What Is Curve AI Analyst: Talk to Healthcare Analytics and Campaign Reporting

Curve AI Analyst is a HIPAA compliant AI assistant built into the Curve platform that lets healthcare marketers ask questions about their website analytics, attribution, and campaign reporting in plain English and get answers grounded in their own data. It runs inside Curve, where the tracking data already lives, so nothing has to be exported to a spreadsheet, pasted into a general purpose chatbot, or handed to a vendor without a BAA. You ask which campaign filled the calendar last month. It answers. Protected health information stays inside the compliant boundary the whole time.

That is the short version. The longer version matters, because of how long healthcare marketers went without it.

Healthcare marketing has always been the leftover

Every other industry got real analytics. An ecommerce team has known its cost per acquisition by campaign, by creative, by hour of the day, for the better part of fifteen years. A software growth lead can trace a deal back to the blog post that started it. None of that is exotic. It has been table stakes for a long time.

Healthcare got a pixel it could not keep.

The standard client side stack was designed to send raw browser data directly to an ad platform. That is fine when the page is a running shoe. It is a serious problem when the page is a bariatric surgery consult, a suboxone intake form, or an oncology second opinion request. The URL by itself can be protected health information. So can the referrer, the form field, the query string, the appointment type sitting in the booking widget. Regulators noticed. Hundreds of class action complaints have been filed against healthcare organizations over tracking since 2023, and enforcement has not slowed down.

Which left healthcare marketing teams choosing between two bad options. Leak, and hope nobody looks. Or go dark, strip the tracking off the site, and try to buy patients with no idea which half of the budget is working. We wrote about that choice in detail in leak or go dark, and the third option for telehealth tracking, and it is worth reading if you have ever had legal ask you to remove a pixel on a Friday afternoon.

Going dark is what most compliance officers push for, and it is why so many healthcare accounts run on gut feel. Leaking is what a lot of agencies quietly keep doing. Neither one answers a question as simple as which campaign produced consults that actually showed up.

The analytics tools themselves were never going to solve it. Google Analytics 4 will not sign a BAA, which is the whole ballgame for a covered entity, and we walk through why in is Google Analytics 4 HIPAA compliant. The pattern across the rest of the category is covered in how HIPAA limits traditional analytics platforms. They were all built on an assumption healthcare cannot make, which is that user level data can safely leave your control.

So healthcare marketing became the leftover. Not because the people running it are less sophisticated than the people running ecommerce. They are often more sophisticated, because they have had to be. It is because the tooling stopped at the compliance line, and for years nobody built the part that came after.

Curve fixed collection first. Curve AI Analyst is the second half

Curve is HIPAA compliant ad tracking, attribution, and analytics for healthcare. The mechanics are server side. Instead of a client side pixel firing browser data straight at Meta or Google, a single Curve script sends events to Curve infrastructure hosted in the United States. From there, PHI pattern detection flags risky payloads, only the fields you have explicitly mapped are allowed out, and identifiers are hashed to each platform's conversion API specification before anything is forwarded. Clean conversions go server side to Meta CAPI, Google, TikTok, Microsoft, LinkedIn, and the rest. A signed BAA is included on every plan. If you want the setup level detail, it is in our guide to HIPAA compliant conversion tracking setup.

That solved the collection problem. Healthcare marketers could finally run ads with real conversion signal without shipping PHI to an ad network.

It left a second problem sitting there. Once the data is flowing, somebody still has to go read it. Open the analytics tab, pick a date range, choose an attribution model, cross reference the paid media board, check whether the connectors actually synced, then form an opinion. That work is not hard. It is slow, which is why most teams look at their numbers monthly instead of weekly.

Curve AI Analyst removes that step. You ask the question the way you would ask a colleague, and the answer comes back from your own data.

What Curve AI Analyst is doing under the hood

It queries the data Curve already holds

This is the part that matters most, and it is the reason a compliant version of this can exist at all. Curve is not a chat layer bolted onto somebody else's analytics. Curve already holds your first party website analytics and your campaign reporting in house, in the same HIPAA covered infrastructure that receives your events.

The analytics side includes unique visitors, visits, pageviews, bounce rate, average session duration, daily trends, channels and sources, full UTM breakdowns, top pages, entry and exit pages, device and browser and operating system, country and region and city, goal and funnel conversion, and attribution models including first touch, last touch, U shaped, and assisted, on a ninety day lookback.

The campaign reporting side pulls that together with paid media performance from Google Ads and Meta, organic search from Google Search Console, CRM outcomes from HubSpot and WhatConverts, and connector freshness so you can tell the difference between no results and a sync that quietly failed.

Curve AI Analyst reads across all of it. When you ask a question that spans acquisition and outcome, it does not have to guess, because both halves are already in the same system.

Every answer is grounded in a real query

Curve AI Analyst does not generate numbers from memory or from a general model's impression of what healthcare benchmarks look like. Each question is resolved through the same query layer that powers your dashboard. The model decides which questions to ask of your data, the platform runs those queries against your organization's records, and the answer is built from what comes back. If the data is not there, you get told the data is not there rather than a confident guess.

It also handles reporting freshness honestly. Curve sees your site traffic in real time, while ad platforms report on their own lag of hours or days. When a question compares the two, the answer says so.

It is locked to your organization

Scope is not something the model decides. Your organization is taken from your authenticated session and bound to the tools before the model ever sees the question, so there is no phrasing of a prompt that reaches another customer's data. Identifier shaped values are stripped before anything moves to the model layer, on top of the PHI safeguards already applied at ingestion. Access is granted per organization, and every conversation is logged and org scoped for audit.

For a compliance officer: this is a read layer over data already collected under your BAA, inside the same boundary, with redaction on the way through.

