A business can look well integrated because a few exceptional people keep its work moving. The customer receives an answer, the transaction closes, and the monthly report arrives. What leadership may not see is the employee who reconciles conflicting records, chases an approval, or translates between departments before any of those outcomes become possible.
I would treat that dependence as an operating-model question. Experienced employees should apply judgment where judgment matters. When they repeatedly repair the same routine handoff, leadership needs to understand what the process is asking them to compensate for. More software may help, but only after the underlying responsibility and information gaps are understood.
Follow one piece of work
Choose a recurring request with a clear business result: onboarding a customer, approving a supplier, or resolving a service issue. Follow it from the first request to acceptance by the receiving team. Ask people to describe what actually happens, including the messages, spreadsheets, and informal checks that do not appear in the process diagram.
For each handoff, identify the information required, the person accountable for providing it, and the decision the next team must make. A timestamp showing that a request was forwarded is useful. It does not establish that the next team received something usable. The important boundary is acceptance: can the receiving team proceed without reconstructing the request?
Separate expertise from routine repair
Imagine a customer manager who repeatedly stops other work to reconcile an account identifier before finance can issue an invoice. That example is hypothetical. The intervention could reflect inconsistent identifiers, unclear ownership of a master record, or a legitimate exception that requires commercial judgment. Treating all three as an integration defect would miss important differences.
Ask what the experienced person adds on each occasion. If they supply missing information, fix the information path. If they supply authority, define the decision right and its limits. If they resolve a genuinely unusual tradeoff, preserve access to their judgment. The aim is to stop using the same person's availability as the answer to several different organizational problems.
Assign an owner to the boundary
Departmental accountability can leave a gap between departments. One team may consider its task complete when it sends the request, while another considers its responsibility to begin only after someone explains what the request means. Each team can meet its local measure while the customer still waits.
I would assign an accountable owner for that boundary and agree on a short acceptance standard with both teams. What must be present? Which exceptions can be resolved locally? Who decides when information conflicts? Where does a rejected handoff return? The owner does not need to perform every task personally. They need authority to resolve the recurring failure and visibility into whether the revised arrangement works.
Make the investment decision concrete
A proposed integration should describe the interruption it removes and the work it enables. Connecting two applications is a technical deliverable. Removing repeated re-entry, reducing avoidable corrections, or making an approval possible without a personal introduction is an operating benefit. Those benefits need their own acceptance criteria.
The same issue appears in security operations when an investigation depends on one experienced employee to reconcile information across tools, teams, or approval processes. Leadership should distinguish the judgment that genuinely requires expertise from the recurring coordination work that a better-defined operating model could address. The objective is not to remove human judgment. It is to make the process less dependent on one person's constant intervention.
Include support, exception handling, access control, and ongoing maintenance in the investment discussion. A connection that works only when its original developer intervenes has moved the dependency rather than resolved it. For AI-enabled handoffs, NIST's AI RMF emphasizes clear roles and communication concerning AI risk [N1]. That supports explicit responsibility; it does not prescribe this article's process design or promise a return on investment.
Protect the knowledge while changing the process
Do not begin by telling the indispensable employee that their involvement is the problem. They may have prevented repeated customer failures by making an incomplete process work. Ask them to help identify the recurring patterns, the exceptions that deserve attention, and the decisions that should stay with an experienced person.
Make that work visible and provide time for it. Otherwise, the organization can ask someone to keep rescuing the process while also documenting how to stop needing the rescue. Recognize their contribution to a better operating model. The desired result is a more resilient process and a more useful role for their expertise, with less dependence on their constant availability.
Test the revised handoff
Run a small pilot with an agreed set of routine requests and representative exceptions. Record rejected handoffs, repeat contacts, corrections, and customer completion rather than counting only the messages sent or interfaces connected. Check whether another qualified employee can follow the documented route when the usual troubleshooter is unavailable.
Do not infer financial savings simply because interruptions fall. Determine whether the released capacity was used, which work improved, and whether service quality held. A useful review can reveal that the process is easier to operate without yet supporting a cost-reduction claim. That is still evidence leadership can use to choose the next investment.
Consider the information burden at the handoff as well. Ask the receiving team to identify which fields it actually needs and who is responsible for correcting them. Sending a larger attachment is not a substitute for agreeing on an authoritative record. Give the integration owner a way to resolve conflicting entries and tell both teams which version governs. During the pilot, review a few rejected requests together and determine whether the acceptance standard was unclear, the required information was unavailable, or an employee made a genuine mistake. Each finding points to a different intervention and a different owner.
The executive question is where personal intervention belongs. I would protect it for the decisions that require experience and remove the routine ambiguity that keeps pulling experienced people away from them. Start with one recurring interruption, establish an accountable handoff, and verify that the business can complete the work without calling the same person every time.
Sources and References
[N1] NIST AI Risk Management Framework 1.0, Core https://airc.nist.gov/airmf-resources/airmf/5-sec-core/
2023 framework; accessed November 12, 2026. Scope: GOVERN 2.1 clear responsibilities; GOVERN 1.5 periodic review; GOVERN 1.7 safe decommissioning. Voluntary AI-specific guidance. Broader business applications in these drafts are author recommendations, not NIST requirements.

