Monday morning: 80 pages of client requirements

Last Monday at 9 a.m., I was revising a quote in a café when Lin Shan—my client—sent 80-plus pages of requirements and two rounds of clarifications. The file was on my laptop, but scrolling back and forth made it easy to miss a constraint. I used to force myself through documents this way. I miscopied a key requirement into my proposal and lost three hours to rework. In that moment, I was certain: the client wasn’t the problem. I just didn’t have a less painful way to read the document.

What Neon does, and who is already using it

Gemini 4 Neon is Google’s new model. It sits below its premium models but focuses on improvements in long-document handling and office automation. It’s currently available to Google AI Pro and Ultra subscribers, or through pay-as-you-go usage. The article mentions Nate publicly sharing a workflow where AI rewrote his LinkedIn bio, so Neon is already comfortable with this kind of content cleanup. For us, the more practical use is putting contracts, requirement packs, and meeting notes together, asking Neon to list the delivery scope, payment milestones, and unresolved questions, and then checking those points against the source.

Replicating it today: start with the cheapest step

The cost of testing this can be low. If I already have a Google AI Pro or Ultra subscription, the extra cost may be $0, with about 30 minutes of setup time. Without a subscription, I can start with pay-as-you-go usage: the article lists $3 per million input tokens and $15 per million output tokens. I wouldn’t start with a large plan. The technical barrier is simply being able to upload a file and use the chat box; no coding knowledge is needed. My first step is to open Gemini, start a new conversation, upload a client document I’m authorized to use, and ask it to “list the key requirements and label the corresponding page numbers.” I wouldn’t ask it to make the final decision for me.

Choosing based on the stage I’m at

If I’m just starting out and don’t receive many client documents each week, I’d use the free allowance before worrying about a model upgrade. With one or two steady clients whose files regularly exceed 50 pages, I’d make “extract first, verify next” a fixed workflow. If I’m scaling and several people handle requirements, I’d test the accuracy and cost per job before deciding whether to upgrade. There’s no urgency to try it now; I can keep my current process until long documents become a bottleneck. If I do test it, I’d begin with a small task on a non-sensitive internal document, rather than letting it take over client communication from day one.