Demo: Refills

Pre-Demo Intel

Company: Refills (refills.com)

What: DTC telehealth — GLP-1s, ED, hair, longevity

Volume: 15–20K tickets/month, scaling fast

Channels: 70% in-app patient chat, 30% email (Gorgias)

Current AI: MyAsk AI (getting replaced)

Moving to: Own CRM with open APIs

People on the Call

Harrison: Found us, did research, sold on concept — your ALLY

Matt Osborn: CEO — cares about scale, cost, competitive edge

CTO: Technical architecture, integration, "why not build it" — KEY skeptic

Partner: Implementation, timeline, operational fit

Discovery Recap (What They Told Us)

Scaling to 100K patients without 200 agents — MyAsk only deflects, doesn't resolve

Manual dosage change process: 15–20 min per ticket across 3 systems

Fragmented stack: Gorgias + custom CRM + Notion + MyAsk — agents switching between everything

Competitive gap with Hims/Ro on patient experience

What's in the Sandbox

16 KB articles (products, shipping, returns, subscriptions, dosage, FDA disclaimers)

4 live workflows with mock tools that fire real tool calls

Mock CRM: getPatientProfile, getOrderStatus, getPrescriptionDetails, submitDosageReviewRequest, updateShippingDate, reportMedicationIssue

Demo patient: Ryan Collins (ryancollins@gmail.com)

Watch For

Harrison will nod along — use him to validate points. "Harrison, this is what you were describing, right?"

CTO will probe on architecture — have MCP/API answers ready. Don't get pulled into a whiteboard session.

Matt will care about ROI — frame everything in headcount saved and patient satisfaction.

Partner will ask about implementation — "weeks, not months" is your answer.

Opening & Recap

3–5 min

"Harrison, Matt, great to see you again. And great to meet you both — thanks for making the time."

?Quick pleasantries only. Let Harrison introduce the CTO and partner if he hasn't already.

"Are we good on time for the next 45 minutes? I want to make sure we leave room for questions."

✓Wait for confirmation, then immediately set the agenda.

"My goal today is three things. First, I'm going to jump straight into showing you the AI agent handling your four use cases live — order status, subscriptions, medication issues, and the big one, dosage changes. Second, I'll pull back the curtain and show you what's behind it — the workflows and the knowledge base. Third, Coach — how your team stays in the loop. Sound good?"

✓Wait for yes. Then recap discovery.

"Before I dive in, I want to quickly recap what we covered last time so everyone's on the same page. Harrison and Matt, you told me the biggest challenge right now is scaling patient support without scaling headcount. You're at 15–20K tickets a month, growing fast, and MyAsk is deflecting but not resolving. The hard tickets — dosage changes, prescription questions — still take 15–20 minutes across three different systems. And the goal is to get to 100K patients without needing 200 agents. Does that still capture it?"

♪Let them confirm or add context. If the CTO or partner adds a new concern, note it — address it during the demo.

The Bold Claim

30 sec

"Before I show you anything — what I'm about to walk you through is going to show you how Refills can resolve 70% of patient tickets end-to-end, including the complex clinical ones, without a human ever touching them. Not deflect — resolve. That's the bar."

!Let it land. Don't rush past it. Brief pause, then jump straight into the demo.

Live Demo — Order Status + Subscription + Medication Issue

~5 min

"Let me jump right in and show you. I'm going to open a test ticket as a patient named Ryan Collins and walk through a real conversation. Watch what the agent does — it's going to look up my order, adjust my subscription, and file a medication issue report, all in one thread."

✓Open a new test ticket in the sandbox. Make sure screen share is visible.

Test Ticket 1 — Live Conversation

Hey, I placed an order last week and haven't gotten it yet. Can you check on it for me?

Agent pulls getOrderStatus, returns tracking info, shipping details, estimated delivery.

✓Let the agent respond fully. Point out the tool call firing in the sidebar.

Ok cool. While I have you — I'm flying to Europe on May 3rd and won't be back until the 12th. Can you push my next shipment back so it doesn't arrive while I'm gone?

