AI can help people draft faster, search faster and produce more. But what happens when the work they accelerate sits inside a workflow that was already slow, fragmented or poorly designed?

A proposal gets written in minutes, then waits days for approval. A report takes seconds to generate, but leaves the next team struggling to understand what decision it supports. More work gets completed, yet customers experience the same delays.

The individual gains may be real. The business outcome may barely change.

For leaders steering AI initiatives, this distinction matters. Improving the performance of a task and improving the performance of a business require different questions, different interventions and different measures of success.

Two horizons, two different ambitions

We can think of these as two horizons.

Horizon 1 focuses on personal productivity. It equips individuals with AI assistants that help them complete tasks within the existing way of working.

Horizon 2 focuses on workflow performance. It examines how work moves across the business and redesigns that flow using data, AI, software and human judgment.

Both can create value. But scaling Horizon 1 does not automatically deliver Horizon 2.

An organization depends on the connections between people as much as on their individual capabilities. Work passes through teams, systems, decisions and approvals before it becomes an outcome someone values.

At every handoff, something can go wrong. Information may be incomplete. Ownership may be unclear. Different teams may use different definitions of “ready.” An approval may sit in a queue because the person responsible is managing several competing priorities.

Making one task faster leaves those constraints untouched unless the intervention also changes what happens around it.

Faster proposals, the same customer wait

Consider a hypothetical sales team using AI to prepare customer proposals. Account managers can now produce a first draft in a fraction of the time. From their perspective, the improvement is substantial.

But each proposal still requires pricing confirmation, a delivery assessment and approval for any unusual terms. The information needed for those decisions is scattered across systems. Reviewers repeatedly send drafts back for clarification.

The sales team produces proposals faster. The review queue grows. Customers still wait.

There may still be a worthwhile benefit: account managers have more time for customer conversations. But the business cannot assume that faster drafting means faster sales, more revenue or lower delivery risk. Those outcomes depend on the rest of the workflow.

A Horizon 2 initiative would examine the entire path from customer request to an approved, deliverable offer.

It might connect the information needed for pricing and delivery decisions, establish standard requirements for a complete proposal, and route exceptions to the right person. AI could help assemble relevant information, identify omissions and support drafting within that redesigned process.

Success would then be assessed through measures such as time to an approved offer, the amount of rework and the accuracy of delivery commitments.

The drafting task remains important. Its contribution is now connected to a business outcome.

The handoff determines whether a gain travels

This also changes how we think about output quality. A document is useful when someone can reliably act on it. Producing it quickly is only part of the equation.

If an AI-assisted output contains unsupported assumptions or misses essential context, the next person may have to spend additional time checking and correcting it. A saving at one desk becomes extra work at another.

That possibility makes the handoff a critical design question. What does the recipient need? What must be checked before the work moves forward? Who is responsible when something is incomplete or uncertain?

Answering those questions helps turn individual assistance into dependable organizational capability.

Why workflow redesign requires leadership

This also explains why Horizon 2 can be harder to pursue. Providing people with a tool can often fit within existing responsibilities. Redesigning a workflow requires agreement across them.

Teams may need shared definitions, access to common data and clarity about who can make which decisions. Leaders may need to resolve conflicting priorities or change measures that reward local output while overlooking delays elsewhere.

AI does not resolve those coordination problems simply by becoming more capable. They require management attention.

Start with a business problem and follow the work

The practical starting point is a workflow with a meaningful business problem: customers waiting too long, repeated rework, unreliable decisions or an expensive backlog.

Trace the work from the initial need to the completed outcome. Look for where it waits, where information is lost and where people repeat work because an earlier step was incomplete. Then determine which changes would improve the result and where AI can contribute.

This approach also makes measurement more useful. Tool adoption and reported time savings can indicate progress, but they need to be connected to what the organization wants to achieve.

If the goal is faster customer resolution, measure the time to a resolved issue and whether it stays resolved. If the goal is more efficient delivery, examine rework, quality and the cost of completing the workflow. If the goal is additional capacity, establish how the freed time will be used.

The central question is whether a local improvement survives the handoffs and reaches the customer or the business.

Make the next AI initiative count

Personal productivity is a valuable ambition. It can relieve pressure, improve tasks and create room for better work. Leaders should pursue those benefits with a clear understanding of what they can accomplish.

They should also make a deliberate choice about when the workflow itself needs to change.

For your next AI initiative, start with the outcome you want to improve. Follow the work across the organization. Identify the constraint that prevents a better result.

Then ask whether AI will help remove that constraint—or simply help people reach it faster.