What AI-Native Research Actually Looks Like: Lessons from Calendly and Included Health

Recruitment has been the last piece of the research process to change.

Planning, moderation, synthesis, analysis: all of it has been reshaped by AI over the past two years. Recruitment stayed stuck, and has been synonymous for pulling lists, managing spreadsheets, writing screeners, sending mail merges, chasing booking links, assigning NDAs, and mailing gift cards. For some teams that is ten or more tools for one study, and each step carries it’s own governance risk: PII moving between systems, consent that has to be captured, tax obligations on incentives, and participants who can be overcontacted by a stakeholder who meant to email 50 people and emailed 1,000.

That risk used to be manageable because access was narrow, but that’s no longer the case with the building products becoming increasingly democratized.

Product roles are blending, and stakeholders across the org now want direct user access. The question every Research and Research Ops team is facing: how do you keep the quality bar high without giving up speed, compliance, or participant experience?

That is why we built Rally Agent and Rally MCP, both now open in beta.

Rally MCP puts Rally's tools inside the agentic systems your company already uses: Claude, Cursor, ChatGPT, Microsoft Copilot. Rally Agent is an out-of-the-box research ops recruitment agent available in Slack, Claude, and in Rally's new agentic interface. Both sit on top of the same research infrastructure we have spent four years building, which means your permissions, templates, governance rules, approval workflows, and role types carry over completely. If someone can't launch recruitment in Rally, they can't launch it through the agent. If someone can't see PII in Rally, they can't see it through the MCP. Nothing destructive happens without a human approving it first.

The philosophy is simple: the human work in research stays human. Everything else gets assisted or automated. And the agent doesn't draw the line on who can do what. You do.

At our launch webinar, we brought on two customers who have been living inside this for months: Mahad Bullo, Senior Staff Researcher at Included Health, and Ray Mason-Murillo, the first Research Ops hire at Calendly. Here is what they told us.

Research is already democratized. The question is whether it's governed.

At Included Health, researchers, designers, and PMs all run research, with Rally acting as what Mahad calls the “nerve center” of their practice. Since adopting Rally, the team has 4x'd its research volume. Governance carries extra weight in healthcare, where HIPAA and PHI protections are not optional, and Included Health covers 84 million lives.

Calendly had the opposite starting problem. Plenty of people were doing research when Ray arrived, but you’d have been hard pressed to find out about any of it. There was no central visibility and no tracking of volume at all.

So while Ray can’t say whether volume went up, he’s seen a major improvement in study quality resulting in the team trusting insights more than ever before. Product managers, designers, and go-to-market folks are still running their own research, just more responsibly and with higher visibility across the organization.

"AI native" means something different depending on where you sit

For Ray, it means getting out of the enablement bottleneck. He used to be the person cleaning up study titles and teaching everyone the naming conventions and indexing standards. Now a designer or PM describes a study, the agent names and structures it the way Ray taught it, and the cleanup work goes away.

For Mahad, it means asking what was impossible before. His example: connect a participant in Rally to their Marvin interview and their Maze prototype results, and suddenly mixed-method work is accessible to people who were never trained in those methods. Same for reporting, where skills let his team produce anything from a quick single-interview insight to a full readout, faster and in more layers.

Both landed in the same place on the interface. Mahad said he could not remember the last time he opened the Rally UI, other than a call with his CSM the day before, as he’s been working almost exclusively through the MCP. 

The first thing each of them tested

Mahad went straight to building a study from a plan. It worked, so he went into Rally to check whether it had actually captured everything and found that it had. From there he chained Rally's MCP to Sprig's and wrote a skill so that creating a Rally study also creates the in-product prompt. One less thing to remember on every study.

He is now pointing the same approach at panel health. Building the panel was the exciting part. Managing it over time is the part that gets overwhelming, and that is where he expects the MCP to earn its keep.

Ray started somewhere completely different, which he flagged live as a difference in how the two of them think about Research Ops. His pressing problem was a disorganized research repository, so he tested search. Pick a research topic, return the prior studies and specific interviews that match using transcript search, flag the strong matches, and say plainly when something looks like new territory. It came back fast and accurate with almost no fine-tuning. He checked the results himself, and they held up. His reaction: this does not feel like a V1.

Where they see their roles going

Ray is moving from doing research operations to enabling it. His 6 to 12 month goal is C-suite access: executives using plain language in Claude to access prior research without feeling like they are in the weeds. Meeting them where they already work, rather than asking them to learn a new tool they will never open.

Mahad wants two things: first, a formal education and enablement program, which he named as a real gap at Included Health and one he thinks standardized skills and MCP workflows finally make possible. Second, deeper mixed-method work: an observability layer built with the data science team that follows a customer across Rally, Amplitude, and product usage, feeding data-rooted personas that Rally can then recruit against.

He’s already running three to four studies a week with the team behind one of their flagship products, answering everything from ID card copy to bigger strategic questions. The point he keeps making internally is that research is not the slog product teams remember. The infrastructure exists so teams can spin up a study, recruit from the panel, and report out quickly. The old mental model is the thing that needs dislodging.

The line that summed it up

When governance is built into the agent, you stop worrying about whether someone is going to break a rule, because the rule is already embedded. Research and Research Ops teams are the ones positioned to architect the systems that let research scale safely. Rally Agent and MCP are the tools for building them.

Rally Agent and Rally MCP are open in beta. If you're a customer, reach out to your CSM to get access. If you're not, get in touch and we'll set up a deep dive.
‍