Executive summary
A changing division of work creates room to rethink the customer experience and the services an enterprise can offer. Leaders should define that future operation, then work backward to a transition they can sustain.
Today’s division of work can quietly become tomorrow’s design constraint. As AI and agents make different allocations possible, leaders have reason to reopen that choice.
In Microsoft’s 2026 Work Trend Index, 58% of surveyed AI users said they were producing work they could not have produced a year earlier. [1] Anthropic’s October 16 study describes employees building agent instructions and monitoring tools as some execution shifts to agents, alongside continuing human judgment and interpersonal work. [2]
These are survey responses and observations within one AI-intensive company, respectively. Neither establishes a universal future of work. They support reconsidering inherited assumptions.
My view is that the future operating vision should set the direction; current operating reality should shape the path and pace.
An organization can become so occupied with queues, policies, approvals and continuity that every AI proposal starts inside the existing allocation of tasks. Bottom-up improvement can produce useful local gains while leaving inherited handoffs and the overall design intact. Customers may still carry information between teams, while more useful services remain unexplored.
When the starting map assumes people collect, reconcile and move every case, AI is cast as an assistant to those steps. That can obscure a service in which agents coordinate routine execution within delegated limits and people direct the operation, own relationships and make consequential judgments.
Customer protection, clear authority and service continuity remain essential throughout. Essential obligations can outlast the procedures used to satisfy them: preventing unauthorized commitments remains non-negotiable even if the email chain used for approval changes.
Leaders should ask: If we were designing this operation today, around the outcomes our customers and business need, how would we organize the work using the capabilities now available?
That opens a larger ambition: a simpler customer journey, more responsive service, or support for customers whose needs were previously too costly to coordinate. The illustration shows how direction and evidence can work together to make that ambition practical.
Top-down direction starts with the desired customer and business outcome. Sketch the operation that could deliver it with people who perform, receive and understand the work. Work backward to responsibilities, knowledge, system access and other dependencies, then choose a feasible first intervention.
In suitable work, agents could execute more within defined authority. People would increasingly define outcomes and priorities, design and direct work, contribute domain knowledge and customer understanding, judge results, handle unfamiliar situations and improve the operation. This is what execution to orchestration should mean in practice. Some roles will remain centered on direct service or specialist practice; established automation may remain the best execution mechanism.
Consider a hypothetical business customer onboarding service. The desired experience is one coordinated start: customers provide core information once, see what remains outstanding and begin using an agreed service when requirements are met. Agents might gather authorized information, request missing details and coordinate routine setup through existing systems. People retain customer relationships and unusual commercial decisions; a designated person accepts readiness against the Outcome owner’s criteria. Required eligibility checks and approvals remain. The design could make tailored onboarding support feasible for customers currently served through generic instructions.
Initially, authorize agents only to assemble information and prepare a readiness recommendation for one standard service and customer group; people approve and carry out setup. Test missing documents, conflicting terms and failed handoffs; compare customer effort, elapsed time, errors and review workload. The results would inform whether and how to delegate routine setup.
Treat the future design as a working hypothesis. Evidence from employees, customers and actual operating performance should refine the path and, where warranted, change the allocation of work or service promise. Improvement means updating maintained knowledge, instructions, tools and workflows, then evaluating the changes against representative cases.
Leadership must make the choices individual teams cannot settle: the customer promise, the boundary of delegated action, priorities across functions and the capacity devoted to change. The Outcome owner needs authority and resources to coordinate the result; specialist decisions retain their proper owners.
Give knowledgeable employees protected time to design, test and learn. Explain changing responsibilities and provide practical development. Fund ongoing service, review and recovery as well as implementation, including a staffed fallback during transition.
The first intervention should test an important part of the future design while the existing operation remains dependable.
What would we design differently if today’s division of work were a choice rather than a fixed constraint?
Short, practical perspectives on AI-era operations, governance, and operating-model transformation.
Subscribe
RePerspective Labs