Breaking through the automation ceiling

For decades, automation meant doing the same things you were already doing, only faster. Forms went digital, manual handoffs got eliminated, and systems that had never spoken to each other could be seamlessly connected. That meant increased efficiency, decreased risk of error, and an overall better experience across a multitude of business processes.

But with the advent of AI, we’re seeing new possibilities. Rather than just continuing to move faster, we can start expanding the boundaries of what automation means and what it can achieve. Processes that once ground to a halt at the words “human intervention required” are now within reach of true, governed autonomy. So why aren’t we doing it already?

The challenge is that with expanded possibility comes expanded complexity. More systems, more agents, more data, and more expectations require more than automation – they require orchestration. And that changes the strategy for every organization investing in process improvement. Organizations trying to simply layer AI onto fragmented tools aren’t really transforming … they’re just buying into hype.

The organizations that will succeed in this new era aren’t those who just automate more, but those who orchestrate differently, connecting agents, people, processes, and systems into coordinated, intelligent architecture.

Let’s pause and consider the distinction between generative AI and agentic AI. While generative AI helps users create (new strategies, text, code, ideas, and more) based on user prompts, agentic AI — ideally — takes action without waiting for prompts. When organizations talk about incorporating AI into their businesses processes, they are generally talking about agentic AI.

But part of what’s keeping organizations stuck in this purgatory stems from how loosely the word “agentic” is being used. Le Clair draws a careful distinction between systems that are truly agentic and those that are what Forrester calls “agentish” — meaningful, but a different category entirely.

Most AI systems in production today are insight agents: Large language models used for search or retrieval-augmented generation (RAG), providing useful outputs but lacking a meaningful action component. Slightly more advanced are solver agents — taking a deterministic process and automating a specific workflow step using a model. Le Clair offers anti-money laundering alert review as an example: A model reviews flagged transactions, filters out roughly half the false positives, and hands off only the genuine alerts to human analysts. Valuable, but “agentish.”

Worker agents go further: The model itself does the orchestration, routing work to pre-built automations, RPA bots, APIs, or human endpoints. Even more transformative, but still technically agentish. (Psst – Nintex Agent Designer falls into this category. Watch it in action!)

Le Clair’s is not a pessimistic statement, but a clarifying one. It means organizations that are feeling behind are not actually behind — they are navigating the same frontier as everyone else.

And it means the practical opportunity right now is in worker and solver agents, not executive agents: Getting the “agentish” implementations right, governed well, and built on foundations strong enough to support what comes next.

The most consistent message from those who are actually getting results from automation and AI is one that rarely features in keynote demos or vendor marketing: It starts with understanding your processes.

Not automating them. Not enhancing them with AI. Just understanding them.

This sounds obvious, but is surprisingly rare. Most organizations carry their process knowledge in the heads of long-tenured employees, in email threads, and in undocumented workarounds that evolved organically over years. When automation projects stall, as many do, it is often not because the technology failed, but because the technology was pointed at a process that nobody had fully mapped.

“You really can’t automate something you don’t understand. Step one is understanding it.” 

—Kevin Schaal, CEO, AiGS (Nintex Premier Partner) 

Larry Bennett leads the process excellence team at GM Financial, where his mandate is deceptively simple: Document how work actually happens across the enterprise, and make that knowledge accessible to everyone who needs it.

It is slow work. There is no enterprise-wide mandate forcing departments to participate. Adoption happens department by department, built on demonstrated value rather than top-down directive. In the early days, the benefits were not immediately obvious and progress was gradual.

But the documentation repository has grown. And as it has, teams that had operated in silos began to see how their work connects to others. Automation opportunities that were invisible when processes were examined individually became apparent when viewed end-to-end. Audit and compliance teams found documented foundations to map controls against. And AI tools, now in the early stages of layering on, have something real and reliable to work with.

Hawaii Employers Mutual Insurance Company (HEMIC) came to Nintex at a different point in their journey — not with a specific broken process to fix, but with a strategic mandate to modernize how work gets done across the organization. Chief Transformation Officer Leila Kagawa describes the challenge facing many mid-sized organizations: The market for automation and AI tools is overwhelming, and the pressure to move quickly can lead organizations toward the wrong solutions.

