The person who built it has left. The agency relationship ended. It still runs, mostly, and nobody can tell you whether it still works - only that turning it off feels risky. This is the least glamorous work we do and the most common reason people call us.
From$6,000
Read-only assessment: what it does, how well, what it costs, what would break it - written up with a recommendation.
Indicative starting price. The fixed fee for your scope is quoted after the two-day diagnosis, before any build begins.
What makes it work
The reflex on inheriting someone else’s system is to rebuild. It is usually wrong, and it is always the most expensive option. Most of what we take over has a sound core wrapped in undocumented decisions - the work is separating the two, not starting again.
Some systems are worth fixing, some worth replacing, and some are quietly costing more than they return and should simply be retired. We will tell you which, in writing, before proposing any build - even when the honest answer earns us nothing.
Architecture, data flows, prompts, permissions, dependencies, costs and failure modes, written down. Whatever you decide next, you stop being one departure away from not understanding your own system.
Capabilities
Tracing what actually runs, what calls what, which credentials and data it touches, and which parts are dead code nobody dares delete.
An evaluation set built from your real historical cases, so for the first time there is a number for how well it performs - not an impression.
What it spends and on what, which providers and models it depends on, and what breaks when one of them is deprecated.
What the system is permitted to do, what data leaves your perimeter, what is logged, and what a bad input could cause.
Monitoring and a rollback path first so it stops being fragile, then remediation against the priorities you agree.
In detail
A category name is not a scope. These are the individual pieces of work inside this practice - take the two that apply to you and ignore the rest.
Tracing what actually runs, what calls what, which credentials and data it touches, and which parts are dead code nobody dares delete.
Building an evaluation set from your real historical cases so that, for the first time, there is a number for how well it performs.
What it spends and on what, which providers and models it depends on, and precisely what breaks when one is deprecated.
What the system is permitted to do, what data leaves your perimeter, what is logged, and what a malicious input could cause.
Architecture, data flows, prompts, permissions and failure modes written down, so you stop being one departure from not understanding your own system.
Monitoring, alerting and a rollback path put in first, so it stops being fragile before anyone starts improving it.
Fixing the prioritised issues from the assessment, against costs you agreed in advance.
Moving off a deprecated model, an expensive provider or an abandoned framework, with quality measured before and after.
A safe retirement plan when the honest answer is that the system costs more than it returns.
What you receive
Around three weeks end to end. That comes from scoping tightly to one workflow - not from skipping a phase. Each still ends in evidence you can check.
Let's talk
The assessment is fixed-fee and read-only - we change nothing without your say-so. You get a written verdict on what it does, how well, what it costs and what would break it. If the recommendation is to retire it, that is what the document will say.
Related practices
