

Scaling in the Age of AI: Why organizations must redesign work, not just automate it
Scaling in the Age of AI: Why organizations must redesign work, not just automate it
A company spends years building the systems it needs to scale. It introduces functional heads, clearer roles, standard processes, approval thresholds, reporting rhythms and managers who coordinate work across teams.
Then AI arrives.
Some work that required specialists can now be prepared by a generalist with AI. Some reports and coordination activities no longer need a manager. Agents can act on information, not just summarize it. And decisions can move faster than the organization’s existing approval system was designed to handle.
The company is not simply adopting another technology. It is revisiting the management system it built for growth.
That is the central argument of our research report, Scaling in the Age of AI. Organizations face two overlapping journeys: scaling the business and redesigning work as AI changes the capacity and contribution of people, teams and digital agents.
The second scaling curve
The first scaling journey changes how an organization works. A small team can rely on shared context, informal roles and direct access to founders. As it grows, it needs explicit priorities, decision rights, functional depth, workflows, controls and ways to develop leaders.

AI changes those same design choices while they are still evolving. It increases the speed and volume of work. It changes who can produce an analysis, who can act on a recommendation, and how information moves across functions.
This is the second scaling curve. AI can accelerate a strong operating model. It can also multiply the weaknesses of a poorly designed one. A faster handoff does not help if nobody owns the complete outcome. More recommendations do not help if decisions still wait in the same queue. And a fluent AI answer does not eliminate the need for domain judgment.
Start with outcomes, not automation
The tempting approach is to list tasks, estimate what AI can automate, and translate that into headcount or productivity targets. We think that starts too far downstream.
Begin with the business outcome. Identify what the workflow must deliver. Then define the full contribution of each role: decisions, judgment, relationships, expertise, execution and the ability to handle exceptions. Only then decide what people, assistants and agents should do.
A role may contain routine activities that AI can perform, while also carrying institutional memory, customer trust, exception handling and the apprenticeship through which future experts develop. Removing the task is not the same as replacing the contribution.
The report sets out a six-step design sequence: business outcome; performance outcomes; role contributions; task and decision allocation; design mechanisms; and review triggers.
Ten dilemmas leaders cannot avoid
There is no universal AI-ready organization chart. Leaders must resolve competing priorities that are both legitimate. Our Dilemmas Resolution Framework (DRF) helps make these choices explicit, including the benefits sought, costs accepted, controls required and evidence that should trigger a review.
The report explores ten dilemmas in depth:
1. Standardize the work or preserve local adaptation.
2. Centralize decisions or distribute authority.
3. Build functional depth or create end-to-end ownership.
4. Flatten the hierarchy or strengthen people leadership.
5. Keep stable roles or deploy fluid skills.
6. Build deep expertise or expand AI-enabled breadth.
7. Capture productivity or build future capability.
8. Move quickly with experiments or build reliable operations.
9. Expand agent autonomy or preserve human accountability.
10. Open access to knowledge or protect confidentiality.
These decisions interact. An organization can make an apparently sensible choice on each and still create a system that is incoherent overall.
Three dilemmas worth discussing now
Will AI make middle managers redundant? Some reporting, routing and follow-up work can disappear. But the need to translate context, coach people, redesign work and own difficult exceptions may increase. The right question is what each management layer contributes, not simply how many layers to remove.
What happens when AI performs entry-level work? First drafts, research and routine cases have historically produced output and helped junior colleagues develop judgment. If those experiences disappear, organizations must deliberately create new routes to expertise.
Who is accountable when an agent acts? Naming a human reviewer is not enough if that person cannot inspect evidence, change permissions, stop the agent or meaningfully review its output. Authority, observability and accountability have to be designed together.
What leadership teams can do next
Choose one important customer or business outcome. Map the end-to-end workflow, including handoffs, rework, exceptions and decisions. Identify where AI changes capacity or decision speed. Then ask which organizational dilemmas become harder or easier.
Record the position you choose, what it will improve, what it may compromise, who owns the outcome and what evidence will cause you to reconsider.
This is not a one-time organization restructuring exercise. It is an ongoing discipline of redesigning work as business conditions and technology change.
Read the research report
Scaling in the Age of AI is a 25-page research report by Sirisha Bhamidipati and Harish Bhamidipati of Align By Design. It examines the second scaling curve, an expanded landscape of organizational dilemmas, ten dilemmas in depth and a first-principles redesign process.
READ THE FULL RESEARCH REPORT (PDF)
If you are considering how AI changes your organization’s roles, management system or decision rights, we welcome a conversation about the dilemmas you are facing.
DISCUSS AN ORGANIZATION DESIGN QUESTION [Get in touch]
INVITE SIRISHA OR HARISH TO SPEAK [Get in touch]
Authors: Sirisha Bhamidipati and Harish Bhamidipati, Align By Design.
