iterthink On The Emerging Need For Human Accountability In AI-Driven Document Workflows

Simon Dilhas, CEO and co-founder of abstract, believes the rapid adoption of AI agents across enterprise operations is bringing renewed attention to human accountability. In his view, long-term organizational value may come from maintaining visibility, responsibility, and traceability across AI-assisted work, particularly as documents increasingly influence critical decisions. Within that context, iterthink, one of abstract’s software products, has been developed as a review layer designed to help organizations understand how documents evolve as people and AI systems contribute to them.

“AI is blending into daily business tasks, and it’s changing how we handle documents. Drafts, reports, contracts, and project files can be put together and improved much faster now, which helps teams keep things moving,” Dilhas says. He observes that this growing efficiency is often accompanied by a subtle shift in behavior. Because AI-generated content frequently appears polished and coherent, teams may place greater emphasis on speed and presentation while dedicating less attention to the reasoning, assumptions, and factual integrity behind the text itself.

For organizations operating in environments where documentation influences financial, operational, legal, or project outcomes, Dilhas notes that this introduces an important governance consideration. For him, a document often represents decisions, commitments, requirements, and responsibilities. “Understanding exactly what changed and why becomes more valuable when AI systems participate in creating or revising those documents at scale,” he states.

Dilhas frequently returns to a concept inspired by an English anthropologist, who described information as the difference that makes a difference. Through that lens, information gains significance when a change produces a meaningful consequence. A modified sentence, adjusted requirement, or revised recommendation may appear minor during review, yet its downstream implications can extend far beyond the document itself. From this perspective, Dilhas emphasizes that visibility into change becomes an essential part of understanding information.

This growing focus on visibility can help explain why, as Dilhas observes, document review is attracting renewed attention among compliance leaders, legal professionals, CISOs, and operations executives. “One recurring concern is about preserving organizational intent and expertise as AI becomes a larger participant in content creation,” he shares. “Another relates to subtle alterations in facts, recommendations, or obligations that can emerge during automated editing processes.” Small textual revisions may influence meaning in ways that require careful human interpretation, particularly when documentation informs high-value decisions.

As a result, responsible AI adoption increasingly appears connected to governance structures that support human judgment. Dilhas says, “Oversight begins with traceability. Organizations may benefit from understanding who initiated a change, how the modification occurred, and which individual approved it.”

That philosophy is reflected in Iterthink’s design. The platform focuses on making document evolution visible through detailed comparison and review capabilities. Instead of emphasizing generation alone, it aims to understand modifications between versions and help teams evaluate their potential significance. The objective is to create an environment where AI participation remains observable, interpretable, and subject to human approval.

For Dilhas, the significance of this challenge extends beyond document management itself. As AI becomes embedded across more business functions, organizations seem to rethink how human expertise and automated systems work together. In that environment, the quality of decision-making may depend on an organization’s ability to understand the information flowing through its processes, especially when AI contributes to creating, modifying, or interpreting that information.

That shift is influencing how Dilhas views the next phase of enterprise AI adoption. He anticipates business environments where multiple AI models operate alongside one another, with teams selecting different systems based on factors such as cost, security requirements, availability, or task complexity. In his view, organizations may eventually treat AI models much like specialized professional resources, matching each task with the most appropriate tool for the job.

As that flexibility expands, Dilhas believes attention may increasingly move toward the systems that manage information around those models. “Access to powerful AI capabilities is becoming more widespread, making the surrounding infrastructure, governance processes, and information controls more important,” he says. “That means organizations may value understanding how information moves, changes, and influences decisions throughout a workflow even more.”

This perspective also informs Dilhas’s interest in ownership, transparency, and control. He points to the benefits of environments where organizations maintain direct stewardship over their data, files, software, and operational knowledge. Amid growing adoption of SaaS platforms and token-based consumption models, he sees increasing interest in software architectures that provide greater visibility into how information is stored, managed, and used. “The closer critical knowledge remains to the people responsible for decisions, the easier it becomes to understand, challenge, and improve those decisions,” he remarks.

That philosophy extends across abstract’s broader portfolio. Whether working with Building Information Modeling (BIM) data, project requirements, or documents, the company’s focus has consistently been on making information more usable, reliable, and traceable. The company is currently aligning the underlying data models of Iterthink and {yourcompany}os, another core product in abstract’s ecosystem, to help create a more consistent information layer. This can enable project data, requirements, and document insights to move more seamlessly between local and online workflows. Iterthink represents an extension of that thinking into written content, helping organizations examine how information changes over time and how those changes may influence subsequent decisions and actions.

For leaders accelerating AI adoption, Dilhas offers a fresh perspective. Discussions around implementation often focus on model selection, productivity gains, and deployment strategies. He encourages organizations to dedicate similar attention to the information governance structures that support those initiatives. Review processes, visibility into change, and clearly defined accountability mechanisms can help create conditions where human judgment remains an active part of AI-assisted work.

“The future of AI may depend on helping people see,” Dilhas states. “Organizations can gain a stronger foundation for judgment when every important change can be understood in context.”