Ask Simon: AI answers clinicians trust
Ask Simon is a conversational AI agent for clinicians. It answers a clinical question in plain language, discloses the sources it drew on, and provides in-product source verification. Adhering to FDA regulatory guidelines and earning clinician trust.
My Role
Product design (contract)
Research
- Regulatory research
- Guidance synthesis
- Requirements definition
- Scope definition
Design
- Interaction design
- Information architecture
- Progressive disclosure system
- Content design
- Visual design
- Design system
Regulatory Research
Adopting FDA regulatory guidelines, and earning clinician trust
The FDA’s clinical decision support guidance landed in January 2026 and changed what an AI tool is allowed to say to a clinician. Reading it closely, I was able to identify four criterion that applied to the current version of Ask Simon. Synthesized what changes needed to be made and provided the product and leadership teams with the design requirements to fulfill them.
Criterion 1: Forbidden inputs
Must not acquire, process, or analyze medical images, IVD signals, or patterns and signals from signal acquisition systems.
Criterion 1: Forbidden inputs
Key details
- Pattern means multiple, sequential, or repeated measurements. Discrete point-in-time measurements, such as routine vital signs at a visit, are not patterns.
- Interpreting the clinical implications of a signal, pattern, or image fails, even when working only with derived data.
Criterion 2: Allowed inputs
Must display, analyze, or print medical information about a patient, or other medical information: peer-reviewed studies, practice guidelines, textbooks, drug labeling, government recommendations.
Criterion 2: Allowed inputs
Key details
- The 2026 update clarified that information does not have to be commonly discussed in clinical conversations. It needs relevance supported by well-understood and accepted sources.
Criterion 3: Outputs
Must support or provide recommendations to a healthcare professional, not specific directives, about prevention, diagnosis, or treatment.
Criterion 3: Outputs
Key details
- Acceptable outputs include lists of options, prioritized lists, and lists of follow-up or next-step options.
- New in 2026: the FDA will exercise enforcement discretion when only one option is clinically appropriate and the software otherwise meets every criterion.
- Software intended to replace or direct clinical judgment fails.
Criterion 4: Transparency
Must enable the healthcare professional to independently review the basis for recommendations, so they do not rely primarily on the software.
Criterion 4: Transparency
Key details
- Time-critical and urgent-decision use cases automatically fail, because there is no time for independent review.
- The FDA explicitly endorses progressive disclosure: prioritizing the most decision-relevant information, and making additional detail available as appropriate.
Research document
The Disclosure System
Three surfaces along the user journey
Replacing a static disclaimer in the footer in Ask Simon V1, there are three disclosure moments directly in the user’s flow: while waiting for a search result, the search response, and a persistent information icon. Each providing progressive disclosure of the most pertinent information to clinicians that’s clean and easy to understand.
While Ask Simon searches
Sources surface one at a time as they are consulted, so the basis for the answer is visible before the answer is.
The response screen
Interactive source count indicators are displayed next to each piece of content.
Source details
Tapping a source count indicator displays the list of referenced sources, their URLs, and a general disclosure describing how Ask Simon prioritizes and handles source references.
Information icon
How Ask Simon works and where it stops is always available in a persistent header icon. Easy access, full transparency, in plain language.
Source Verification
Verification without leaving the conversation
Opening a citation in v1 meant a new browser tab, and most clinicians never came back to finish the conversation. V2 brings the source inline: open it, read it, scroll back, keep going. Sources that can’t be framed fall back to an external link. It looks minor in a screenshot, but continuity of the clinician’s mental model is what keeps Ask Simon useful instead of frustrating.
The modal iframe
Source web pages are displayed directly in the Ask Simon canvas. Keeping the clinician in the app.
The fallback
Some publishers refuse to be embedded. Those fall back to an external link, so the citation is always reachable.
Start Screen
Orientation and trust signals
Ask Simon v1 dropped clinicians into an empty chat with no orientation, which meant first-prompt friction and early abandonment. V2’s start screen opens with tappable example prompts, drawn from real use cases. Offering clinicians a mental starting point to the conversation. And a simple disclaimer about sources being gated to respected and often used organizations is a meaningful trust signal for clinicians.
Example prompts and trust signal
Three example prompts drawn from real use cases, and a plain statement of where Ask Simon will and will not look. A clinician arriving for the first time can see how Ask Simon works and where it gets its information.
Privacy guardrails
Potentially sensitive information can be identified and flagged prior to prompt submission. Providing the necessary guardrails to ensure privacy, and signaling that safety and privacy are important.
Outcome
Shipped June 2026: web, iOS, and Android.
Reading the FDA’s guidance closely surfaced four clinician needs, and Ask Simon v2 answers each one with its own surface:
Three disclosure moments built into the response itself
Sources that open in place instead of a new tab
A start screen that orients the clinician and flags sensitive prompts before they are sent
V2 shipped as a single release carrying all of it, adopting FDA regulatory guidelines and helping to earn clinician trust.
