Strategy · Philippines

A Short Defense of the Written Thesis

September 12, 20233 min read

We call it a thesis. Other studios call it a discovery output, a brief, a product definition, or a requirements document. The name is less important than the artifact: a written statement, produced collaboratively at the start of an engagement, that says in plain language what we are building, why, and for whom.

Every project we take on produces one. This has not changed, even as the tools available for producing and consuming text have gotten significantly more capable. Here is why.

What the Written Thesis Actually Is

The thesis is not a features list. It is not a technical specification. It is a document that answers three questions: what problem does this product solve, who is experiencing that problem, and why does this approach make sense for them.

It is short. Usually four to eight pages depending on the complexity of the engagement. It is written in language a non-technical reader can follow because both the client and the build team need to be able to hold it at the same time.

At the end of a discovery phase, we review the thesis with the client and they sign off on it before the build quote is issued. The thesis is what the quote is priced against. If the thesis changes materially, the quote changes.

The Pressure to Skip It

The pressure to skip or abbreviate the thesis comes from two directions.

Clients who have a clear vision of what they want to build sometimes feel that writing down what they already know is redundant. Why spend a week on a document when you could spend that week building? The answer is that the vision in a client's head is not the same as a shared understanding between the client and the studio. Discovery surfaces the gaps.

More recently, there is a different kind of pressure: clients who have had a conversation with ChatGPT or used another AI tool to generate a product requirements document are arriving with that document and asking us to skip discovery. The document they bring is often impressive looking. It is also often wrong in subtle ways - requirements that are stated as if they were the only option when the problem has multiple valid approaches, features that are included because they appeared in similar product descriptions rather than because this client's users need them.

An AI-generated document is a starting point, not a thesis. It is a useful input to a discovery conversation, not a substitute for one.

What Discovery Finds That the AI Does Not

Discovery finds the constraints that are not in the brief.

The client who wants an inventory management system but has not mentioned that their supplier sends inventory updates via a format that no standard API handles. The healthcare product where one branch operates on a different workflow than the other four. The e-commerce integration where the client's fulfillment partner has a technical requirement that affects the checkout architecture.

These are real things we have found in discovery. None of them would have appeared in a generated document. They appeared because we asked questions and listened to the answers.

Discovery also finds disagreements within the client team. When multiple stakeholders are involved in a product, their ideas about what the product should do are often not identical. A written thesis that all of them read and sign off on is the mechanism for surfacing and resolving those disagreements before they become mid-build conflicts.

Why Written

Writing forces specificity. A spoken conversation can contain agreement that dissolves when both parties write down what they agreed to. A written document that both parties read and sign creates a shared, reviewable record.

In a year when AI tools can generate enormous volumes of text instantly, the discipline of writing something carefully - slowly enough to catch the gaps and ambiguities - is more valuable, not less. Anyone can generate words. The work of a thesis is in the thinking that underlies them.

We are not going to stop producing written theses. We are going to keep doing discovery even when clients arrive with AI-generated documents. And we are happy to explain why to any client who finds the process slower than expected.

Start a project →

Need this built for your business?

Let's scope it together.

Start a project