HEMIC’s approach has been deliberately sequenced. Working with Nintex Premier Partner AiGS, they began by documenting standard operating procedures in Nintex Process Manager — establishing a baseline understanding of how work actually flows before attempting to automate or augment it with AI. The goal is not just operational efficiency, but giving people the clarity to focus on what matters most.

“We always talk about technology in terms of meeting people where they are. We had to find a partner that meets us where we are on our continuous improvement journey.” 

— Leila Kagawa, Chief Transformation Officer, HEMIC 

Kagawa is also thinking deliberately about her workforce. HEMIC’s employees span multiple generations, with different relationships to technology. Automation, she believes, can give experienced employees relief from repetitive work — freeing them for the relationship-intensive parts of their jobs — while creating more standardized and navigable onboarding paths for newer team members.

There is a version of the AI conversation that treats governance as a constraint — something that slows down innovation, that cautious legal and compliance teams impose on enthusiastic technology leaders. We think that framing gets it backwards.

Governance is not the roadblock to AI adoption. It is the thing that makes AI adoption sustainable. 

This is one of the clearest takeaways from Le Clair’s analysis of where organizations are struggling. The AI deployments that have failed to deliver value share a pattern: They moved fast, they skipped the control layer, and they produced outputs that their organizations as a whole were not comfortable enough in to act on.

“You need auditability. You need explainability. You need a level of observability. All of that is independent of how the agents are interacting with each other — but you need to understand those things.” 

— Craig Le Clair, VP & Principal Analyst, Forrester 

Prediction: The next phase of AI and automation investment won’t be defined by how many agents an organization deploys but by how well those agents are orchestrated alongside people, workflows, and systems to drive measurable outcomes.

That is both a technology challenge and a strategic one … one that Le Clair has thought carefully about.

From deterministic to adaptive

Forrester has introduced the concept of Adaptive Process Orchestration (APO) to describe where the market is heading. At Nintex, we call it Agentic Business Orchestration. The underlying insight from both concepts is that the coordination challenges of the near future look fundamentally different from what traditional automation platforms were built to handle and require fundamentally different approaches to undertake.

Where traditional intelligent automation platforms managed APIs, RPA bots, and workflow steps, the emerging challenge involves coordinating all those plus potentially hundreds of agentic AI models with different access requirements, different latency profiles, different governance needs, and the human workers who need control over when and where they step into various processes.

What makes the agentic moment genuinely different from previous waves of automation, Le Clair argues, is not that it allows organizations to do the same things faster, but that it allows them to reimagine the processes themselves.

In the deterministic world, humans served as middleware: The connective tissue between systems, interpreting context and making judgment calls that could not be encoded in rules. AI agents can soon overtake that role, making connections across information domains that no human could efficiently maintain, and routing work to the appropriate next step based on understanding rather than predetermined rules.

That shift does not make humans irrelevant. In fact, it makes human judgment more valuable when applied where it is truly needed rather than consumed by coordination and routing.

Le Clair invokes science fiction writer William Gibson, who said, “The future is already here — it’s just not evenly distributed.” Before the mobile explosion, he recalls, the BlackBerry was so popular among a small group of professionals that prescient observers could already see the coming cell phone boom, if not its exact form (the iPhone was introduced in 2007).

Today, he sees the same pattern in AI: Pockets of organizations already using agentic capabilities with remarkable precision and confidence, making decisions in ways that required very little configuration and delivered outsized results.

“You see these little pockets of brilliance, and you think: There are legs here. There are good things coming down the road. But we have to be patient, and we have to solve these trust issues.” 

— Craig Le Clair, VP & Principal Analyst, Forrester

The opportunity is real, but the timeline requires patience. But the organizations that are building their foundations now — documenting their processes, building their AI governance, and sequencing their automation investments deliberately — are the ones who will be best positioned to move when the moment arrives.