Why Agentic AI Needs Business Process Orchestration

Why Orchestration is Key to Governance

Agents are goal-oriented and probabilistic. Enterprise operations are audited and deterministic. Business process orchestration resolves that tension by modeling the whole process as a governed BPMN 2.0 diagram, then running agents, robots, and people against it with full traceability. Orchestration keeps processes running. Process intelligence keeps them learning. Enterprises need both, and they are not the same thing.

BPMN is a notation standard from the era of swim lanes and flowchart diamonds, which does not sound like the foundation for anything described as agentic. That is exactly why it works. Non-deterministic technology needs a deterministic frame around it, and the market already has one that both humans and machines can read.

Marshall Sied, Co-founder and Co-CEO of Ashling, sat down with Robert Love, Principal Product Manager for GenAI and Orchestration at UiPath, to work through how process foundations, orchestration, and process intelligence come together in an agentic operating model. Ashling is a UiPath Diamond partner, the highest partner tier, so the conversation ran technical fast.

Start With What a Process Actually Is

A process is a series of activities and inputs that lead to outputs and outcomes. Order to cash is a process. Collections inside order to cash is a process. The definition is simple, and the discipline around it is where programs get sloppy.

BPMN 2.0 is the standard most business process management solutions were built on, and it remains the de facto process foundation in the market. Anyone who has built a Visio diagram has used components of it without necessarily formalizing them.

The reason this matters more now than it did five years ago: a gateway is a decision, and deciding is exactly what you hand an agent when you give it a goal and let it determine its own path. A standard built to describe human and machine decision paths turns out to describe agent decision paths too.

What Business Process Orchestration Does

Business process orchestration connects the moving parts of a process: the inputs, the actions taken through APIs or RPA, and the decisions and branches along the way. UiPath Maestro is built on the BPMN 2.0 standard and coordinates that flow across systems.

In practice, a single process might have a robot updating a record, an AI agent evaluating risk, and a person granting approval. Orchestration means all three follow one governed design. The model defines how the work gets done, enforces timing, and assigns responsibility for each record.

The maintenance benefit is worth naming. Change a risk threshold once and every downstream component inherits it. That is a meaningful difference from updating the same rule in nine places and hoping you found all of them.

Orchestration also covers the full life cycle rather than the design stage alone. Once processes are live, teams can pause instances, retry a specific step, or roll a process back to a previous point. When an IT system goes down, that is nobody's automation defect, and being able to retry that one step or route to an alternative path is the difference between a blip and a backlog.

Process Orchestration and Process Intelligence Are Not the Same Thing

This is a live source of confusion, and the two capabilities are permanently married when a program is set up well.

Process intelligence is not a one-time X-ray of your ERP and CRM logs. It is a continuous improvement capability, which is what makes it useful to a lean team or a continuous process improvement function rather than a consulting deliverable that ages on a shelf.

The enrichment loop is the part worth understanding. After a process is modeled, implemented, and executing, pulling logs from the ERP and CRM it touches deepens the picture considerably. You see every variant, every delay, and the cause behind the rework.

Simulation is where this gets genuinely interesting. Ashling recently worked with a client simulating the fit gap of moving from SAP ECC to S/4HANA, modeling both what the change meant for training their people and where agents made sense in the future state for the steps that are not deterministic. Simulate first, harden second, then move it into orchestration.

Why BPMN Is the Right Foundation for Agents

Three reasons this standard carries the agentic era rather than getting replaced by it.

1. Standardized semantics. BPMN provides a common readable schema for humans, machines, and agents alike. Give a goal-oriented, non-deterministic agent a consistent input structure and you get more consistent outputs from it. The organizational benefit is just as real: process owners, automation teams, and the people newly building AI agents all end up describing work the same way, which is not a small thing when those three groups have historically spoken different languages.

2. Composable architecture. Standard semantics let you coordinate multiple agents across tasks and systems with higher confidence and lower risk. It also means you swap components incrementally. A human-in-the-loop step becomes an agent step when it is ready, without replacing a thousand-step RPA workflow with one master agent in a single move. Chaining agents together is technically possible today. Doing it across an enterprise without this kind of blueprint is a different proposition.

3. Cross-vendor interoperability. Enterprises run agents from many places: agents built on the automation platform, coded agents built on frameworks like LangChain, agents embedded in systems of record like Salesforce, and agents on data platforms like Snowflake and Databricks. Maestro orchestrates agents across those ecosystems, including Microsoft Copilot, Azure AI Foundry, and Google Vertex. Open architecture was one of the reasons Ashling backed UiPath early, and it matters more now that every CIO is trying to reconcile a wave of new agents with the architecture-reviewed platforms they already own.