Agent calls updateShippingDate, confirms the new date, shows confirmation.

✓Point out: the agent understood the context, calculated the date, and made the write action.

Perfect. One more thing — my last shipment actually had an issue. The vial looked cloudy and kind of yellowish compared to what I normally get. I used it anyway but wanted to flag it.

Agent advises on safety (don't use compromised medication next time), files a report via reportMedicationIssue, asks if patient needs a replacement.

?Point out the safety guidance + the issue filing happening simultaneously.

Got it, I'll hold off next time. Is there any way to get a replacement for that one since it looked off?

Agent confirms replacement process, provides next steps.

Great thanks for the help

Agent wraps up, confirms all three actions taken.

Section Break — Check / Questions / Tie-Down

Three different actions in one conversation. Lookup (order status), write action (subscription date change), and issue filing (medication report + replacement). Point to the pain from discovery:

"That was one conversation. Three different actions — a lookup, a date change, and a medication issue report with a replacement. Today, that's an agent switching between Gorgias, your CRM, and probably Slack to flag the medication issue. How long does that take your team right now?"

♪Let them answer. They'll say something like "10–15 minutes." Nod.

"Does that make sense? Can you see how that changes things for the team at scale?"

✓Get verbal confirmation. Pull in Matt: "Matt, what would that do for your operation at 100K patients?"

Live Demo — Dosage Change Request

~7 min · The wow moment

"Now let me show you the one Harrison was most excited about. This is the dosage change workflow — the most complex thing your team handles. The patient wants to move up faster than the titration schedule. Watch how the agent navigates this."

✓Open a new test ticket. Fresh conversation.

Test Ticket 2 — Dosage Change (The Kill Shot)

I want to get bumped up to the highest dose. I've been on this one for a while and I'm not seeing the results I want.

Agent calls getPrescriptionDetails, pulls current dose + titration schedule, explains where the patient is in the progression and what the next step would be per their provider's plan.

✓Point out: agent pulled the prescription, knows the schedule, and explained it clearly — not a canned FAQ response.

I get that there's a schedule but I feel fine. No side effects at all. Can't we just skip ahead? I'm paying the same price either way.

Agent holds firm on the titration protocol, references provider notes, explains why the schedule exists (safety), empathizes but doesn't cave.

?This is the key moment. Point out: the agent held the clinical guardrail. It didn't just agree with the patient. It followed the protocol.

What if I just ask my doctor directly? Can you at least send them a message that I want to move faster?

Agent calls submitDosageReviewRequest, notifies Dr. Mitchell, confirms the request was submitted, sets expectation on timeline for response.

✓THIS is the Harrison moment — "agent emails the doctor while still chatting with the patient." Point to it explicitly.

Ok. And what if she says no? Can I get a second opinion or switch providers?

Agent handles gracefully — explains the process for provider changes, stays within policy, doesn't make promises it can't keep.

Alright fine. I'll wait for Dr. Mitchell to get back to me. Thanks.

Agent confirms the review request is submitted, summarizes next steps, closes cleanly.

This Is the Kill Shot — Make It Land

Do NOT rush past this. Let it breathe.

"I want to pause here. What you just saw — that's your hardest ticket. A patient pushing to skip ahead on their dosage. Your human agent today has to pull up the prescription in the CRM, cross-reference the titration schedule in Notion, hold the clinical line without upsetting the patient, and then go into a separate system to flag the doctor. That's 15–20 minutes across three systems. What just happened took one conversation turn."

!Let the silence work. Don't fill it.

"Harrison — you said this was the one that excited you most. The agent emailing the doctor while still chatting with the patient. Is that what you envisioned?"

♪Let Harrison validate. He'll sell it to the room for you.

"Can you see how this changes things for your team?"

✓Get verbal buy-in from the room. Pull in each person if needed.

Workflows & Knowledge Base

~5 min

"Now let me pull back the curtain and show you what was behind everything you just saw."

