The Control Room for Every Stack9 AI Agent
Stack9 AI Studio is where your own AI Agent configurators tune prompts and tools, run evaluation datasets, replay any conversation step by step, and promote or roll back a version — every change still passing the evaluation gate and formal change control.
The Authoring Loop, Step by Step
Every agent starts as a draft. A configurator defines it, tests it against real questions, and only promotes it once it clears an evaluation gate agreed on in advance.
How a configurator builds and improves an agent
An AI Agent configurator opens Stack9 AI Studio
- Step 01
Define the agent
Its goal, its instructions, and the tone it should take
- Step 02
Grant access
Choose the tools and knowledge this agent may use
- Step 03
Test and refine — the authoring loop
Ask a real question
"Is the physiotherapy invoice on claim CLM-10482 covered under this policy, and how much is payable?"
Read the answer
Drafted only from the policy wording and documents this agent was given access to
Inspect the trace
Which sources it used, which tools it called, the identity each call ran under, and how it reached that answer
Adjust
Sharpen the instructions, widen or narrow the knowledge, tighten the guardrails
- Step 04 · The gate
Run the evaluation suite
Scored automatically against a benchmark set approved in advance — promotion depends on it
Below the threshold — back to the loop
Meets the threshold — continues
- Step 05
Promote the version
Versioned and promoted under change control — and rolled back in seconds if needed
- Step 06
Live
Available to back-office users, portals and integrations
- Step 07
Real-world feedback
Ratings, overrides and escalations captured against actual answers
- Configured in Stack9
- Managed AWS service
- Data & measurement
- A person / user
- Loop / feedback
The evaluation gate in Step 04 is also a governance control — see how Stack9 governs AI in production on AI Governance.
What AI Studio Gives a Configurator
Configure Without Code
Tune each agent's system prompt, tool access, skills, guardrails and model selection without code, and author the standard phrases it reuses.
Evaluate Against Real Cases
Build, extend and run evaluation datasets — benchmark questions with predefined correct outcomes — and score a version before it can be released.
Replay Step by Step
Open a full conversation and follow it turn by turn — what was asked, every tool call, the identity it ran under, and what it did with the answer.
Trace to Effect
See how an agent's output flowed into the rest of the solution — which record it wrote to, which workflow step it triggered — before adjusting it further.
Compare Versions
Compare two versions side by side on accuracy, cost and behaviour, so a change is attributed to exactly what moved rather than guessed at.
Promote and Roll Back
An authorised configurator promotes a version to live, or rolls back to a prior one, in seconds — every promotion still passing the evaluation gate and change control.
Score Per Agent
Each specialist agent is scored on how often it resolves without escalating, its accuracy and the feedback it receives — a low score queues it for review.
Every promotion still passes the evaluation gate and the same formal change control described in Security & Architecture.
Where AI Studio Ends and Development Begins
- The AI layer is around 80% configurable in AI Studio. The remaining effort sits in building the evaluation datasets themselves and in function-calling development for complex, client-specific tool integrations — treat the percentage as indicative, from one scoping.
- Evaluation datasets are authored, not pre-built. Benchmark questions and their correct outcomes are built per engagement with your own subject-matter experts, before the first agent is configured.
- A new data source or tool needs a developer the first time it’s connected. After that, configurators build and tune agents against it independently in AI Studio without further engineering.
- April9 remains responsible for deeper platform, infrastructure and model-provider changes. AI Studio governs agent configuration; changing the underlying runtime, models or infrastructure sits with April9 under formal change control.
Put Your Configurators in Control
Stack9 AI Studio is the native interface for configuring, evaluating and promoting every agent you run.