Services / Architecture & AI Assessment
Know what your system costs, where it can fail, and what to do next.
I inspect your current architecture and AI workflows, then give you a written map of the system, the risks worth acting on, and options with effort estimates. The work has a defined scope and a decision you need to make.
$3,500, one time. I scope the systems and access with you before work starts.
The decision this supports
You have software that works, but the cost, failure paths, or next investment aren't clear. Perhaps the AI bill is growing, a vendor recommends a rebuild, or a team wants to add an agent that can change customer data. I look at the system you actually run so you can choose what to fix, replace, postpone, or leave alone.
For a focused, by-hand code review of an app built with AI coding tools, see the AI App Review. This assessment looks across the architecture and operating cost of a wider system.
What you receive
A current-state map
The important components, outside services, data paths, and ownership boundaries, based on code, configuration, and conversations with the people running it.
A cost and risk read
Where AI spend is measured, which workflows consume it, and where permissions, failure handling, or dependencies create material risk. Unknowns stay marked as unknowns.
Practical options
A prioritized choice among changes such as right-sizing a model, adding a missing control, refactoring a subsystem, replacing a component, or keeping the current design. Each option includes an effort estimate and tradeoffs.
A decision walkthrough
A written assessment you can share with your team and a conversation to work through what should happen next. Implementation is a separate scope if you want me to do it.
How I look at a running system
I build and operate AI systems myself. My write-up on AI cost shows one of the checks behind this assessment: attribute each model call to the feature or workflow that caused it, then judge cost per useful result rather than the invoice total. My degraded-state write-up shows why I ask what users and operators see when one dependency fails.
Those are examples from my own systems, not results from a client assessment. Your report would be based on your code, usage, team, and constraints.
How the assessment works
Set the decision and access
We agree which system and question are in scope. I ask for the code, usage and cost records, configuration, and the people who can explain how it runs.
Trace the real paths
I compare the architecture described to me with the code, integrations, operating data, and failure paths I can inspect. Missing access and uncertain claims go into the record.
Review the options
You receive the written map, ranked risks, and options with effort estimates. We walk through the tradeoffs before you choose the next step.
A good fit when
- You need an independent technical read before committing to a rebuild, vendor, or AI expansion.
- You need to know what an AI workflow costs and why the cost changes.
- Your team needs a prioritized plan that names the risks and the work, without committing to implementation yet.
Tell me the decision you're facing.
I'll ask what system is in scope and what evidence is available, then tell you whether this assessment is the right size.
Request a 30-minute call