Getty Images/iStockphoto
Who owns the workflow? Why enterprises are managing the wrong thing
Enterprises manage applications well but often lack owners for cross-functional workflows. AI automation reveals that application ownership alone doesn't ensure business outcomes.
A few years ago, I was hired to review the systems of a large enterprise that had invested heavily in modernization. On paper, the company looked disciplined: The ERP platform had a strong owner, the CRM stack had its own leadership team, and HR systems were well governed. Yet customer onboarding still took weeks. Everyone assumed this was a technology problem, but it was not. The real issue was that no one owned the workflow end to end.
Sales blamed compliance, compliance blamed operations, and IT kept pointing out that the systems were working as designed. The problem? They were all right. The systems functioned, but the business outcome was failing. When the company introduced AI-based automation, the bottlenecks became even more visible. AI did not create the dysfunction; it exposed that the enterprise had owners for platforms but not for performance.
Enterprises have become very good at assigning ownership to software. There is a "lead" for ERP, a "lead" for CRM, a "lead" for HR systems and others for procurement, finance and customer service platforms. Those ways of owning technology are both familiar and rational, and they are deeply ingrained into how organizations fund, govern and operate technology. For years, they did OK because the main management problem seemed to be getting these big systems up and running and keeping them running.
But businesses do not generate value with single applications, isolated applications or even with suites of applications. They generate value through the workflows that run across and through them. That is where the enterprise accountability problem begins.
Accountability, fragmentation and the hidden cost
When no one owns the process, organizations pay for fragmentation through hidden costs that are often normalized. The pattern recurs everywhere. Customer onboarding stalls because sales, compliance, finance, legal and ops each own a piece of the process, but no one owns the entire path. Recruiting moves slowly because HR, IT, security and the hiring manager operate on different time scales and measure success differently. Procurement bottlenecks arise when business requesters, sourcing teams, finance approvers and supplier data managers prioritize their own needs and when no one is responsible for the overall cycle time.
The simplistic explanation is complexity. Complexity exists, but it is sometimes a euphemism for a more accurate yet less politically correct diagnosis. Many enterprise workflow challenges are more than technical problems; they are ownership defects. The enterprise has not decided who is responsible for the outcome when work crosses systems and functions. Without that clarity, friction persists in areas that exist between formal lines of responsibility.
That is why the prominent discussion about AI today is often misplaced. AI did not invent the accountability problem; it merely revealed it. Agents, orchestration platforms and automation technologies are challenging enterprises to look at the real work because these technologies span the very boundaries that traditional governance models tend to ignore: systems of record, systems of instruction, networks, environments, clouds and regions.
An AI agent might interface with numerous systems, aid in decision-making, invoke actions and handle exceptions across processes. That is exactly what makes it useful. That is also what makes gaping ownership holes impossible to obscure. If no one has real ownership of the end-to-end process, AI does not eliminate fragmentation. Instead, it collides with fragmentation. Dissonant policies are revealed more clearly. Ambiguous decision rights harden into something that cannot be wished away. Escalation paths that used to be handled informally must increasingly be formalized if automation is to be trusted.
The same is true of orchestration. Orchestration tools are more than system connectors; they reveal the seams between systems. They illustrate where data definitions conflict, where approvals are ambiguous, where process logic collapses under exceptions and where no one has the right to redesign the entire flow. In that regard, AI and automation are not so much the source of a new crisis of accountability as its stress test.
Why application-centric ownership breaks down in a workflow economy
The bigger problem is that the company now operates in a workflow economy but has an ownership model designed for an application economy. Applications are stable units of governance. Workflows are flexible, cross-organizational and outcome-focused. Applications are monitored for uptime, release quality, compliance, adoption and other metrics. Workflows are measured by cycle time, quality, conversion, cost-to-serve, customer satisfaction and resilience.
The real distinguishing factor is not whether a company runs ERP, CRM, HR and procurement systems; it is whether those systems can deliver the rapid, reliable and flexible business processes needed. AI makes flexible business processes more urgent because it relies on defining workflows. It delivers the greatest value not within a single tool but across a sequence of work.
That means ownership must shift up a level. While system owners still exist, they alone are insufficient. Organizations need owners who are responsible for entire workflows -- especially those that significantly influence growth, efficiency, compliance and customer experience. Otherwise, they will continue managing components and fail at the level where results are generated.
What genuine workflow ownership looks like
True workflow ownership is not about creating yet another committee or replacing application leaders but about holding someone accountable for where value is generated. A workflow owner is responsible for the end-to-end performance of the process, including its metrics, exceptions, redesign priorities and inter-organizational coordination. It is time to identify a person with sufficient authority to convene stakeholders across organizational lines and enough line of sight to identify where the process is breaking down.
This office is not intended to replace the functions of system owners. Platform leaders still maintain responsibility for the platform's integrity, security, capabilities, vendor relationships and technical road maps. Workflow owners have a business focus within those systems and are accountable for business performance. This is the difference that makes the difference: One role keeps the tools healthy. The other ensures that the work moving through those tools accomplishes what it is supposed to.
Too many organizations assume that accountability will naturally emerge from this collaboration. Sometimes it does, as when transformation programs are tightly monitored by executive leadership. But such informal cooperation seldom withstands the day-to-day operating pressures. Strong accountability must be baked into the operating model's system design, with clear decision rights and shared outcomes-based metrics for success.
When companies come to realize the gap, they often find that addressing it is more than just naming a new owner. Governance needs to articulate who is empowered to make cross-functional process decisions. Metrics for system-level reporting need to shift toward end-to-end outcome measures. Incentives need to discourage local optimization that damages the broader flow of work. Process visibility and process redesign must be enabled by the technology architecture, rather than the architecture becoming yet another band-aid applied over fractured process layers.
And this is where AI's many strategies do or do not work. When an enterprise uses AI in a workflow with no clear ownership of the entire workflow, it typically sees success from experiments with limited strategic value. This often involves isolating one part of a process and treating it as an experiment in a box. It might enhance certain tasks within the box, but it will not reshape the entire process, most of which takes place outside the box. In contrast, when workflow ownership exists, AI can be directed toward specific goals within defined point-to-point boundaries, with established escalation procedures and quantifiable performance metrics.
That is why we believe workflow ownership is not optional: It is required if an enterprise is to develop mature AI rather than tinker around the edges. Organizations will successfully scale AI by making the underlying work governable. Making it "smarter" will come later.
Why does this matter?
The enterprises best positioned to capture AI value will ask, "Who owns the outcome when the work touches everything?" This question goes beyond production quality, exception handling, trust, compliance and performance issues. Managing applications is still necessary, but no longer sufficient. The next era of transformation involves governing workflows end to end, a task once reserved for platforms. AI did not create this need, but it does reveal that owning applications is no longer enough to run modern operations.
Can enterprises implement intelligent technology? Most already have. Here is the real question: Will they redesign accountability to show how and where value is created? Until then, enterprises will continue to assign responsibility to systems and expect performance from workflows. That divide is where sensing intelligence is currently draining a large share of its value in enterprises.
Someone needs the authority to step back and look at the forest rather than the individual trees. This digital forest ranger will be responsible for identifying and addressing areas across the workflows (forest) that need attention or improvement. My goal is to give you an idea of what your ranger's job description might look like.
David Linthicum is a globally recognized thought leader, innovator and influencer in AI, cloud computing and cybersecurity. Prior to 2024, he spent more than 30 years at Deloitte Consulting LLP, where he most recently served as the managing director and chief cloud strategy officer.
