Field note
The interesting number is not always the rate on the invoice. It is what the whole system asks people, processes, and other tools to do around it.
Read the field noteAI costs · Finance operations · Practical systems
AI is changing finance.What happens next?
I’m Jackie. Close to Code is where I explore what AI costs, where it genuinely helps, and what changes when it becomes part of real finance work.
FollowingAI costs and operational consequences
MethodsAnalysis, experiments, and Python
Core questionIs it useful enough to justify itself?
01 / trace the costcount the surrounding system
02 / follow the worksee where the effort moves
03 / decide the valueask what is actually worth it
The cost behind the call
The cheapest model may be the most expensive workflow.
A low token price can look like savings until retries, rework, review time, integrations, and failure handling enter the picture. The more useful question is: What does the entire workflow cost?
What I’m exploring
The costs, workflows, and consequences of practical AI.
Close to Code follows the questions that appear after AI becomes part of real work. What are we paying for? Where does it help? What new work does it create? Who owns the decision?
What are we actually paying for?
Model costs are only one part of the bill. I’m looking at APIs, subscriptions, usage, review time, rework, and where the cost moves.
Read: The invoice is not the cost modelWhere does AI genuinely improve the work?
Accounting workflows, reconciliation, payments, expenses, exceptions, and the work technology moves rather than removes.
Read: Exceptions are the workflowWhat happens after the demo?
Retries, inconsistent outputs, ownership, failure handling, human review, and the details that determine whether an idea holds up.
Read: Human in the loop is not a strategyWhat changes when software starts recommending?
Decision rights, incentives, accountability, job design, institutional knowledge, vendor dependence, and hidden assumptions.
Read: Anomaly detection is not an explanationHow I approach the questions
Look past the price and the promise.
I want to understand what an AI system costs in context, what work it changes, and whether the result remains understandable to the people responsible for it.
Count the whole system
Look beyond the model or subscription price. Include review time, integrations, retries, rework, and downstream effort.
Follow the work
When technology saves time in one place, look for where the work moved and who inherited it.
Keep ownership visible
Make it clear who can recommend, approve, change, and stop the process. Keep the result explainable.
J
Hi, I’m Jackie
I like understanding why something costs what it does.
I work in accounting and finance operations, and I can get a little absorbed in understanding why a process keeps breaking or why a tool that was supposed to make life easier somehow created three new problems.
Lately, a lot of that curiosity has been focused on AI. I’m interested in what it costs, where it improves finance work, and what new issues appear once it becomes part of a real system.
Close to Code is where I share what I’m building, testing, and learning along the way.
The email version
Let’s look at what happens after the demo.
I’m preparing an occasional Close to Code newsletter about AI costs, finance workflows, system behavior, and the questions that appear once technology meets real work.
Newsletter signup is being connected. Until then, the journal has the current field notes.
Read the field notes