Wednesday, 5 August 2026
As organisations accelerate investment in autonomous artificial intelligence, many are discovering that deploying AI agents is proving easier than achieving meaningful business results. According to ARIS CEO Guillaume Bacuvier, the challenge is not the capability of today’s AI models but the lack of operational knowledge available to guide them.
Speaking ahead of the Ai4 2026 conference in Las Vegas, Bacuvier argues that many companies are expecting AI agents to transform complex business operations without first providing them with an accurate understanding of how those organisations actually function.
His comments reflect a growing debate within the enterprise AI market, where businesses are shifting attention from the performance of language models to the quality of the organisational data and processes that support them.
Technology Is Advancing Faster Than Business Readiness
Large language models have become increasingly capable of analysing information, generating content and assisting employees with a wide range of tasks. However, these systems generally have little knowledge of an individual company’s internal procedures, approval structures, regulatory obligations or decision-making frameworks.
Without that context, AI agents may perform well in demonstrations or isolated pilot projects but struggle when deployed across large organisations where thousands of interconnected business processes influence everyday operations.
Bacuvier believes that improving business understanding is now more important than simply deploying more sophisticated AI models.
Operational Context Determines Success
According to ARIS, organisations often underestimate the complexity hidden within their own operations.
Large enterprises typically manage thousands of interconnected workflows spanning finance, procurement, human resources, manufacturing, customer service and regulatory compliance. Each process contains approvals, policies, dependencies and exceptions that have evolved over many years.
When AI agents lack visibility of these relationships, they may complete individual tasks successfully while failing to support broader business objectives or introducing unexpected operational risks.
Rather than relying solely on increasingly powerful AI models, ARIS argues that organisations should first establish a structured representation of how their business operates before introducing autonomous automation.
Measuring Business Value Instead of AI Adoption
The company also challenges the way many organisations measure the success of AI initiatives.
Counting the number of deployed AI agents or completed pilot projects provides little indication of whether technology is improving operational performance. Instead, Bacuvier argues that businesses should focus on measurable outcomes such as faster customer service, lower operating costs, improved compliance, reduced risk and higher productivity.
This reflects a wider trend within enterprise technology, where executives are increasingly demanding evidence that AI investments deliver tangible financial returns rather than simply demonstrating technical capability.
Preparing Organisations Before Deploying AI
ARIS recommends that organisations establish several operational foundations before expanding the use of AI agents.
The first is achieving complete visibility across end-to-end business processes so that AI systems understand how individual tasks contribute to wider operations.
The second is defining governance structures, including approval responsibilities and decision-making authority, before allowing autonomous systems to execute business activities.
Compliance requirements should also be incorporated directly into operational workflows rather than being treated as separate controls after deployment.
Finally, organisations should evaluate AI programmes using business performance indicators rather than technology adoption metrics alone.
Practical Examples from Large Enterprises
ARIS highlights several organisations that have invested heavily in documenting and simplifying their operational processes before introducing more advanced automation.
At Boots UK, the company says a connected process architecture covering more than 2,000 business processes enabled a finance workflow to be redesigned from 220 individual steps to approximately 40, significantly reducing execution time while creating a more structured operational environment for future AI deployment.
Italian aerospace and defence group Leonardo has also developed thousands of interconnected process models as part of a digital representation of its business operations. This framework is intended to provide the governance and organisational knowledge needed to support future AI applications across engineering and manufacturing activities.
These examples illustrate a growing recognition that operational transformation often begins with understanding existing business processes before introducing new technologies.
The Next Stage of Enterprise AI
Bacuvier’s argument reflects an important shift taking place across the enterprise AI market. Early adoption focused largely on the capabilities of increasingly powerful language models. Attention is now moving towards the quality of the business environment in which those models operate.
Technology providers are investing more heavily in systems that capture organisational knowledge, map business processes and embed governance directly into AI workflows. The objective is to create AI agents capable of operating within clearly defined business rules rather than simply responding to prompts.
As enterprises continue expanding AI across finance, manufacturing, supply chains and customer operations, operational context is becoming as important as model capability itself.
For many organisations, the next competitive advantage may not come from deploying more AI agents, but from ensuring those agents have a complete understanding of the business they are expected to support.
