Department-wide redesign produces larger value

The speaker describes a change in the main enterprise bottleneck. In his account, execution was still the difficult part in 2024: even when a business knew what it wanted, getting an AI system to complete the work was a major constraint. He argues that the situation has moved on. The harder question is now:

How should the work be designed so that AI can perform it well?

This is not mainly a question of choosing a better model. It is a question of designing an operating process: which steps should be automated, which should include a human check, and which should remain human because of their risk or because automation would add little value.

The speaker presents this as his view of the changing bottleneck, not as an independently measured fact about every company. His broader framing is that AI execution and knowledge work have advanced substantially. The remaining challenge is to understand a particular business deeply enough to redesign its work around those capabilities.

Point solutions versus department transformation

A point solution improves one narrow task. For example, a company might automate one part of prospecting or one part of accounts-payable work. That can be useful, but the task does not exist in isolation. It may depend on information from another task, require a handoff, or affect a later decision. Improving one point can therefore leave much of the surrounding department unchanged.

A department-wide transformation asks a larger question: how should the department work as a connected system? Instead of placing an agent beside one existing task, the team examines the related workflows and decides how the whole set of activities should be divided between AI and people.

The distinction can be represented like this:

Point solution
    one task -> automate or assist -> local improvement

Department-wide redesign
    connected workflows -> redesign dependencies and handoffs
                       -> allocate work between AI and people
                       -> measure value across the department

The second approach does not mean that every step becomes autonomous. It means that automation decisions are made with the surrounding work in view. This connects directly to the earlier FDE process: first discover how the department actually operates, then re-engineer its workflows, and finally deploy agents into the customer's existing environment.

Why broader scope can create more value

The speaker's reasoning is that value can appear in several different forms at once:

Value category	What it means in general	Example of the question a redesign asks
Revenue uplift	The business generates more revenue.	Can the redesigned work help the revenue process perform better?
Cost savings	The business spends less time or money on the work.	Which connected activities can AI handle or accelerate?
Risk mitigation	The business reduces the chance or impact of an unwanted outcome.	Where should a human review remain part of the process?

These categories are distinct. A transformation does not need to produce value only by reducing headcount or operating cost. It might also improve revenue or reduce risk. The talk does not provide a formula for combining the categories, so they should be treated as separate ways to evaluate an outcome.

The causal idea is simple: when tasks are connected, the total result may depend on more than one task. A narrow improvement can be blocked by an unchanged handoff or dependency elsewhere. A department-level redesign can consider those relationships together. This does not prove that a larger project will always have a better return. It explains why the speaker believes a larger scope can expose more opportunities than a single-task automation.

The reported return ranges

The speaker contrasts the returns he reports for local automation with the returns he reports for broader transformations. He says point solutions often produce roughly 5–10% ROI, while department-wide efforts have produced figures such as 25%, 50%, or 75% ROI.

Those numbers are claims about reported client outcomes in the talk. No baseline, measurement period, definition of ROI, or method for separating revenue uplift, cost savings, and risk mitigation is supplied. They should therefore be read as the speaker's evidence for his business argument, not as a universal rule that department-wide redesign always produces one of those percentages.

The practical lesson is not to assume that a bigger scope automatically creates value. The lesson is to test whether the department's processes are sufficiently connected that a local fix would miss important dependencies. If they are, the FDE's work should cover the wider system rather than optimizing one isolated task by default.

From business case to technical design

This argument prepares the transition to the next part of the talk. Once the company chooses department-wide redesign, it needs a way to preserve the context discovered by FDEs and strategists, represent the workflows, and help build agents over them. The next speaker gives the technical account of that proposed FDE agent.

The division of labor remains important: the agent is intended to support the work of understanding and redesigning operations, while the FDE motion supplies the customer-specific context. The goal is not merely to add an AI feature to one task. It is to make a department's connected work operate better around AI.