Three months ago, fresh from the Knowledge 2026 show floor, I wrote about 11 things the event had quietly told us about enterprise AI.
(Okay, “quietly” might not be the right term, since Knowledge was loud in nearly every sense. But, apparently, they … ahem … want it that way.)
Now, we know that the ServiceNow ecosystem changes at a breakneck pace, so it’s understandable that even the most concrete announcements might have shifted a bit over the course of Q2. In early May, we were thrilled to reinforce that enterprise AI was moving beyond experimentation, toward governance, context, and workflow orchestration.
Three months later, we can report that most of the observations we shared still look accurate. A few have become more concrete, while others have shifted in ways that deserve a closer look. Here are a few that now have enough evidence behind them to revisit.
Verdict: More accurate by the day.
One of the original article’s clearest conclusions was that AI governance would need to scale alongside AI adoption. In the months since Knowledge, ServiceNow has continued turning that idea into a more detailed product architecture.
By June 2026 AI Control Tower release included generally available capabilities for discovering models and agents in Databricks, Snowflake, and Hugging Face; classifying AI assets by risk at intake; governing customer-selected model providers; monitoring agent deviation; screening outputs for sensitive data and security vulnerabilities; and managing approval and lifecycle workflows for AI assets.
ServiceNow has also added prebuilt compliance content for laws including the EU AI Act and state-level AI legislation. That does not make regulatory compliance automatic, and organizations should be skeptical of any suggestion that software alone can do so. It does show that governance is moving closer to day-to-day workflow execution, where approvals, evidence, incidents, exceptions, and remediation can be managed continuously.
Verdict: Accurate, with a notable expansion.
The original piece argued that models generate excitement, but workflows generate outcomes. ServiceNow’s more recent announcements have reinforced that position while expanding it beyond agents built directly on the ServiceNow platform.
Through ServiceNow Action Fabric, the company says external agents, including agents built with systems such as Claude, Copilot, or a customer’s internal technology, can initiate governed ServiceNow workflows without operating through the traditional ServiceNow interface.
In May, it was reasonable to think about ServiceNow primarily as the environment in which enterprise agents would be built, governed, and deployed. The emerging model is broader: agents may originate in many different platforms, while ServiceNow provides access to the workflows, business rules, approvals, records, and guardrails required to complete work.
This supports the original orchestration argument, but it also introduces a new level of complexity. Companies are not simply managing one collection of native AI agents. They may need to govern an ecosystem of agents from multiple vendors, models, development tools, and business units, all attempting to interact with the same enterprise processes.
Verdict: Accurate, but incomplete.
At Knowledge, context was discussed as the link that helps AI understand organizational structures, relationships, ownership, historical behavior, and business intent. No arguments there.
ServiceNow’s Action Fabric announcement explicitly connects its Knowledge Graph and Context Engine with workflows, playbooks, and business rules. Understanding a situation and being authorized to act on it are not the same thing. An AI agent may correctly identify that a new employee needs a laptop, account access, payroll enrollment, and training assignments. It still needs access to governed processes that determine what can be ordered, who can approve it, which systems can be changed, and what evidence must be recorded.
Enterprise context now includes the operational boundaries within which that information can be used. So, yes, context may still be the entire game. But the game now comes with referees, permissions, audit trails, and an extensive rulebook.
Verdict: Accurate, although autonomy is advancing faster than expected.
The original article resisted the idea that credible enterprises were preparing to remove humans from workflows entirely. That remains a fair assessment, particularly in sensitive or regulated work.
However, more recent customer announcements show that organizations are setting increasingly ambitious autonomy targets for routine work. One example shows a company expanding its ServiceNow deployment across IT, HR, and shared services and is targeting up to 60% autonomous resolution of IT support requests.
Humans may be removed from large volumes of repetitive work while remaining responsible for exceptions, approvals, security decisions, complex cases, and overall system accountability. That is consistent with the original claims, but the dividing line between autonomous and human-led work is moving as organizations gain confidence.
Verdict: Accurate, with more tension than the original suggested.
The final broad argument in May was that enterprise AI success would depend on connecting models, data, context, governance, workflows, and people into a coherent operating environment.
ServiceNow’s recent activity supports that case. Its June partnership with Accenture combines agentic AI with managed security services, risk workflows, regulatory monitoring, OT and IT risk management, and tools intended to help companies migrate away from legacy risk platforms.
But the broader story is not completely settled. Recent coverage has also highlighted investor concern that AI could reduce traditional software seat counts, change pricing models, and allow new competitors to bypass established application interfaces. ServiceNow is responding by positioning itself more heavily around orchestration, governance, security, and consumption-based AI usage.
That makes our original “connect every dot” observation feel right, but for a slightly different reason. Connecting the dots not only provides a path to better AI, but may be how enterprise platforms defend their relevance in a world where users increasingly begin with AI assistants rather than application menus.
The primary themes from Knowledge 2026 have largely held. Enterprise AI is becoming more operational, governance is becoming more technical, and workflow integration remains central to producing meaningful outcomes. The biggest change (yes, even over a few months) is speed.
In May, many of these ideas were presented as platform direction. Today, ServiceNow is reporting increased agentic deployments, expanding cross-platform governance, opening workflows to externally built agents, and pointing to large customers preparing for significant levels of autonomous work.
At the same time, it is important not to confuse announcements, product availability, deployment growth, and projected savings with universal customer success. The market is still working through data quality, adoption, security, operating models, pricing, and the difficult question of which workflows should become autonomous in the first place.
So, all of these words to say that our original conclusions hold, with one minor amendment: the winners will need to connect the dots, but also confirm that the dots are accurate, securely connected, and owned by someone who knows how to operate the "off" switch.