need better tools to manage the intersections between processes, highlighting the challenge organizations face in realizing full value from their AI and automation investments,” said Camunda executives. Bridging the gap between AI vision and reality, Camunda executives argued, requires a move beyond standalone, siloed agents toward “agentic orchestration,” which enables teams to blend deterministic and dynamic orchestration of business processes, leveraging agents to add dynamic reasoning to deterministic processes so they can adapt in real time. Apparently, professionals responsible for automation within their companies tend to agree, as 88 percent said AI needs to be orchestrated across business processes if organizations are to get maximum benefit from their AI investments, while a full 90 percent said AI needs to be orchestrated like any other endpoint within automated business processes to ensure compliance with regulations. At the same time, 85 percent also said their organization has not yet reached the right level of process maturity to implement agentic orchestration, suggesting a need to increase process maturity and AI maturity in parallel. “Deterministic orchestration has always established structured guardrails. By blending it with dynamic orchestration patterns to leverage reasoning across AI agents, people and systems in end-toend processes, enterprises can build a foundation for AI agents they truly trust,” said Camunda’s Petersen. “This is enterprise agentic automation in practice, and it is how organizations will turn today’s AI experiments into durable, businesscritical capabilities.” ABBYY, for its part, found that 99 percent of organizations that augmented GenAI deployments with complementary tools, such as process intelligence, Document AI and retrieval augmented generation (RAG), reported improved outcomes, including more consistent results (50 percent), greater accuracy (43 percent), stronger trust (43 percent), and cost savings (42 percent). The failure to address existing pain points, challenges and the gaps between AI hype and reality threatens more than the rate of growth and further investment in AI technologies. In addition to the ding it could make to a partner’s advisor status, there is the very real chance businesses could pull back on AI-oriented programs altogether. After all, Accenture’s recent study found that regular AI agent usage among employees dropped 10 points since the summer, while a soon-to-be released survey from RingCentral found that nearly 40 percent of organizations have paused or cancelled an AI project, mostly because expectations, workflows or training weren’t aligned from the start. The ultimate lesson for technology advisors could be to focus less on accelerating usage and more on improving how AI implementations and outcomes are designed, measured and supported. o AI Infrastructure Pain Points According to Gartner, about half of AI based spending this year will be on infrastructure. Perhaps that’s not so surprising considering that infrastructures are still catching up to new and evolving AI workloads. Add to that fact how network operators don’t seem to have a lot of confidence in the ability of their current AI infrastructures to meet the demands of AI-driven applications in the next few years. A survey from A10 Networks, a provider of security and infrastructure solutions found that 79 percent of organizations were either “somewhat confident” or “not very confident” in their infrastructure’s ability to handle the performance, latency and availability demands of AI-driven applications. Only 13 percent said they were “very confident.” What are the biggest limitations or pain points in your current infrastructure when it comes to supporting AI workloads? Security constraints (current security applications/solutions can’t inspect or protect AI-related traffic effectively) 49% Legacy systems that are inflexible or hard to integrate with new AI platforms 39% Scalability limitations (hard to scale out infrastructure quickly for AI demand) 38% Lack of visibility/monitoring for AI workloads (potential for “blind spots” in traffic or performance metrics) 33% Insufficient computing power (CPU/GPU) for AI processing 30% Network bandwidth or latency bottlenecks for moving large data/AI traffic 29% Data storage or data management bottlenecks for AI (throughput, I/O, etc.) 19% Other 4% Source: A10 Networks Also not so surprising, security constraints emerged as the top challenge, as 49 percent of respondents named security as their biggest limitation for current infrastructure to support AI workloads. Four in 10 already have deployed new AI-specific security solutions into their infrastructures, showed A10’s data. o 19 WINTER 2026 | CHANNELVISION
RkJQdWJsaXNoZXIy NTg4Njc=