AI’s second wave is about everything you ignored

When execution is slow, inefficiencies elsewhere are easy to ignore. Communication gaps, unclear requirements, and weak decision-making don’t stand out; they’re absorbed into the overall pace. 

Once execution becomes fast, those hidden problems surface immediately, because there is no longer slack in the system to hide them. 

Take business analysis quality. 

If requirements are vague or inconsistent, teams traditionally work through them via back-and-forth. A task that could have taken one week stretches to two. It’s inefficient, but manageable, and not painful enough to demand change. 

Now introduce AI. 

With clear, high-quality input, the same task might take half a day instead of a week. But if the requirements are flawed, the back-and forth still adds a week. Suddenly, the difference isn’t 2x - it’s 10x or more. 

What used to be a tolerable inefficiency becomes an obvious bottleneck. 

The same applies to communication. 

Previously, teams had time to correct misunderstandings between roles or across different teams. Delays added extra effort, but they were absorbed into what felt like the normal pace of development. 

In a high-speed environment, every step becomes a direct input to the next: discovery feeds analysis, analysis feeds specifications, specifications feed planning, planning feeds testing. 

If each step is correct, the next one moves at lightning speed. If something is off, the impact compounds immediately. 

A small misalignment can now slow the next stage by an order of magnitude. 

If organizations stop at the first wave, the gains will be limited.

You might cut delivery time from two weeks to one. That’s valuable, but it’s not transformative.

The real multipliers emerge only when the second wave is addressed. When requirement quality improves, when handoffs are clean, when communication happens at the right time, and when decisions are made quickly, the entire delivery pipeline accelerates.

This is where five- to tenfold gains become possible.

The same technology that exposed these bottlenecks can also help resolve them. In the first wave, AI replaces or accelerates tasks. In the second wave, it closes the gaps between tasks. 

Consider business analysis again. 

An AI agent can be equipped to ask the same questions a developer would eventually ask, but before development begins. Instead of reactive back-and-forth during implementation, the hard questions are surfaced upfront, where they are cheaper and faster to resolve. 

The same principle applies across roles. 

During development, agents can perform continuous validation and verification. They can simulate user interactions, test flows, and identify issues before the work reaches QA. They can run overnight, effectively extending the team’s capacity without adding headcount. 

At Nortal we use our Specifying Agent - "Socrates the Questioner". This Claude Desktop skill transforms vague feature ideas into structured specifications through guided, Socratic dialogue. 

It identifies ambiguities early, defers technical decisions to the right people, and produces clear, actionable specs. In doing so, it addresses the very human problem of unclear requirements by front-loading the questions that would otherwise disrupt development later. 

The highest productivity gains don’t come from making fast tasks faster. 

They come from fixing the issues that were always slowing teams down, but never visible enough to demand attention. 

AI is changing the economics of software delivery. It compresses execution time so dramatically that everything else is forced into the spotlight. 

Organizations that recognize this shift and act on it will move beyond incremental improvements. 

They won’t just build faster. They will remove the constraints that were always limiting them.