Welcome back! Last session you learned how to run an Opportunity Scan, poke around a Diagnostic and Sessions inside it, and review a Presentation. This time we’re going deeper: next-level analysis from a Scan, customizing Diagnostics and Prompts, focusing a Scan on a specific population segment, and saving context for the agent to carry forward. We’ll also take a first look at Explorer.
1. Next Level Analyses
Let’s pick up where we left off. Click Opportunity Scan in the left sidebar, then List at the top of the window (if not already on the List view). Look for the User Training scan from your last session — it should be near the top. Click that Scan, then click the Presentation, then click Build to get into editing mode.
Let’s say there’s a slide you were expecting that didn’t make it into the Presentation. First step: ask the agent whether that data already exists as an unused Finding. Try something like
Check unused findings for information on SNF spend
The agent will rifle through all the Findings from the Scan and report back what it found (or didn’t find).
If the Finding exists, ask the agent to add it to the Presentation. You can let it decide where it fits best, or you can be specific:
Add this finding as a new slide 11
Either way, the agent handles the insertion.
If the Finding doesn’t exist yet, you’ll need to go back into the relevant Session and ask the agent to create it. The easiest way to get there is via the small arrow next to the title of the relevant slide — available both on the full Presentation screen and in the thumbnail panel on the Build screen. Click that arrow and you’ll land directly in the Session that produced that slide.
Give it a try — click the arrow on any slide that has a chart or table. (Text-only summary slides won’t have an arrow, since they pull from multiple Sessions and Findings rather than a single source.)
In the chat window at the bottom of the Session, ask the agent a follow-up question. For example:
Rerun the first key result filtered to the last 3 months and surface as a new key finding
Or drill into something specific — if the agent ran an analysis for the full population, ask it to rerun it for a subset, like a particular provider group or partner.
Once the agent promotes the new Finding, click on it in the Findings pane on the right to view it. Then click the Add button in the top right corner of the result and select your Presentation name — this makes the new Finding available to the agent in the Build window.
Head back to the Presentation tab and refresh the window so the agent picks up the new content. In the chat, tell the agent to find the result you just added and incorporate it at the right spot. A tip: give it the name of the Finding or some identifying detail — the agent has a lot of unused results to sift through, so the more specific you are, the better.
You can repeat this process as many times as needed, building out the presentation with additional analyses as you go.
2. Exporting CSV Files
But what if the Finding is a large table that doesn't fit on a slide? Or maybe you need to generate a chase list to share with a clinical team for follow up?
Any Finding can be exported directly from its Session. To test it out, go back into Session 3 of the User Training Diagnostic and click the Expand Findings button. Look for the Finding titled something like Hospital Admissions per K by Admit Type and click the Export button in the top right of that Finding. The corresponding CSV file will appear in your Downloads folder. (Note: If you are unable to export, your permissions may need to be updated, so please reach out to your Ursa contact for assistance.)
3. Diagnostic Suites
Let’s head back to the Opportunity Scan screen to cover one more important concept: the Diagnostic Suite. Click the New Scan button and expand the Diagnostic Suite dropdown. You’ll see a Suite section at the top (including Cost of Care Opportunity Analysis) and a Diagnostics section below it listing individual Diagnostics.
The dropdown lets you pick a single Diagnostic or an existing Suite. But what if you want to bundle several individual Diagnostics into a new group? Easy — click the Diagnostic button in the left menu (stethoscope icon), select Suite at the top, and click New Suite. The screen will list all available Diagnostics. Give your Suite a name and description, select the Diagnostics you want to include, and click Create Suite. Done — it’ll be available to any user running an Opportunity Scan.
4. Entities
Remember the second dropdown in the New Scan menu, labeled Entity? (If it’s already fuzzy, go ahead and click Opportunity Scan — New Scan to jog your memory.) The Entity component lets you define filters for how you want to slice your data — one Entity per payer group, per ACO, and so on. Your Ursa team will load this screen with the common entities for your organization during Compass implementation, but you can add a new one whenever you like. Here’s how to set one up.
Click Configuration in the left menu (gear icon), then select Entities at the top. Hit New Entity to see what goes into one. Give it whatever name your organization uses, then use the Level dropdown to pick an existing level (Payor, Partner, etc.) or type a new one to create it on the fly.
