Skip to main content

The work behind the AI output

By

 I’ve been using ChatGPT and Codex a lot lately - for research, analysis, writing, and the many small tasks that make up a working day. One thing I’ve started doing is saving the prompt and context behind a document, so I can come back to it later. It began as a practical habit. Then I realised it pointed to something bigger.

 I noticed this while working on a proposal for new AI capabilities. When it was time for review, the proposal itself wasn’t the whole story. People also needed to understand how we had got there: what information we’d considered, what assumptions we’d made, and what might happen next.

I’ve run into a similar thing with client proposals. Work can start in an early conversation, before there’s a formal project space or an agreed way to share context. Later, I, or someone else, might need to pick it up and carry it forward. The conversation may be hard to find, or the reasoning behind the document may never have been written down. Then the final proposal is there, but the path to it isn’t.

 So I’ve been thinking about what it means to leave behind more than the finished document. Could we also save the useful parts of the process, the sources, criteria, prompts, decisions and checks, so someone else can understand the work and continue it?

I don’t expect a future AI run to recreate the same wording. Models change, sources change, and even the same prompt can lead to a different result. What I want is for the next person to have enough context to review the work, update it and understand what changed.

There’s a more formal version of this idea in research. The World Bank, for example, has guidance on documenting AI-assisted analysis, including the prompts, inputs, outputs and human review. The point isn’t to promise perfect reproduction; it’s to make the process visible enough to inspect and revisit.

I’ve been trying a small prompt that turns an AI session into a handover package: a README, notes on the method and sources, a short run log, and the final document if it’s available. I’ve played with it on real work, and it helped another person continue with better context. That’s a modest result, but it made the idea feel useful.

I think of it as a tiny context warehouse for one piece of work. It won’t replace project knowledge-sharing tools, and it doesn’t need to. It’s for the smaller, everyday tasks—or early conversations—that haven’t found a home in a larger system yet.

I’ve put the prompt and its documentation in ai-session-reproduction-prompt. The question I’m left with is simple: if AI helps us produce work faster, how do we make it easier for someone else to pick that work up? This could be a staring point…

Useful resources: World Bank: Documenting AI use for reproducible research

 

Comments

Post a Comment

Popular posts from this blog

How to audit an Azure Cosmos DB

By
In this post, we will talk about how we can audit an Azure Cosmos DB database. Before jumping into the problem let us define the business requirement: As an Administrator I want to be able to audit all changes that were done to specific collection inside my Azure Cosmos DB. The requirement is simple, but can be a little tricky to implement fully. First of all when you are using Azure Cosmos DB or any other storage solution there are 99% odds that you’ll have more than one system that writes data to it. This means that you have or not have control on the systems that are doing any create/update/delete operations. Solution 1: Diagnostic Logs Cosmos DB allows us activate diagnostics logs and stream the output a storage account for achieving to other systems like Event Hub or Log Analytics. This would allow us to have information related to who, when, what, response code and how the access operation to our Cosmos DB was done. Beside this there is a field that specifies what was th...
Post a Comment
Read more

Why Database Modernization Matters for AI

By
  When companies transition to the cloud, they typically begin with applications and virtual machines, which is often the easier part of the process. The actual complexity arises later when databases are moved. To save time and effort, cloud adoption is more of a cloud migration in an IaaS manner, fulfilling current, but not future needs. Even organisations that are already in the cloud find that their databases, although “migrated,” are not genuinely modernised. This disparity becomes particularly evident when they begin to explore AI technologies. Understanding Modernisation Beyond Migration Database modernisation is distinct from merely relocating an outdated database to Azure. It's about making your data layer ready for future needs, like automation, real-time analytics, and AI capabilities. AI needs high throughput, which can be achieved using native DB cloud capabilities. When your database runs in a traditional setup (even hosted in the cloud), in that case, you will enc...
Post a Comment
Read more

Azure Well-Architected AI workload Assessment

By
  AI is everywhere, part of the IT solutions we build and run today. Having an AI service, a good model, and data is not enough. As for cloud, the real difference is how we build, manage and run the whole solution. Microsoft created the Azure Well-Architected Framework for AI Workloads exactly for this reason — to help teams design AI systems that are reliable, secure, and cost-efficient. The assessment has six main categories that we cover in the next section. Based on the results, we can gain a good understanding of the current AI workload estate and a list of actions to improve how you run and manage your AI workloads . Designing the AI Application The first step in building your AI application is to consider how you will structure it. Using containers for tasks like data processing or model inference helps maintain consistency across the system. This approach makes it easier to update, move, and manage different components. When you have multiple steps in your workflow, such as...
Post a Comment
Read more