A conversation with Theo Ahadome, Head of Commercial, Unite

Unite is an AI-powered health insights platform that lets patients gather their complete medical history across 250,000+ U.S. hospitals and act on it through personalized recommendations. We work with pharma to drive better disease understanding, treatment selection, and patient identification, with a particular focus on oncology and rare disease.

The first customer story focused on how we use Athena BDA to identify the right brands and stakeholders to target. Athena helped solve the "who" for us. What we wanted to solve next was the "how": once we know the right people to reach, how do we consistently convert them into meetings? The new AI messaging assistant has been a game changer in that sense. It combines a pre-trained AI layer that knows how to read Athena's data with a customer-side build where you layer in your own messaging logic and execution. It runs inside Claude's Cowork product, and we put our version together in about a day.

There's a decent chunk of setup involved when you first roll this out, but once it's running, the output speaks for itself. The whole thing took us about a day end-to-end, and it runs inside Claude's Cowork product.

The way it's structured: Athena have done the foundational work to train the AI on their data, and as the customer you layer in your own messaging logic and connect the execution. There are three pieces to it.

1. The Athena BDA foundation. This piece is pre-built. Athena provide a comprehensive set of instructions that trains the AI on how to read every field in their data: what each Intent Signal represents, how to interpret pipeline news prompts, how to use conference intelligence to personalize outreach. You build the instructions into the Cowork project and turn on Claude in Chrome for direct access to the Athena Portal. It's a bit clunky via the browser today, but it works fine, and we're excited for the upcoming MCP integration to make it more seamless.

2. Our conditional messaging layer. This is where the customer build really happens, and it was the biggest piece of work from our side. The AI is only as good as the messaging logic you give it, so we invested heavily here.

We started with strict accuracy guardrails. In pharma, hallucination is non-negotiable. We needed to make sure the AI didn't hallucinate sales messaging or details about a drug. So a big part of our setup is keeping the AI tight to our approved messaging and the Athena data overlay.

On top of that, we built layered messaging examples. We share a high-level overview of Unite, standard email templates, and then more tailored versions for our key therapy areas. Where it gets really powerful is the ability to layer personalization further down. For example, for a Phase 3 oncology brand with a testing requirement, the AI knows to lead with our patient identification messaging. For brands with an in-clinic infusion administration route, it pulls in our IV infusion case study. We also built company-specific rules. When a contact works at "Company X," it always mentions our existing MSA. When at "Company Y," it references previous work within the company.

3. Execution via Gmail and LinkedIn. Athena provided the playbook for the execution layer, and we configured it for Unite. The AI sends outreach via our own Gmail accounts, rotating across three inboxes at 25 emails per inbox per day. That keeps deliverability healthy. We've extended the same setup to LinkedIn, where the AI generates and sends tailored connection messages and follow-ups. The whole motion is now genuinely multi-channel.

Our current strategy revolves around four use cases, all powered by the same underlying setup.

Job Changes. Each month we pull the top job changes from Athena and feed the list into the AI. It generates a fully personalized email sequence for each contact, and we run the campaign across our three inboxes.

Pipeline News Prompts. Same workflow for the monthly pipeline news. We review the prompts, identify the highest-priority drugs and stakeholders, and the AI handles the messaging.

Conferences. This is where the multi-layered personalization really shines. We use the Athena conference module to identify high-quality targets at upcoming events, often with detail on what they're speaking about, and the AI uses that context to draft outreach that genuinely engages with their work. It's a long way from the generic "let's grab coffee at the conference" emails everyone else is sending.

Brand-specific campaigns from the Key Drug Database. We use the Key Drug Database to identify high-fit brands, build contact lists for those brands, and then generate brand-specific messaging.

Honestly, it's been a huge success. In the first week we turned this on, we got 10 replies and 5 meetings. From our best campaigns previously, we'd be feeling good about 1-2 replies in a week. We've actually had to peel our campaign volume back since then, just to keep meeting flow manageable.

The biggest unlock has been deliverability. We've always been confident in our messaging. When we get an email in front of someone, we know we can hold their attention. The challenge has historically been getting it into their inbox in the first place, and this new setup has dramatically improved that. I think it comes down to two things working together.

First, we're sending from our own email infrastructure rather than a cold email tool. In our experience, cold email tools kill your deliverability, even when you do everything right. Sending from our real Gmail accounts at sensible volumes gets us into the inbox.

Second, the level of personalization is unlike anything we could do before at this scale. We're putting the brand name someone works on into every email, with messaging built around the specific drug, indication, and administration route. That's a world apart from the fake personalization you see in most cold emails (mentioning what school someone went to, or congratulating them on a five-year work anniversary).

Two things, and they're linked.

The first is the browser dependency for accessing the Athena portal. The Cowork browser extension works, but it's slower than it should be, and you can feel the friction every time the AI needs to pull fresh data. The second is the same kind of problem on the Gmail side. There's no native Cowork integration for sending across multiple mailboxes today, so we're using browser automation for the inbox rotation as well. Both work, but neither is elegant.

For Athena, the answer is the upcoming MCP integration. For Gmail, it's a question of native multi-inbox support landing in Cowork. Once both are in place, this whole motion goes from "fast" to "instant."

The big one is the Athena MCP. The moment that's live, every interaction with Athena data becomes faster, cleaner, and the AI can pull intelligence in real time without the browser hop. That alone is a major upgrade.

What I'm most excited about is what the MCP unlocks beyond just speed. The big one is alerts. We'll be able to flag the criteria we care about (specific therapy areas, drug attributes, target companies, geographies) and the AI will proactively surface activity against them. So when a contact who matches our criteria changes jobs, or when one of them is speaking at a conference in our region, we get told about it. That turns Athena into an always-on intelligence layer for our team.