The data platform pattern is the one showing up most in client environments. Teams built agents on their data platform that query and reason well, then hit the wall where those agents have no action hooks into the ERP, the CRM, or the customer communication stack. Orchestration is what turns querying into acting.

Where Orchestration Sits in the Stack

Gartner's Business Orchestration and Automation Technologies category describes the convergence enterprises have been working toward: process orchestration, connectivity, low-code development, and agentic capabilities consolidated into one platform rather than a patchwork of point tools.

Most of the foundational layer is already in place at large enterprises: RPA, API integration, intelligent document processing, workflow management, low-code development, process modeling and simulation, and event-driven triggers. What has been missing is the connective fabric across all of it, with multimodal AI and multi-agent coordination underneath, and process observability holding the governance line.

That fabric is business process orchestration. The category name is new. The need has been visible in enterprise architecture diagrams for years.

Case Management Makes Orchestration Concrete

At FUSION 2025, UiPath introduced Case Management and Process Apps in Maestro, which is where orchestration stops being an abstraction for a lot of teams.

Case management handles long-running work: a claim, a loan, a dispute, an investigation, a service request. The defining traits are multiple stages, multiple participants, and different policies, SLAs, and escalation rules per stage. Agents, robots, and people work the case under one timeline, one set of SLAs, and one history.

Per stage, teams set the SLA, define what happens on a breach, and define what happens when the SLA hits risk threshold, including who gets notified in each case. Tasks are specified per stage, including which run in parallel, whether they are RPA, document processing, API, or agent tasks. Exception paths get modeled as secondary stages, so a withdrawn claim is handled as a recognized path rather than an unhandled break in a linear flow.

Process Apps generate a role-based interface from the model itself, with an SDK available for teams that want to build their own. A claims officer sees the open cases, the compliance rate, the SLA risk, and inside an individual case, the claim details, the attached documents, the agent recommendations, and the full audit trail of what the agents did, what the robots did, and what the people did.

Ashling views case management and process apps as the execution scaffolding for agentic work, because staged, multi-participant cases are how business operations actually run.

A Property Claims Example

Property claims illustrate the whole argument. A claim requires multiple people to finalize, approve, and close, and it arrives as a pile of documents across systems: ID uploads, photos of the damage, records, and receipts.

The four stages exist regardless of how the work gets done: intake, review, settlement, and closure.

Manual. People handle emails, attachments, incident reports, and coverage checks, searching for duplicates and reading through logs. Slow, difficult, and hard to repeat consistently. Skilled people spend their day moving information around instead of making the decisions only they can make.

With robots. Documents get collected and processed and systems get populated, which recovers real speed. The decision-making still queues behind people, so delays persist and repeatability stays uneven.

With agentic automation. Agents read the packet, classify it, check for duplicates and fraud indicators, and recommend next steps. Robots continue handling the mechanical work. People focus on exceptions, approving and validating what the agents surface.

The same pattern travels across industries. In finance, agents analyze transactions, robots reconcile them, and people approve exceptions. In insurance, agents interpret documents and flag anomalies while adjusters make the final call. In public sector work, agents classify cases and humans make sure outcomes stay fair and equitable.

Robots Are Not Going Anywhere

Duplicate claim checks and fraud checks generate many sub-processes per case, each with its own need for coordinated human review. That expands the addressable surface for agents, and it raises the bar on how tightly the handoffs between agents, robots, and people have to be orchestrated.

At the same time, plenty of steps have to be 100% accurate every single time and will be examined by auditors and internal controls. Payment processing is the obvious one. Deterministic, rules-driven automation is the right tool there, and it will stay the right tool. Nobody should want an agent independently approving a million dollar claim.

The takeaway from the webinar's 2026 outlook: this stops being framed as a choice between traditional automation and agents. Combining rules-based automation with probabilistic automation inside one governed process is the path forward, and the two do not have to live in separate lanes. Enterprises will also decide they cannot compromise on visibility and control once they understand both the power and the limits of purely agentic approaches.

AI Explainability Is the Real Adoption Blocker

Ashling's view on why enterprise agentic adoption has lagged the noise around it: explainability.

Moving from deterministic to probabilistic technology is a cultural shift for a large company, not a technical migration. That shift is manageable inside a digital product engineering team. Doing it enterprise-wide, where auditors, internal controls, and regulators are all in the room, is a different order of problem.

This is why the executable log matters as much as the execution. Full traceability across every action, whether taken by a human, a robot, or an AI agent, including the specific LLM calls and tool calls an agent made, is what converts a black box into something an enterprise can approve. UiPath applies the same standard to agents built outside its platform, which is what makes the traceability argument hold across a mixed estate.

If you are building the measurement side of this, Ashling's guide to measuring agentic AI success covers the frameworks that go past cost savings.