A Hosted AI Service,
or a Private Model?
Hosted AI services give you the most capable models through a simple connection, billed by usage. A private model runs in your own cloud account or on your own hardware, so your data does not go to an outside AI provider. Each has real strengths, and the right answer depends on your data, your volume and the work itself.
The Short Answer
Start with a hosted service when the data can be shared under the provider’s terms and you need the strongest reasoning. Consider a private model when data must stay in your environment, when volume is high and steady enough that usage fees add up, or when you need control over exactly which model version runs. Smaller private models can be less capable, so test them on your real work first.
Side by Side
| Consideration | A private model | A hosted AI service |
|---|---|---|
| Where your data goes | Stays in your cloud account or on your hardware. | Sent to the provider under its terms. |
| Capability | Strong for focused tasks; smaller models can struggle with complex reasoning. | Access to the most capable models available. |
| How the cost behaves | Mostly the cost of the servers, whether busy or idle. | Pay per use; low at small volume, rises with usage. |
| Setup | A deployment project: hosting, security and testing. | Quick to connect. |
| Upkeep | Updates, monitoring and capacity are your responsibility or ours under an agreement. | The provider runs and updates the models. |
| Model changes | You decide when to change versions. | The provider may update or retire models on its own schedule. |
Which One Fits?
A private model is likely the better choice when
- Data cannot leave your environment under your policies
- Usage is high and steady, so per-use fees add up
- The task is focused, such as classifying or extracting
- You need a fixed model version that will not change
A hosted service is likely the better choice when
- The task needs the strongest reasoning available
- Usage is low or unpredictable
- The data can be shared under the provider’s terms
- You want to prove the idea before investing in hosting
Four Questions That Decide It
Answer these about the specific work you want AI to do.
What data will the model see?
List it. Customer records, contracts and health or financial details change the answer more than anything else. Your own policies and advisors decide what may be shared.
How hard is the task?
Sorting and extracting often work well on smaller models. Open-ended reasoning and writing usually benefit from the larger hosted ones.
How much will it run?
Estimate requests per day. Usage fees are cheap at low volume; a private server costs the same whether it handles ten requests or ten thousand.
Who will look after it?
A private model is a system to run: updates, monitoring and capacity. Decide who owns that before choosing it.
Many Systems Use Both
A common design routes each task to the right place: a private model handles the steps that touch sensitive data, and a hosted service handles the steps that need stronger reasoning on data that can be shared. Building the application so the model can be swapped keeps you free to change the mix as prices, models and your needs change.
Typical Starting Points
Work is priced by scope, not by the hour. You get a written scope with a fixed price before anything starts, and larger builds are split into milestones. The ranges below show where engagements typically start.
See Starting Prices
| Engagement | When it fits | Investment |
|---|---|---|
| Automation Opportunity Review | You know the workflow hurts, but not yet what, if anything, to build: one workflow reviewed, with the smallest worthwhile fix in a written action plan. | $495 fixed fee |
| Custom Engagements | New products, AI agents, private AI, custom applications, modernization and data platforms, scoped and priced in writing. | Quoted in writing |
| Operations Automation Partnership | After a build: monitoring, maintenance, vendor and API changes, and small improvements. | From $750 / month |
Hosting, domains, licenses, subscriptions, and processing fees stay in accounts you own and are billed to you directly by each vendor whenever practical. Where we must place an approved charge on your behalf, that amount is collected before we incur it.
After the Build
Every build ends with documentation and a handoff, so your team can run it. If you would rather not, a monthly partnership covers monitoring, vendor and API changes, and small improvements under its own written agreement.
See Partnership LevelsRelated Problems and Services
Before You Decide
- Is a private model as good as a hosted one?
- For focused tasks it can be good enough, and sometimes excellent. For complex reasoning the largest hosted models are usually stronger. We test candidate models on samples of your real work before recommending one.
- Is private AI cheaper?
- It depends on volume. At low or uneven volume a hosted service is usually cheaper. At high, steady volume a private deployment can cost less per request, but the servers cost money even when idle.
- Does using a hosted AI service mean the provider trains on our data?
- That depends on the provider, the product and the account terms you accept. Read those terms, and ask your own advisors whether they meet your obligations. We do not give legal or compliance advice.
- Where would a private model run?
- In your own cloud account or on hardware you control. The accounts stay yours, and third-party hosting costs are billed to you directly.
- Can we switch later?
- Yes, if the application is built for it. We design AI applications so the model behind them can be changed without rebuilding everything.
This page is general information to help you compare options, not legal, financial, tax, employment or other professional advice. Figures are illustrations based on the stated inputs, not client or market data, and are not a promise of savings. We do not guarantee any particular business, time-saving or technical outcome. Our Terms of Service govern.
Full terms: Terms of Service
Unsure Where Your AI Should Run?
Tell us what the AI would do and what data it would touch. We will tell you whether a hosted service, a private model or a mix fits.