One-time engagement
Demo-to-production plan
A two to three week plan for taking your AI system from a working demo into production: what to keep, what to replace and what to build first, with the model and vendor calls made and the reasoning behind each.
The demo works. The plan is what makes it hold with real customers.
- 2 to 3 weeks
- Priced on the call
- The plan, in writing, and a readout
Is this for you?
The demo works and customers are next
Someone has seen it, liked it and asked when they can have it, and the honest answer is not yet.
The prototype became the product
What was built to show the idea is now carrying real traffic, and nobody chose it for that.
The next quarter is spoken for
You know what you want to ship and need the order, the calls and the reasons written down before the team starts.
Not this? The production readiness audit tests a live agent on your real use cases; the architecture review covers a system already in production that is about to scale.
What you get
- 01
The plan
Each part of the system marked keep, replace or build, with the reasoning, so your team can defend every call.
- 02
The build order
The work in the order it should happen, each step with what it unblocks and what it takes.
- 03
The model and vendor calls
Which model, which vendors, build or buy, with what each turned on and what it costs to reverse.
Illustrative pages. Yours carry your system’s scores, numbers and fixes.
The calls and the reasoning
What to build first
- 1CostFix
- 2CostFix
- 3CostFix
- 4CostFix
- 5CostFix
Demo-to-production plan
Six decisions, made and written
- What to keep and what to replace
- The model and vendor calls
- The eval baseline
- Retrieval and data
- Observability from day one
- Cost at launch
What the plan decides
Six decisions that separate a demo from a system that holds with real customers. I make each one with you and write the reasoning down.
01What to keep and what to replace
I go through the demo part by part and mark what can carry real traffic, what needs rebuilding and what should go.
I measureKeep, replace or build, part by part
02The model and vendor calls
I pick the models and the vendors for each step, the cheapest that passes with a fallback behind it, and write the reasons down.
I measureThe calls, and what each one costs to reverse
03The eval baseline
I turn the use cases the demo was built to show into the first test set, so every change after the plan is measured against something.
I measureThe first eval set, built from the demo’s own use cases
04Retrieval and data
I plan how the documents, the index and the permissions will hold once there are real customers and real volumes behind them.
I measureThe retrieval and data design for production
05Observability from day one
I plan the tracing so that from the first production request, every step can be seen and replayed.
I measureWhat gets traced, and where, from the first request
06Cost at launch
I estimate what one resolved task will cost at launch and at ten times, so the price and the plan agree before anyone commits.
I measureCost per resolved task at launch and at ten times
2 to 3 weeks, written plan and readout
The three weeks
- Week 1
Read the demo
Access in place, the demo read end to end, the use cases it was built to show collected with the people who built it.
- Week 2
Decide
Every part marked keep, replace or build; the model and vendor calls made with you; the build order drafted.
- Week 3
Write and hand over
The plan written, a readout with your team, the questions answered. A smaller system finishes at the end of week 2.
The next step
Start with a
20-minute call.
Tell me what the demo does and who is waiting for it. You leave knowing whether the plan fits, how long it would take and what it would cost.
Bring
- What the demo does, in a sentence
- Who is waiting for it, and by when
- What worries you about shipping it
Why me
I have taken systems from first demo to customers more than once. At TWINii I laid the foundation for a real-time AI product from October 2025, and ten expert twins went live at its public launch a year later. At HumanFirewall I took the first product from a brief to live in 45 days. The plan is the document I would have wanted on day one of either.
The track record“Mandeep is the level-headed, methodical thinker you want on your team. He is capable of looking at the big picture, anticipating problems and planning before hand.”
Rajat HansCo-founder and CTO, Kenko AIAsked on most first calls
The questions.
How long does it take?
Two to three weeks from the day access is in place. A single agent or a small app is usually two; several services, a retrieval layer or more than one platform usually three.
What access do you need?
The demo as it runs, the code, the design if there is one and an hour with whoever built it. Nothing needs to be installed anywhere.
Do you build it?
No. The plan is written for any team to build against. If you want me on the build, I can join part-time as your fractional CTO or AI architect, and NeuroGen, the studio I founded, can build it when someone other than me accepts the work.
Do you sign NDAs?
Yes, yours or mine, before I see anything.
What if the demo is further along than I think?
Then the plan says so, part by part, and is shorter. You pay for the read, not for a long document.
What does it cost?
It depends on the size of the system, so it is fixed on the first call, in writing, before anything starts.
How is this different from the architecture review?
The review reads a system that is already in production and about to scale. The plan takes a demo that has not met customers yet and decides what it needs to before it does.