Decision review / AI proposals
The demo worked. That was never the question.
Every AI proposal demonstrates well, because the demonstration is assembled from material chosen to make it demonstrate well. What you would be buying is the behaviour on your data, under your obligations, with your staff, on the morning it is switched on for everybody. That is a different question, and it is the one worth paying to have answered.
What we assess
Five questions a demonstration cannot answer.
The technology is rarely the risk. What it touches, who relies on it, and whether anybody can tell if it is working are the risks.
What data it would actually see
Not what the deployment diagram says. What a user will paste into it on a Tuesday, and whether anything in your environment stops them.
What you have already promised
Client agreements, regulatory obligations and insurance terms that made commitments about where data goes and who processes it. Those were signed before this tool existed and they still bind you.
Retention, training and residency
What the provider keeps, what they may train on, where processing happens, and what survives the end of the contract. These are contract questions with technical consequences.
How you would know it worked
A pilot with no success criteria cannot fail, which is why so many of them are quietly extended. We define what would count as working before anything is switched on.
What it costs to stop
Once a workflow depends on it, reversing is a project rather than a decision. Knowing that cost in advance is what makes the first commitment a safe one.
What you get
A deployable answer, or a reason not to deploy.
Written so that it can be handed to the vendor, to your insurer, or to whoever asks why the decision went the way it did.
An exposure map
What data the proposed deployment touches, which obligations that engages, and where the gaps between the two sit.
A scoped first step
If it is worth doing, the narrowest version worth doing first, with the criteria that decide whether it goes further.
A written position
Proceed, proceed narrowed, or not yet. With the reasoning, and the conditions that would change it.
Who you are working with
Assessed by someone who ships software as well as advice.
VanDien.io is led by Christopher Moskowitz, who most recently ran engineering for Jefferies, a full service capital markets, investment bank, and who has designed, built and released four commercial applications alone, across C++, C#, .NET, React and TypeScript, and Swift. An assessment of what a system does with your data is written here by somebody who has had to implement that behaviour rather than only specify it.
Related reviews
The commercial side of the same proposal.
Common questions
Are you against AI deployments?
No, and the recommendation is frequently to proceed. It is more often to proceed with a narrower first scope and a defined way of telling whether it worked, because the common failure here is not a bad tool, it is a deployment nobody can evaluate afterwards.
Our staff are already using these tools unofficially. Does that change things?
It usually changes the question from whether to adopt to what you are already exposed to. That is worth establishing before you sign anything, and it often reframes the proposal in front of you.
Can you assess a proposal for a tool we have not chosen yet?
Yes. At that stage the more useful work is defining what the tool would have to do to be worth buying, which makes the eventual comparison an evaluation rather than a demo.
Do you review the contract as well as the technology?
The parts that bear on data: what the provider may retain, what they may train on, where it is processed, what survives termination, and what they commit to if it goes wrong. For a broader commercial read, that is a quote review.
Send the proposal and tell us whose data it would touch.
info@vandien.io · (551) 236-3191 · Ridgewood, NJ, serving the New York metro