One confusion turns up every single time: a tool that costs nothing gets read as artificial intelligence that costs nothing
The studio and its templates are free. The work the agents do is metered, and three of the four cost drivers are set at design review by people who never see an invoice.
Prepared by Redress Compliance · August 19, 2026 · Oracle Fusion estates worked inside. 35 to 45 estates, 2024 to 2025.
Executive summary
Zero license cost was read as zero running cost. Teams met consumption in the first week of genuine use, across the 35 to 45 Fusion estates worked inside during 2024 and 2025.
Sandbox prototypes crossed an entitlement line the moment somebody moved them into a live pod, because production custom agents sit behind a separate subscription.
Shipped templates covered more of the early use case list than expected, so custom builds were rarer than the business case assumed.
Nobody had written down which model tier each agent runs on, which made every later cost conversation guesswork.
What is the studio and what does it cost?
It is the environment for configuring, extending and orchestrating AI agents inside Fusion Applications, and it arrives with the Fusion Cloud subscription you already hold at no separate license charge.
Oracle set out the launch detail on the Fusion Insider blog, and the capability material sits on the AI Agents page.
Three ideas have to stay apart to answer the cost question honestly. The tooling and templates are free, the work the agents do is metered, and running custom agents at production scale is separately entitled.
What comes included at no cost
- The configuration environment: everything needed to define an agent, wire its tools and chain agents into a sequence.
- The shipped template library: prebuilt patterns mapped onto processes that already exist in HCM, ERP, SCM and CX.
- Model selection: choosing which model an agent uses, which is where running cost is actually decided.
- Testing outside production, which is where the whole first phase of any sensible rollout belongs.
What is not free
Consumption and the production entitlement. Every action draws against the unit meter once the monthly grant is spent, and custom agents at production scale sit behind a separate subscription, marked in the custom agent documentation.
Neither is a trap. Both are simply invisible from inside a tool that opens without a purchase order, which is exactly why they surprise people.
How much will the shipped templates already cover?
More than the business case usually assumes, and the gap is worth measuring before anybody writes code.
In the rollouts reviewed, the first workshop produces a long list of agent ideas and a strong instinct that most of them are bespoke. They rarely are.
| Pillar | Typical shipped pattern | What it displaces | Where money appears |
|---|---|---|---|
| HCM | Employee question handling and onboarding | Tier one service desk tickets | High volume, so tier choice dominates |
| ERP | Invoice narrative and approval summarization | Manual review commentary | Period end spikes drive the peak |
| SCM | Supplier and order status summaries | Status chasing by email | Scheduled runs create a floor |
| CX | Service case drafting and triage | First draft response writing | Grounding breadth drives the draw |
Three questions that shrink the bespoke column
- Is the underlying process standard Fusion? If the steps come out of the box, a shipped pattern probably addresses them.
- Is the difference in the data or in the logic? Different data is configuration. Different logic is construction, and only the second is expensive.
- Would a business user notice? If the answer takes longer than a sentence, the difference is not worth maintaining.
Run that scoring in an afternoon with the process owners in the room. It is the cheapest hour in the programme and it usually removes most of the custom pipeline.
- Model tier and grounding breadth modelled per agent, against the monthly grant
- The promotion boundary mapped, so no prototype crosses it unnoticed
- Your Fusion agreement read clause by clause with replacement language
When does the studio start to cost money?
At two moments, and opening the tool is neither of them. The first is when agent activity passes the monthly unit grant your subscription carries.
The second is when a custom agent moves into a live environment and crosses the separate entitlement line.
The four places cost appears anyway
| Driver | What it is | How to see it early | The control |
|---|---|---|---|
| Model tier | Which model the agent reasons on | Run the same task on two tiers and compare draw | A written model policy, enforced at design review |
| Grounding breadth | How much data is read before answering | Count retrievals per run in the test pod | Scope grounding to the narrowest useful set |
| Agent chattiness | Retries, replanning and multi turn loops | Look at the spread, not the average, across runs | Cap steps per run and log the outliers |
| Promotion to production | Moving a custom agent into a live pod | Nobody sees it. It is a permission, not a purchase | Approval gate on promotion, named owner per agent |
Three of the four are design decisions taken by people who have never seen an Oracle invoice. That is why a purely commercial control fails here and a design review control works.
