Most of a from-scratch AI budget goes to infrastructure that is identical in every organization. This works out how much, using your own numbers.
On ibl.ai you own all the code and the data, run it model-agnostic across any LLM, and pay with no per-seat pricing — so you can deploy anywhere, from your own cloud to a fully air-gapped network.
Last updated:
A time-and-materials AI engagement is priced as hours multiplied by a blended rate. What that price buys is rarely broken down, and the breakdown is the whole story. Every AI deployment needs the same foundational layers: permissions-aware retrieval, an evaluation harness, guardrails and prompt-injection defense, role-based access control, audit logging, and model routing. None of it is specific to your organization. A competitor in your sector would need it built identically. Only the remainder — your data model, your workflows, your integrations, your agents — is genuinely yours, and it is the smaller share. This calculator separates the two. Enter your team size, blended rate, and timeline, then set what proportion of the effort goes to that undifferentiated foundation. The output is the amount you would be paying to build something that already exists, alongside what the same programme costs when the platform is licensed and only the integration is scoped.
Engineers multiplied by the blended rate, 160 billable hours a month, across the full timeline — before any overrun.
The same engagement at the overrun you expect. This is the figure to compare against, because it is the one that gets invoiced.
The share of the invoice funding retrieval, evaluation, guardrails, access control, audit logging and model routing — none of it specific to your organization.
The licence, a year of compute, and integration scoped against a platform that already runs.
What the from-scratch engagement costs above the licensed path in year one. Negative means the build is cheaper at your inputs, which happens when the foundational share is genuinely low.
Time not spent constructing a platform. For most programmes this is the more consequential number, because the capability arrives sooner.
The first-year difference expressed as a proportion of the T&M invoice.
| Segment | Metric | Typical | With AI |
|---|---|---|---|
| Mid-size enterprise (5 engineers, 12 months, $185/hr) | First-year cost to production | $1.78M at the estimate, $2.22M at a 25% overrun | $594K licensed path — licence, compute and 8 weeks of integration |
| Large programme (12 engineers, 18 months, $250/hr) | Spent on foundational platform work | $5.6M of an $8.6M invoice at a 65% foundational share | $0 — those layers already exist and are amortised across every customer |
| Public sector T&M award | Time to first production workload | 12–18 months, with a ceiling conversation before it | Weeks, because the construction phase does not exist |
| Any organization, year three | Maintenance of the foundational layers | A funded backlog item for every model release and protocol change | Delivered upstream, with the source code in your possession |
The comparison isolates one variable: how much of the platform already exists when the engagement starts. Both paths are costed at the same blended rate, with the same engineers, so the result reflects scope rather than a pricing advantage.
The T&M path multiplies engineers by rate by 160 billable hours a month across the timeline, then applies the overrun you expect. The undifferentiated share of that total is calculated separately, because it is the portion funding work that is identical across organizations and available off the shelf.
The licensed path costs the platform licence, a year of compute, and integration effort billed at the same rate — scoped in weeks rather than months because the system being integrated is finished.
Two things the model deliberately excludes, both of which favour the licensed path: ongoing maintenance of a bespoke system (every model release, protocol change and security fix becomes a funded backlog item), and the value of holding source rights at the end. Excluding them keeps the arithmetic conservative and easy to audit.
The default 65% foundational share is a planning figure, not a measurement of your programme. Categorise submitted timesheets into platform infrastructure versus organization-specific work and substitute your own number — that single ratio explains most AI budget outcomes.
ibl.ai is the agentic AI platform where you own all the code and the data. You self-host the entire stack inside your own perimeter, run it model-agnostic across any LLM and switch anytime, and pay by usage with no per-seat pricing — so you can deploy anywhere: your cloud, on-premise, GovCloud, or fully air-gapped.
Full source code under a perpetual license, running on your infrastructure. Not API access to someone else's platform — the stack itself is yours.
Run any LLM — Claude, GPT, Gemini, Llama, Command, or your own fine-tune — and switch providers without rewriting the platform.
Usage-based billing against a budget cap you set. Cost tracks what your organization actually uses, not how many people you employ.
Your cloud, your VPC, on-premise, GovCloud, or a fully air-gapped network with no outbound connectivity.
1.6M+ users across 400+ organizations run the platform this way, including NVIDIA, MIT, and Syracuse University.
ibl.ai is family-owned and operated from New York, NY — a U.S.-headquartered, domestically-owned long-term partner, not a vendor that sells licenses and moves on.
See how ibl.ai deploys AI agents you own and control—on your infrastructure, integrated with your systems.