is the secret. Sheldon says that UiPath Maestro is the technology that the company is offering that works to provide the glue here. Case management and process apps for UiPath Maestro has been introduced at Fusion 2025
Also in product news, new UiPath IXP agentic document processing enhancements securely optimise data and document-driven workflows across an enterprise. Capabilities are here for advanced data extraction, validation and agentic looping, enabling the scaling of complex tables, pages and fields. Meanwhile, features like Autopilot for schema creation, Extraction Agents and Validation Agents reduce manual effort and deliver accuracy, with a proprietary customizable LLM to drive what the company calls “last-mile performance” today. Purpose-built for high-value use cases such as contract comparison and system-of-record validation, these developments integrate seamlessly with UiPath Agents and Agentic Solutions to accelerate time-to-value and optimise document workflows across industries.
Testing, Testing, 1, 2, 3
The audience is used to a slew of new product announcements being showcased in a product keynote, so onwards to UiPath Test Cloud, where features now accelerate and scale testing across every step of the software development cycle. Performance testing allows QA engineers to simulate real-world user traffic and workloads to assess how applications perform under stress, from identifying bottlenecks to ensuring system stability and validating speed and responsiveness to meet business and customer expectations at scale.
In addition, Test Cloud now offers Studio Web for software testing, self-healing test automation, autonomous testing and chatting with UiPath’s AI-driven Autopilot in Test Manager, delivering a true end-to-end solution for agentic testing.
“Powerful new AI security and compliance features are built in, giving organisations the confidence to deploy and manage agentic automation with control and enterprise-grade oversight and meeting international standards, at every stage. Expanded capabilities include extended agent guardrails for controlled agency and sensitive data use protection, coupled with content moderation for real-time PII masking. And UiPath’s Unified Audit 2.0 platform offers a single source of truth to streamline compliance, accelerate investigations and enforce accountability across all automation actors, enabling the integration of audit logs for automations, agents and governance events into existing compliance and security workflows,” said Sheldon and team.
From Forward to Fusion to…?
UiPath Fusion was the evolution (and open rebranding) of the company’s previous user, customer, partner and practitioner symposium – the artist (we mean “event”) formerly known as UiPath Forward. Dines spoke openly during his keynote to express the reasons for this renaming. To summarise his thoughts, he is suggesting that Forward was all about automation acceleration to propel companies forward faster, but that Fusion is all about a newly forged amalgam between agentic AI in its multifarious forms and the presence of business value, when manifested and validated in tangible return on investment figures.
What comes next after UiPath Forward and Fusion then?
This is Las Vegas, it’s Fabulous, obviously.