The Short Answer
ibl.ai is a platform-plus-engineers option that deploys Azure OpenAI, Copilot Studio integrations, Gemini, Bedrock and OpenAI models on your own data, model-agnostic, inside your cloud, and hands over the source: you own all the code and the data. Whichever partner you choose, get a signed transfer of everything it builds, because the default does not favor you.
Plenty of firms can do this work. Microsoft, Google and AWS each run partner programs, and AWS alone announced 60 validated partners across three new agentic categories in November 2025, including consulting services. Finding a competent firm is not the hard part.
The hard part is what the buyer holds when the engagement ends. That is decided in the contract, and most of it is decided before the first line of code.
Is your data safe with Copilot, Gemini or Bedrock?
On training, all three providers give a clear written answer, so a consultancy cannot use it to set itself apart. Microsoft says prompts, responses and Graph data "aren't used to train foundation LLMs."
Google's terms say it will not use Customer Data to train or fine-tune AI/ML models without the customer's permission or instruction. AWS says Bedrock content is not shared with any model providers and is not used to improve the base models.
These are real commitments, and they are the reason a data-training objection rarely decides this purchase. What they do not address is the layer a consultancy builds on your behalf, which is where your institutional knowledge ends up.
What does a deployment partner actually build on your data?
A deployment partner builds the layer between the model and your systems: connectors, the ontology that maps your concepts and relationships, prompts, skills, agents, evaluation sets and guardrail policies.
Those assets outlast any one model, and they carry what your organization knows.
They are also what you are paying for. Licences and tokens come from the provider, while the engagement fee buys this layer. If it cannot move, your model choice cannot really move either.
The platforms are not locked boxes, and portability is worth testing specifically. Microsoft documents exporting Copilot Studio agents as solutions between environments, and also notes that not all agent components and properties are included.
Who owns the code a consultancy writes for you?
Under US law the consultancy's side of the table often starts with the copyright, unless a contract moves it.
The Copyright Office's Circular 30 says a work is made for hire when an employee creates it, or when it is commissioned in one of nine listed categories under a signed agreement.
Software is not named among the nine, which run from contributions to collective works through to atlases.
A work-for-hire label on a software contract may therefore not do what the parties assume. The dependable mechanism is an assignment, and 17 U.S.C. § 204 requires a transfer of copyright ownership to be in a signed writing.
Many good consultancies do assign deliverables, and many reasonably keep their pre-existing tools and license them back. Both are normal.
The risk sits in the gap: a deliverable that depends on a framework you only hold a limited licence to. This is general information, so have counsel read the clause.
How do the three kinds of deployment partner compare?
The three kinds of deployment partner differ most in what remains after they leave.
We compared them at length in Who Should Build Your AI Platform in 2026?. The table applies that comparison to Copilot, Gemini and Bedrock, and every cell still depends on the contract you sign.
| Partner model | Best at | Model scope | What to check you keep |
|---|---|---|---|
| Systems integrator | Scale, process and change management | Usually broad; often follows its alliances | Assignment of deliverables; license to any retained accelerators |
| Vendor professional services | Depth on one vendor's stack | That vendor's catalog | Exports of agents, prompts and evaluation data |
| ibl.ai | Agentic AI Operating System plus engineers who build with your team | Model-agnostic across any LLM | Platform source under a perpetual license, in your repositories; confirm how custom work is assigned, as with any partner |
An integrator can be the right call for a transformation spanning ERP and operating-model change. Vendor services can be the right call if you have committed to one stack for the long term. Those are legitimate choices, provided the ownership terms are written down.
What does per-seat pricing mean for a Copilot or Gemini rollout?
Per-seat pricing means the bill grows with headcount, whichever partner runs the rollout.
Microsoft lists Microsoft 365 Copilot at $30 per user per month, paid yearly, and Google lists Gemini Enterprise Business from $21 per seat per month, with larger editions sold through sales.
At Microsoft's list price, 5,000 seats comes to $1.8 million a year before services and before the qualifying Microsoft 365 license. That is the wrong shape for tools used heavily by some people and rarely by others, and the same arithmetic scales with every new hire.
The usage-priced options exist too.
Bedrock bills on consumption, Copilot Studio offers credit packs at $200 a month for 25,000 credits plus a pay-as-you-go meter, and Google offers a pay-as-you-go Gemini Enterprise edition.
A good partner should model these against your real usage before recommending seats.
How does ibl.ai deploy Copilot, Gemini or Bedrock models on your own data?
ibl.ai treats the models as interchangeable components of a platform you hold. Agents on the Agentic OS use Azure OpenAI, with Copilot Studio integration, Google Gemini, Amazon Bedrock or OpenAI GPT, in your own tenant or account.
Your data stays where it lives. Connectors reach your systems in place, and the ontology over them belongs to you, so changing the model does not mean rebuilding what your agents know.
The platform source arrives under a perpetual license, and that license is the clause to read in our contract, as with anyone's.
The platform runs for 1.6M+ users across 400+ organizations, and Syracuse University runs ibl.ai on its own Google Cloud. ibl.ai is family-owned and operated from New York, with no per-seat pricing, and our engineers work alongside your team.
What should you ask a deployment partner before signing?
Ask questions that have contractual answers, and ask every bidder the same ones. Five cover most of the exposure:
- Assignment: Does the statement of work assign copyright in every deliverable to us, in a signed writing?
- Retained tools: Which of your frameworks do our deliverables depend on, and what license do we get if we part ways?
- Independence: Can our own team operate and change this without you, and can we see it demonstrated before final payment?
- Portability: If we change model provider next year, which assets move unchanged, and which get rebuilt?
- Data paths: Which systems does every prompt touch, who can read the logs, and do any of them sit outside our tenant?
A partner with a good answer will welcome all five. If you want to see ours, the service pages and the full code license set out what we deliver.