✓Navigate to the KB in the Lorikeet admin.

"Everything the agent just did draws from two places: the knowledge base and the workflows. I built out 16 articles from your site — your products, shipping policies, return policies, subscription management, dosage titration schedules, FDA disclaimers. This is the agent's brain. It doesn't hallucinate answers because it only draws from what's here."

✓Scroll through a few articles. Show the dosage one briefly — don't dwell.

"Then we have the four workflows you just saw in action — one for each use case Harrison flagged. Each one has its own logic, its own tools, its own guardrails."

✓Click into the dosage change workflow to show the structure. They just saw it work — now show them why.

"Behind these workflows are mock versions of your CRM tools — getPatientProfile, getOrderStatus, getPrescriptionDetails, and action tools like updateShippingDate and submitDosageReviewRequest. In production, these connect directly to your APIs."

CTO Hook

Address the CTO directly here:

"[CTO name] — one thing I want to flag. Everything I just showed you — the articles, the workflows, the tools — I built all of it programmatically through our MCP integration. The same Claude integration you're already using. That means your engineering team can build, test, and deploy changes to the AI agent the same way they ship code. No drag-and-drop builder. No vendor lock-in."

♪Let the CTO react. If they ask a technical question, keep it to 2–3 sentences. "Happy to do a technical deep-dive on that in a follow-up with your engineering team."

Section Break — Check / Questions / Tie-Down

"Does that make sense? What questions do you have before I show you Coach?"

Coach Agent & MCP

3–5 min

"The last piece I want to show you is Coach. This is how your human agents stay in the loop when the AI needs oversight or when a ticket escalates."

✓Show the Coach interface briefly. Don't go deep — just show the concept.

"Coach gives your team real-time visibility into what the AI is doing. They can review, approve, or step in at any point. And every resolved ticket feeds back into the system — the agent gets smarter over time."

CTO Section

Address the CTO directly.

"[CTO name] — I mentioned earlier that everything was built through MCP. Let me explain what that means for your team. Lorikeet exposes the entire platform through an MCP server — the same protocol Claude uses. Your engineers can build and manage the entire AI support operation through Claude Code, Cursor, or any MCP client. Articles, workflows, tools, guardrails — all of it is code. It deploys like code. It version-controls like code."

"You told me you're heavy Claude users already. This means the integration isn't a new paradigm — it's the same paradigm you're already in. You're not learning a new platform. You're extending the one you already use."

♪Let the CTO engage. If they ask deep technical questions, keep answers concise: "Happy to do a technical deep-dive on that in a follow-up with your engineering team. The short answer is [yes/here's how]."

Section Break — Check / Questions / Tie-Down

"Does that make sense? What questions do you have on the technical side?"

Multi-Stakeholder Engagement

2–3 min

!Before moving to the close, make sure every person on the call has been engaged. Don't let anyone sit passively.

"Before we wrap the demo portion — I want to make sure everyone's had a chance to weigh in."

"Harrison — you've seen the most of this. Does what I showed today match what you envisioned when you found us?"

✓Harrison is your ally. Let him sell the room. He'll say something enthusiastic — let it breathe.

"Matt — from a CEO perspective, thinking about scale and unit economics — what's your take on what you've seen?"

♪Matt cares about cost and competitive positioning. If he asks about ROI, frame it as: "At your current volume, the AI resolving even 50% of tickets saves you [X] agent hours per month."

"[CTO name] — I know you're thinking about this from an architecture standpoint. What questions are top of mind for you?"

♪Address technical questions concisely. Offer a follow-up deep-dive if needed.

"[Partner name] — from an implementation standpoint, what are you thinking?"

♪Partner will ask about timeline and lift. Answer: "Implementation is weeks, not months. We handle the heavy lifting on setup."

Power Tie-Down

1–2 min

"Based on everything I've shown you today — the four use cases live, the dosage change workflow, the knowledge base and workflows behind them, Coach, and the MCP integration — can you see yourself using Lorikeet and adopting it into your organization? Or do you foresee any challenges?"

