Megan Smith

Listen, then look at the data

Not every decision is user-led. What users ask for tells me where the pain is. The data tells me how big it is and who it affects.

Customer Service

I train the support team before release so they can answer questions on day one: what changed, why, and how it works. Training is often ongoing, not a one-time demo.

I'm also an active participant in solving support tickets. Partnering closely with Customer Service reps helps me know when a problem is really a problem, keeps me in the loop early, and lets me jump on solutions quickly. It also gives me great sources for client interviews.

Sales

I help Sales understand the problem a release solves and who it's for, so they can talk about the value to practices, not just the features.

I also bridge the gap between what Sales hears and what the product does. Practices rarely describe a problem the way we do. One office asks about "claims falling through the cracks," another about "money we never got paid for." I help Sales recognize the different ways clients ask the same question, so they can connect the dots to the right solution.

Marketing

I partner with Marketing on adoption and on how a feature is represented and explained. True to pictures over pages, I sometimes bring mock logos or draft marketing articles to get the conversation started.

My goal is to make sure we never misrepresent what a feature does, while still highlighting the benefits it brings to the practice. I also share the customer stories and feedback that show the impact.

Measure and close the loop

I track adoption against the metrics we set before building. What customers, support, sales and the data tell us feeds right back into the next round of outcome decisions.

Prototype instead of describe

I use Claude to turn an idea into something people can see and react to, from small changes to whole new pages. Decisions no longer wait until someone can draw them in Figma.

See pictures over pages →

Everyday tools, custom skills

Claude does the majority of the heavy lifting across the product lifecycle, from PRDs and release notes to market analysis. I've created my own skills for each one, so tasks that used to be manual are now simple and automated.

I use ChatGPT to keep my voice and tone consistent. I've used it since January 2023, so it knows how I write. I also use it for quick marketing images, like in-app Pendo guides and social media announcements.

Analytics and tracking

I connected Claude to Pendo and Salesforce. It pulls and tracks how the product is being used and adopted, and updates the metrics every day, so data-informed decisions take minutes, not days.

See how data shapes my outcome decisions →

Claim batches sent: RCM vs. Hero

How often offices send their claim batch from the RCM Claims page, compared with the legacy All Claims page in Hero.

US offices (max reach)US offices

3,853

Pendo "Offices-US only" segment

Offices sending from RCMRCM offices

1,140

29.6% of offices · 21,480 sends

Offices sending from HeroHero offices

420

10.9% of offices · 9,120 sends

Offices sending from bothBoth

95

8% of RCM senders still also send from Hero

RCM share of batch sendsRCM share

70.2%

21,480 of 30,600 batch sends

Reach across US offices · last 30 days

• RCM only 1,045 (27.1%)
• Both 95 (2.5%)
• Hero only 325 (8.4%)
• Neither 2,388 (62.0%)

Weekly trend · Jul 19 – Oct 4 · batch sends

• RCM: batch sent
• Hero: batch sent

Show weekly table

Week of	RCM sends	Hero sends
Jul 19	200	7,700
Jul 26	220	7,300
Aug 2	250	7,500
Aug 9	300	7,100
Aug 16	420	6,800
Aug 23	600	6,200
Aug 30	1,100	5,400
Sep 6	2,600	3,900
Sep 13	3,900	2,900
Sep 20	4,700	2,300
Sep 27	5,300	2,000
Oct 4	5,600	1,900

Sent from RCM = clicks on Send All In Batch on the RCM Claims page. Sent from Hero = clicks on Send All Batched Claims on the legacy All Claims page.

Example only: these are not real numbers, just a sample of the adoption metrics I track.

Build my own AI agents

I build personal AI agents to take care of parts of my day, like a work dashboard and a reading tracker. Building them myself teaches me what agents are good at, and where they still need a person.

From dashboards to insights

Traditional analytics and dashboards leave the interpreting to the user. AI can layer on top of the data a practice already has to surface insights, spot patterns, and point out what matters.

From insights to next actions

Agents can go a step further: notice a problem, recommend what to do about it, and get the work ready for a person to approve. In RCM, that means taking repetitive work, like eligibility checks and denial follow-up, off the team's plate.

Keep people in control

In healthcare, people should be taking care of people: informing, communicating and keeping it personal. Let AI do the legwork, not make the final call. I design agents to gather, compare and prepare the work, then hand the decision to a person to review and approve. That keeps trust high while the work gets faster.

Where I'd take it next

I see agents moving from working alongside the team to working ahead of it: