Michael Jaszczyk is CEO of NEWWORK Software, which provides a digital workforce platform that makes AI usable for enterprise organizations.
Not long ago, I watched an AI agent do something that summed up almost everything enterprise organizations get wrong about AI. The agent was given a real operational question and answered it well. It navigated to the right data, correctly identified the problem and produced a well-founded recommendation.
But then it stopped. With no next step to take, I saw how the intelligence is real, but it never translated into execution.
The gap between a good answer and tangible, finished work is where most enterprise technology, especially AI, either proves itself or falls short. This is the gap that most of today’s piloted AI projects ignore.
According to McKinsey’s 2025 State of AI survey, while a large majority of organizations now use AI in some form, just 39% report any measurable impact on enterprise-wide profits. In fact, most of those respondents said that less than 5% of their organization’s enterprise-wide profits were attributed to AI use.
As adoption spreads quickly, the benefits aren’t following—proof that current implementations don’t yet realize the potential of true enterprise AI.
Agents Alone Don’t Amount To Enterprise AI
As shown by the agent that produced an accurate and actionable recommendation, today’s models are not the problem. The first wave of generative AI was all about outputs like writing, summarizing information or searching and analyzing datasets. This changed how we work, but it did not change how enterprises run their business.
At the end of the day, a person still has to read the output, decide what to do with it, route for approval, update the right systems and notify relevant individuals downstream. The next wave of AI has to move beyond identifying the work. It has to complete it.
This is where things start to break down. Much of what is marketed as “enterprise AI” today is simply a capable model pointed at key company data. Agentic frameworks that can plan, reason and call upon tools or coordinate with other agents are helpful. But on their own, they are still a construction kit.
A construction kit only works when someone else supplies the operating model around it: who owns processes, which policies apply in different situations, which systems have authority over others, what data an agent has access to, what decisions require approval, how exceptions are handled and where recorded evidence is housed.
Inside a real enterprise, no one supplies this context by default. An organization may have a dozen working agents and still have nothing it can safely run on.
So what actually constitutes “enterprise AI”?
Enterprise AI Can’t Exist Without Five Critical Components
True enterprise AI comes down to whether a handful of capabilities exist together in a single, governed layer or are scattered across separate, siloed functions or tools. Enterprise AI must have these five critical layers:
1. Business context: Enterprise AI must understand the full business context behind how the organization operates, including its data, documented knowledge, policies and history of past decisions. Without that, it’s all guesswork.
2. Complete connectivity: Because enterprise work crosses finance, HR, service and a long tail of systems that no one wants to replace, connectivity is crucial. AI must operate across what’s already running. Fresh starts to consolidate all information on a single platform simply aren’t feasible for enterprises.
3. Governance: Governance is the true accelerator of AI value. Enterprise AI must be able to distinguish a person from an agent, know what each is permitted to do, scope every agent’s authority, keep an accurate audit trail and enable users to revoke access when needed. When these controls exist, an organization can decide where AI acts on its own, where it needs sign-off and where a human remains as the decision maker.
4. The workflow state: A model reasons well, but enterprise processes run long, they branch and they stall and resume. A credible enterprise AI platform must separate the reasoning engine from the execution fabric. A workflow layer has to carry out a task through all the necessary steps. Without this layer, every meaningful business process becomes an exception someone must clean up by hand.
5. The operating surface: The final layer, the operating surface, is where enterprise AI becomes visible to people. Effective enterprise AI needs one place where employees ask for what they need, managers handle approvals and process owners inspect decision-making.
Implementing enterprise AI doesn’t require the business to believe that every decision should be autonomous. The point is to let organizations decide, on a process-by-process basis, how much autonomy is appropriate and how much human judgment to retain.
Enterprise AI, Defined
Enterprise AI begins when AI leaves the prompt box and enters the operating model. Prompt-based AI supports research, summarization and content generation, and recommendation-based agents can identify the next best action. But true enterprise AI goes further by executing work inside of an operating model.
Enterprise AI is defined less by whether an enterprise has agents up and running and more by whether those agents can work inside the business with full context, permission and policy knowledge, with real workflows and records of what they accomplish.
Once that foundation exists, AI pilots can finally begin moving to full production. Work will move without waiting on a dozen handoffs, and processes will improve because each outcome informs the next.
For executives evaluating the success of current projects and pondering where to go next, these possibilities raise a sharper question than “do we have agents?” Ask whether your AI can operate inside your enterprise, under total control, and actually finish the work it starts. Until it can, all you have is intelligence. True enterprise AI unlocks the execution.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