!This is the most important question of the call. Stop talking. Wait for the answer.

♪If they say yes — move directly to pricing. If they raise a concern — address it, then re-ask.

?If the CTO raises "why not build it" here, go to the objection handler below. Don't get defensive.

Pricing & Next Steps

5–7 min

✓Temperature check came back positive. Now present pricing.

"Let me walk you through how pricing works."

"We have three plans. Our Start plan is $1,500 a month. It includes the core platform, KB, workflows, and one channel. I don't think this is the right fit for you, and I'll tell you why in a moment."

"Our most popular plan is the Scale plan at $4,000 a month. It includes everything in Start, plus multi-channel support, advanced workflows, Coach, full API access, and priority onboarding. This is where most companies at your stage land."

"Then we have Enterprise, which is fully custom for high-volume or highly regulated environments."

"Based on what you've shared — 15–20K tickets, two channels, complex clinical workflows, and the need for API-first integration — what I recommend for Refills is Scale. The multi-channel support covers both your in-app chat and email. The advanced workflows handle the dosage change complexity. And the API access means your engineering team can manage everything through MCP."

✓Pause. Let it land.

"Which one of those do you feel serves you best?"

♪Wait for response. If they lean Scale, confirm. If they push back, don't negotiate — reframe value.

once plan is confirmed

"Great. Here's what next steps look like. First, I'll give you sandbox access so your team can poke around and validate everything you saw today. Second, we schedule a technical deep-dive with your CTO and engineering team to walk through the integration, API, and MCP setup. Third, we lock in a timeline and get you live."

"Harrison — you mentioned timing was urgent. You said 'I wish we had this yesterday.' What does the timeline look like to get started?"

✓Push for a specific date. "Can we get the technical deep-dive on the calendar this week?"

Objection Quick Reference

"Why not build it ourselves?"

"You absolutely could. The question is whether you should. Building the agent is 10% of the work. The other 90% is the orchestration layer — guardrails, escalation logic, Coach, quality scoring, the feedback loop that makes it better over time. That's what we've spent two years building. Your engineering team's time is better spent on your core product — the patient experience, the CRM, the clinical tools. Let us handle the support AI so they can focus on what moves the needle for Refills."

"What's the implementation timeline?"

"Weeks, not months. What I showed you today was built in a few days. Production deployment depends on your API readiness, but we typically have companies live within 2–4 weeks. We handle the heavy lifting on KB build-out, workflow design, and tool integration. Your team's main lift is connecting the API endpoints."

"What about HIPAA / compliance?"

"We're built for regulated environments. Eucalyptus is a telehealth company and they run on us. We can walk through our security architecture in the technical deep-dive, including data handling, PHI, and audit logging. Happy to share our compliance documentation ahead of that call."

"What if the AI gives wrong medical info?"

"That's exactly why we built guardrails. The agent only draws from your approved knowledge base — it doesn't hallucinate medical advice. For clinical decisions like dosage changes, the workflow is designed to defer to the provider, not make the call itself. You saw that in the demo — the agent held the line and escalated to Dr. Mitchell. And Coach gives your team real-time oversight to catch anything that doesn't look right."

Refills-Specific Reminders

Don't let the CTO derail into a whiteboard session — offer a follow-up deep-dive

Don't trash MyAsk AI — "it's great for deflection, this is resolution"

Don't get into Gorgias migration details here — save for implementation call

Don't over-explain the tech to non-technical people — save MCP depth for CTO only

Do use "patients" not "customers" — their language

Do let Harrison sell the room — he's already bought in, use him as an ally

Do reference Eucalyptus when asked about telehealth experience — same complexity, proven results

Do frame everything for Matt as headcount saved — "100K patients without 200 agents"

Do use Harrison's quote back to him — "I wish we had this yesterday"

Do lead with the live demo — show don't tell, then explain the how

Do point out tool calls in the sidebar during demos — makes it tangible, not magic

Do remember they're "not married to any of it" — they're open to full stack replacement