Three of the four cost drivers are set during design, by people whose job is quality rather than spend. That is where the control has to live.
The decision rule for whether an agent stays cheap
- Does it run on the lowest capable tier? If somebody moved it up during testing, that decision is permanent unless you reverse it.
- Does it read the narrowest useful set of data? Every extra source is paid for on every run, not once.
- Is it started by a person rather than a clock? Human triggered agents scale with value, scheduled agents scale with the calendar.
The Fusion AI agents playbook
The model policy, the promotion gate, and the consumption baseline to establish before you agree to any pack.
Get the playbook →What 35 to 45 Fusion estates showed
Across the Oracle Fusion estates worked inside during 2024 and 2025, and the studio rollouts reviewed since, one confusion turns up every single time.
The four patterns that recurred
- Teams treated zero license cost as zero running cost, then met consumption on the first week of genuine use.
- Prototypes that behaved perfectly in a sandbox crossed an entitlement line the moment somebody moved them into a live pod.
- The shipped templates covered more of the early use case list than anybody expected, so custom builds were rarer than the business case assumed.
- Nobody had written down which model tier each agent runs on, which made every later cost conversation guesswork.
The last one is the cheapest to fix and the most consequential. Model policy set in week one is worth more than any negotiation you will run in month twelve.
Research briefingReading the Oracle SaaS renewalWhere subscription scope, consumption meters and entitlement boundaries meet inside a Fusion agreement.
Where the common advice on the studio is wrong
The advice you will hear from every direction is that the studio is free, so you should start building custom agents to differentiate your processes. We disagree, and not for compliance reasons.
The template you configure and then leave alone is the cheapest asset in the stack. It survives quarterly updates with the least attention and stays comfortably on the free side of the entitlement line.
Custom construction is where consumption climbs, where regression testing lands on your team every quarter, and where the separately subscribed boundary sits. Build custom where the process genuinely is your advantage.
Everywhere else, the unmodified template is the sophisticated choice, not the lazy one.
The tooling, the template library and the orchestration are included.
Model tier, grounding breadth, chattiness and promotion.
By people whose job is quality rather than spend.
The wider entitlement picture sits in our Fusion AI agents licensing reference and in the agentic applications subscription.
Your first five moves
- Write a model policy in week one, because it is worth more than any negotiation you run in month twelve.
- Score the agent idea list against the shipped templates in one afternoon with the process owners present.
- Default down and promote by exception, so routine template actions run on the lowest capable tier with a named approver for any upgrade.
- Baseline real consumption against your monthly grant before agreeing to any pack or bundled block.
- Put an approval gate on promotion to production and a named owner on every agent, because that boundary is a permission rather than a purchase.
Frequently asked questions
Does the studio carry a license charge?
No. The studio, the template library and the orchestration tooling carry no separate license charge on a Fusion Cloud subscription.
So what actually costs money?
Consumption and the production entitlement. Agent actions draw against the unit meter once the monthly grant is spent, and production custom agents are separately subscribed.
What is the confusion that recurs?
A tool that costs nothing gets read as artificial intelligence that costs nothing. Teams met consumption in the first week of genuine use.
Where do the costs actually appear?
In four places, and none of them is the tool: model tier, grounding breadth, agent chattiness, and promotion of a custom agent to production.
Why does a commercial control fail here?
Because three of the four drivers are design decisions taken by people who have never seen an Oracle invoice. The control has to live at design review.
Are custom agents usually necessary?
Less often than the business case assumes. The shipped templates covered more of the early use case list than anybody expected.
How do you tell configuration from construction?
Ask whether the difference is in the data or the logic. Different data is configuration. Different logic is construction, and only the second is expensive.
What keeps an agent cheap?
The lowest capable tier, the narrowest useful grounding, and a human trigger rather than a clock. An agent answering yes to all three stays unremarkable.
What is the hidden boundary?
Promotion to production. Nobody sees it happen because it is a permission rather than a purchase, which is why it needs an approval gate.
What is the cheapest fix available?
Writing down which model tier each agent runs on. Without it every later cost conversation is guesswork.