The Notes field is where you put entity-specific context that you want the agent to keep in mind when running diagnostics. For example, a particular partner may not have any Ambulatory Surgical Centers (ASCs) in its network — so you’d want to tell the agent not to recommend shifting surgical procedures to ASCs for that partner. That kind of guardrail lives right here.
Field Name is the name of the field in the Ursa Data Model — you can use the plain text name (Primary Payor ID) rather than the raw SQL column name (primary_payor_id). Field Value is what you want the agent to filter to. So if Field Name is Primary Payor ID, Field Value might be MSSP. Click Save Entity when you’re done, and your new Entity becomes available to any user running a scan or individual diagnostic.
You can also edit any existing Entity by clicking on it, adding additional context in the Notes window, and clicking Save.
5. Vintages
The third field on the New Scan menu is labeled Vintage. A Vintage is simply the timeframe label for a Scan. If you run the same Scan on a regular basis, tagging each run with a Vintage (May 2026, June 2026, Q1 2026, etc.) keeps things organized and easy to find later. Free-text it however makes sense for your organization.
6. Scan Presentation Guidance
The last dropdown on the New Scan menu is Presentation Guidance. This section allows you to pick a prompt that gives instructions to the agent on how to design the Presentation generated within the Opportunity Scan. You may recall from Session One that big prompt block at the bottom of the initial teal window in the Build screen of the Presentation— the one that started with “The starting request from the user is.” That’s what lives here.
You can customize the existing prompts or create new ones by going to Configuration (gear icon) in the left menu and selecting Scan Presentation Guidance. For example, you may have one format that you want the agent to use when generating your quarterly board presentation, but the outputs of the monthly Avoidable ED Analysis should have a different look and feel.
7. Global Prompts
Another thing worth knowing about: Global Prompts. These also live within Configuration.
The first one is Global Session Addendum. This is where you put instructions that apply to every Diagnostic and Opportunity Scan in your organization. It’s called “Addendum” because it gets appended to the very end of each Session Prompt — and the agent tends to pay extra attention to whatever appears last. Use it for things like behavioral guardrails (“only do one tool call at a time”) or org-wide constants (“always filter to where At Risk Patient = 1”).
The key thing to remember: the Global Session Addendum is for things that are always true across the board. If you need instructions specific to one Diagnostic or Scan, those live elsewhere. We’ve already seen how Entities handle population-level filtering. In a future session, we’ll dig into how to write or edit a prompt for a specific Diagnostic.
The second set, labeled Scan Followup Transition Guidance, controls what happens when a user does a next-level analysis from within an Opportunity Scan. You may remember from earlier in this Session how you can click the small arrow on a slide to go to its source Session and create additional Findings. This text tells the agent how to handle and respond to those follow up questions. The default prompt is designed to allow a nontechnical user to take advantage of this functionality while protecting against the agent (or user) going beyond the scope of the Session and potentially creating inaccurate results. You can edit this prompt if needed, but the recommended approach is to leave it as is.
8. Intro to Explorer
Last up for today: a first look at Explorer. We’ll keep it brief here and go deep in the next session.
Explorer is the part of Compass that feels most like a general-purpose chatbot — but it’s aimed at advanced users who know the data well enough to sanity-check what the agent is doing. By default, Explorer gives the agent solid instructions for reviewing and answering questions. The agent searches the data model, proposes an approach, asks any clarifying questions it needs, and then runs the analysis. Think of it as a more collaborative, back-and-forth experience than Opportunity Scan.
Let’s give it a spin. Click Explorer in the left menu (Compass icon) and paste in this prompt:
What percent of my patient population have a behavioral health condition? And how does their ED and IP hospital utilization compare with the rest of the population?
Click Start Session.
You’re now in a Session window — same setup as what you saw in Session One, but this time you’re in the driver’s seat. The agent will come back with a proposed approach and any clarifying questions. Answer it in the chat and hit Enter. From there, it will run its analysis and promote any Findings it thinks are worth surfacing. Keep the conversation going — ask for additional queries, different cuts, whatever you need — until you have what you’re looking for.
We’ll let you keep playing with Explorer on your own for now — that’s a wrap on Session Two. Next session: a deeper dive into Explorer and its Personas, plus a full walkthrough of how to create a new Diagnostic from scratch, whether for standalone use or as part of a suite.