Using the AI Advisor
The AI Advisor is a helper that knows your model. It can help you add data to your shapes, set up what-ifs and explain your results. You stay in charge: it proposes, and you decide.
What it can and can't do
The Advisor reads your model and refers to your elements by name. What it does depends on where you ask.
| Where you are | What it can do |
|---|---|
| Editing a model | Answer questions, spot problems and propose changes. You review a proposal and apply it. It can't change the model by itself. |
| Building a what-if | Propose a draft of changes and check it against the model. You apply the rows you want. Nothing is saved until you save. |
| Reading results | Explain what the numbers mean and why they came out that way. It's read-only here, and it only quotes numbers from your run. |
| Reading a study | Read the study's scenarios and compare them. It can't change the model or the study. |
Where to find it
- In an element's editor: use the sparkle button, then pick a suggested question or type your own.
- In the what-if builder and on a study: choose Ask the Advisor.
- In Results: choose Ask the Advisor about this run, or Ask Advisor to explain on the Overview tab.
If you don't see these, check Show AI Advisor features in Settings.
Creating and defining model behavior
Use this when you're starting a model, or when a step doesn't behave the way you expect. Good prompts describe the real process in your own words.
- Help me build a model of my process.
- What do you need to know to build my model?
- What kinds of processes can Quodsi simulate?
- Is the capacity here right for the arrival rate feeding it?
- What looks wrong with how this step is set up?
- Does this arrival pattern match the demand I described?
- Do the routing conditions on this connector cover every case?
- Is this model ready to simulate, and if not, why not?
- Is the run length long enough for the behaviour I care about?
When you ask for a change, the Advisor shows a proposal that you review first. Quodsi checks a proposal against the model's rules before you see it.
Defining and setting up what-if scenarios
Use this when you know the client's question but not yet how to test it.
- What should I vary to cut waiting?
- Which levers matter most for this?
- Is this draft a fair test?
- How many nurses do we need?
- What if demand grows?
- Should this be a what-if lever for this decision?
If the Baseline names a constraint, the Advisor starts there. For an open question like "how many nurses?", it tends to propose a range, which turns the what-if into a study. It keeps broad studies small, and it explains each change in your terms.
A what-if can't add a new step, a new path or a new kind of resource. If you ask for one, the Advisor explains that it's a change to the model itself.
Understanding and analyzing outputs
Use this when you have results and need to explain them. Ask from the tab you're looking at, so the Advisor knows what you mean.
- What happened in this run?
- What should I look at first?
- Is this process keeping up with demand?
- Which step is limiting throughput?
- Is the limit the step itself or a resource it seizes?
- Why is cycle time so high?
- Am I paying for capacity I do not use?
- Is the bottleneck the same in every replication?
- What does this study say?
- Which scenario is best?
- What does this column mean?
The Advisor pairs each Quodsi term with its plain meaning, and it names the tab where you can see a number it cites. That makes it a good way to learn the results as you go. The Metric glossary has the same definitions.
Getting better answers
- Be specific about the decision. "Should we add a nurse or a room?" beats "How can this be better?"
- Ask from the right place. Open the activity or result you mean before you ask.
- Ask it to explain itself. If you can't follow an answer, you can't defend it. Ask why, or ask which number it's based on.