Governance and orchestration emerge as biggest barriers to scaling agentic AI

In attendance at SS&C Blue Prism Launch Live, experts warned organisations must first tackle visibility and control if they want to scale agentic AI safely

Businesses are racing to move agentic AI projects from experimentation and into production, and there are two issues that are emerging as the biggest barriers to scale: orchestration and governance.

In attendance at SS&C Blue Prism Launch Live 2026, conversations repeatedly returned to these themes, highlighting how organisations are grappling with control, accountability and compliance as AI agents gain greater autonomy within business workflows.

The event comes after the vendor’s launch of WorkHQ, its control plane for agentic automation, adding to its portfolio of enterprise intelligent automation, robotic process automation and agentic AI offerings.

As organisations look to move beyond experimentation, the focus is shifting from model capabilities to operational control.

From the event, CRN has outlined some key conversations that cover what people want to know about agentic AI, and why they matter for scaling the technology.

Orchestration and scaling agentic AI

As organisations move beyond traditional robotic process automation (RPA) and begin deploying AI agents, maintaining control is becoming one of the biggest challenges facing enterprises.

For years, businesses have already automated processes through RPA, however, “as we enter the agentic workflow era, we’re looking to add 30, 40, 50 per cent of the types of processes you weren't able to automate before.

“Because of the limitation of what RPA could do, now you can add in AI to take on some of the cognitive workload that humans had to do, and leave the rest for humans where there is very high-level, high-value work,” explained Jamie Kelly, director product marketing SS&C.

But with that opportunity comes new risks.

“When we look at some of the particular challenges, visibility is a key challenge.

“[Challenge] two is ambiguous accountability.

“How much authority you give to an agent is a key decision that you have to make.

“Ultimately, who owns what decisions the AI makes, and that is a foundational thought that must go into each and every one of your planning,” he continued.

Kelly argued that organisations must clearly define responsibility for AI-driven decisions before deploying agents at scale.

The key difference between agentic automation and traditional automation is that failure hits differently.

Kelly explained, “Traditional automation, when something went wrong, it stopped.

“The fact it stopped meant no actions that happened after it could further damage your business.

“Agentic failure can hit differently because if you give AI the authority to auto approve a particular job, then it will continue regardless of if it's taken the wrong deduction, and it keeps moving forward.

“That’s what we call silent continuation. Things just move on without necessarily you being comfortable with what it's come up with.”

Unlike traditional automation failures, which often stop a workflow altogether, agentic failures can continue operating and amplify mistakes across interconnected processes.

Control comes in two levels.

The first is controlling the individual parts, specifically, creating guardrails for individual actors performing work. How much scope of authority is being given to each individual agent? How much scope is it being given to do individual tasks? What are the security measures in place?

The second is orchestrating the automation system as a whole.

Orchestration allows visibility into what agents are doing, how many tokens they’re using, the work that is successfully being completed, and how human involvement in the loop can be reduced to speed up processes.

By controlling both levels, organisations are able to have greater insights into what each individual agent is doing, as well as how this is impacting the entire chain of automation.

Orchestration is key when it comes to scaling agentic AI across an organisation, rather than using it for odd tasks in multiple different areas.

While orchestration focuses on maintaining control and visibility over AI agents, governance addresses the broader frameworks needed to ensure those systems can operate safely.

Agentic AI governance and where partners can add value

Drew Sonden, director of strategic growth at SS&C Blue Prism opened his talk about governance by asking the room for a show of hands to signal who in the room was concerned about the pace of change and potential AI harm the technology could cause, and was met with an overwhelming number of hands.

“The majority of people here have some concerns over the pace of change that AI is offering and the risks that it poses to our organisations, and I guess that means that whilst governance isn't the most exciting topic, it's more relevant and important than ever within our automation practices.”

Governance frameworks are key in making sure businesses operate smoothly, but especially important when agentic AI is thrown into the mix.

AI and machine learning are not new technologies, but their capabilities have vastly exceeded the speed at which we have been able to keep up.

This is leading to what Sonden refers to as the ‘AI chasm of compliance’.

“Despite the immense power of AI, and potentially because of some of the concerns that people have around it, what we see is a constant cycle of organisations looking to implement AI programmes, making rapid early progress and then stalling.

“The reason that these programmes are stalling is not because the technology is not capable, but because a lot of organisations are trying really hard to move fast, and things like governance tend to get forgotten.”

Sonden argued that governance considerations are often overlooked in the rush to deploy AI, leaving many pilot projects unable to progress into production.

“It’s a cycle of repeated deliveries that show promise but then fizzle out because there's no way they can move from test into production.”

This is where partners have an opportunity to add value, helping customers establish governance frameworks, design operating models, and implement controls that allow AI initiatives to move safely from proof-of-concept to production.

Although, this presents a new problem, he explained.

“If we put too many constraints in front of the use of AI, then individuals within our organisations will attempt to circumvent them, and we go back to the problem where we have an area of grey AI IT, and we don't have control over what's implemented within our organisations.”

It is key for organisations to establish governance frameworks early in their AI journeys, ensuring projects can move from proof-of-concept to production while remaining compliant and secure.

The challenges surrounding governance and orchestration present a significant opportunity for channel partners, with organisations increasingly seeking trusted advice to design AI operating models, establish governance frameworks and implement controls around autonomous agents.

The speakers argued that governance and orchestration will be the difference between organisations running isolated AI experiments and deploying agentic AI successfully at scale.