The questions healthcare marketers actually ask it

The useful test is not whether it can produce a paragraph. It is whether it answers the questions you would otherwise spend a Tuesday afternoon chasing. Things like:

• Which channel produced the most consult requests last month, and how does that compare to the month before?
• Which landing pages are bouncing, and which ones are quietly carrying the account?
• Where in the booking funnel are people dropping, and is it worse on mobile?
• Which campaign actually filled the calendar, not just which one generated form fills?
• What did organic search bring in this quarter compared to paid, and which queries are driving it?
• How many visitors are on the site right now?
• Which of my sources look strongest under assisted attribution rather than last touch?
• Did my Google Ads connector sync this week, or am I looking at stale numbers?

The last two tend to surprise people. Attribution model and data health questions get skipped when answering them means flipping between four tabs, and they are exactly the ones that change a budget decision.

What Curve AI Analyst does not do

Being clear about the edges is more useful than overselling. Curve AI Analyst is a read surface, deliberately.

• It does not write SQL, and it does not run free form queries against your database.
• It does not create or edit goals, funnels, or event mappings.
• It does not turn connectors on or off or change how your destinations are configured.
• It does not replace the dashboard. The dashboard is still where you go to configure things and to look at a chart in detail. This is the faster path to the same data.
• It does not invent context it was not given. If you ask something outside the data Curve holds for you, it says so.

Read only is a compliance decision as much as a product one. A system that can only answer questions has a far smaller blast radius than one that can also rewrite your tracking configuration mid conversation.

Sentinel and Curve AI Analyst are the same product

If you have seen Curve talk about Sentinel on social, that is the same thing described here. Sentinel is the public facing name, Curve AI Analyst is what it is called in the product and in our documentation, and there is no difference in what it does.

What it changes about a healthcare marketing week

The honest before and after is not dramatic in a screenshot. It is dramatic in a calendar.

Before, the monthly reporting cycle is a person exporting numbers, reconciling them against what the ad platform claims, formatting them, and then writing three sentences of interpretation at the end that everyone actually reads. Anything that is not on the standing report does not get asked, because asking is expensive.

After, asking is cheap. You check the thing you were curious about on Wednesday instead of waiting for the report. A practice owner who has never opened an analytics dashboard can type a question and get a real answer, which changes who gets to participate in marketing decisions. In most practices the person with budget authority and the person who can read an attribution report are not the same person, and that gap is where a lot of bad spending comes from.

It also shortens compliance conversations. When the analytics live inside a BAA covered platform and the assistant reading them sits inside the same boundary, the question of where the data goes has a one sentence answer.

Frequently asked questions

Is Curve AI Analyst HIPAA compliant?

Yes. It operates on data already collected inside Curve's HIPAA compliant infrastructure, under the BAA that comes with every Curve plan. Access is scoped to your organization at the session level, identifier shaped values are redacted before they reach the model layer, and the assistant reads data rather than writing it. The compliance posture is inherited from the stack underneath, which was built for healthcare from the first event.

Can I just use ChatGPT for this if I anonymize the data first?

No, and this is worth being blunt about. Anonymizing a healthcare marketing export by hand is not a control, it is a hope. Page paths and campaign names carry condition information, and a row that looks harmless alone can be re identifying next to the rest of the sheet. Pasting patient adjacent data into a consumer AI tool that has not signed a BAA with you is a disclosure no matter how carefully you scrubbed the columns. Curve AI Analyst works because the data never makes that trip.

What kinds of questions can it answer?

Anything that lives in your Curve analytics and campaign reporting. Traffic and engagement, channels and UTMs, page level performance, device and geography, goals and funnels, attribution across first touch, last touch, U shaped, and assisted models, paid media from Google Ads and Meta, organic search from Google Search Console, CRM outcomes where HubSpot or WhatConverts is connected, and connector freshness.

Does it replace the Curve dashboard?

No. The dashboard is where configuration happens and where you go to sit with a chart. Curve AI Analyst is the fast path for the questions that only needed a number and a sentence of context. Most teams use both, often in the same sitting.

Do I need an analyst or a data team to use it?

No, and that is largely the point. The whole category assumed a healthcare marketing team had someone fluent in attribution modeling sitting nearby. Most do not. If you can describe what you want to know, you can get the answer.

Why can't I get this from my ad platform's reporting?

Ad platform reporting only sees its own slice, and only the version of events reported to it. It cannot tell you what happened on your site after the click, cannot see the consult booked through your scheduler, and has no view of organic or CRM outcomes. It also cannot be made HIPAA compliant on request, which is how healthcare accounts got stuck in the first place.

What do I need in place before it is useful?

The Curve script installed on your domains, analytics enabled, and your key conversion events and goals defined. Campaign reporting questions also need the relevant reporting connectors authorized and synced. If you already run Curve, you are most of the way there.

You should be able to talk to your own numbers

Healthcare marketing spent a decade being told what it could not have. No pixel you can keep. No clean attribution. No conversion signal worth optimizing on. Every request for a normal marketing capability came back with a compliance reason it was impossible, and the honest answer was usually that nobody had bothered to build it properly for this industry.

Curve is the first platform to let you talk to your analytics, your marketing, and your campaign reporting in a HIPAA compliant way. First, and the best place for it to happen, because the data is already here. It was collected inside the compliant boundary, it is stored inside the compliant boundary, and it gets answered inside the compliant boundary. You ask. It answers. PHI does not leave.

If you run ads for a healthcare organization and you have been waiting for the analytics half of your job to catch up to everyone else's, this is it. See how Curve AI Analyst fits your stack at curvecompliance.com, and bring the campaign question you have been meaning to answer since last quarter.