How Cavalry Built an AI-Native Outbound Engine That Booked 111 Demos in 120 Days

Impact TL;DR

• 111 enterprise demos booked in 120 days in a market where everyone else is asking for the same demo
• 4.21% reply rate on 106,816 emails sent to 25,201 unique target contacts
• 10.45% of replies turned into booked demos
• 8 campaigns running in parallel, with deliverability orchestrated across multiple inboxes and domains
• Meetings per month doubled from the first months to peak: about 12.5 a month early on, then 53 in March

‍

The problem

They had tested outbound. Now they needed someone to scale it as a system.

Cavalry sells AI for customer support. The buying committee they want to reach (heads of customer experience, e-commerce founders, ops leaders at growing brands) gets the same email from twelve different competitors every week. Siena, Forethought, Gorgias, Octocom, Klark, and a list of newer AI customer service tools all open with "book a demo." That ask stopped working a long time ago, so Cavalry wanted a different approach.

The competition isn't just crowded. It's bigger. Most of the AI customer support companies Cavalry is up against have more revenue, more people, and more money to spend on the same audience. Outspending them isn't a realistic option.

Cavalry's team had already run some outbound and seen it work. What they didn't have was time. Running a serious outbound program (improving data accuracy across target accounts, finding new contacts inside the buying committee, writing the copy, watching deliverability, replacing the campaigns that stop working) takes a team that does only that. Cavalry's team was building the product.

So they went looking for a partner who could take what was already working, build the GTM orchestration around it, and keep it running. Outbound was also only one piece of how they were going to market. They were running ads against the same target accounts in parallel, and wanted the outbound to fit into that wider GTM motion.

‍

The Solution:

Growth Today built Cavalry's outbound around what we call The Roundtable Play, anchored on AI-native personalisation at scale. Every email opens by asking the buyer what they think about a problem they're already dealing with. The end of the email asks for their thoughts, not a demo. Heads of customer experience and founders are much more willing to share how they're handling something than to give up half an hour to a pitch. The conversations that come back are real, and when the buyer's setup matches what Cavalry does, the fit conversation builds organically and the sales rep gets to sell.

The Roundtable Play is closer to account-based marketing than to traditional outbound. Every email is built around the buyer's situation.

For Cavalry, that meant:

• Improving data accuracy across the target account list: new e-commerce brands added every cycle, dead contacts removed, new buying-committee members found as roles changed inside the accounts
• Persona-led messaging across the buying committee: five distinct personas (founders, customer experience leaders, marketing leaders, e-commerce leaders, ops/tech leaders), each getting a different problem hook, question, and reason to care
• Refreshing copy before reply rates decrease: rewriting the copy regularly so the messaging stays fresh
• Orchestrating 8 campaigns in parallel: different angles running concurrently, so the same audience got a different conversation regularly
• Continuous data and signal monitoring: when one campaign started outperforming the rest, we pulled what was working into the others

‍

Use case 1

Persona-led messaging across the buying committee

The Roundtable Play only works if the buyer reading the email feels like it was written for them. We mapped the buying committee at Cavalry's target accounts into five personas. Each persona got a different question, because each one cares about a different problem:

‍

‍

For each persona, we built five layers into the email:

• The opening line: anchored to what was on their mind that week. For a head of customer experience: "Many CX leaders say the toughest part of Q4 is keeping response times down while the team gets significantly more tickets to handle."
• The question: what we asked them. For the same person: "How are you planning to manage the upcoming high ticket volumes this holiday season?"
• The topic: the bigger thing the question fits into (for a customer experience leader, the pressure to be available 24/7)
• The reason it matters: why this topic matters in their role, not just to the company
• The questions they're already asking themselves: three or four things we knew they were thinking about anyway, so the email matched the conversation already happening in their head

On top of persona-level mapping, every email also pulled in AI-assisted personalisation:

• Contact-level personalisation: the buyer's role and tenure, their recent public activity, their current work focus
• Account-level personalisation: the company's stage, any signals (funding, new product, hiring spike), and their tech-stack indicators
• Buying-committee routing: when the email reached someone who wasn't the right buyer, it named the actual decision-maker and asked for an introduction

‍

Use case 2

Orchestrating 8 campaigns in parallel, refreshing what stops working

One angle gets tired fast. Send the same email to the same buying committee week after week, and reply rates decrease.

Across 120 days, 8 campaigns ran in parallel. Each one anchored on a different reason someone in the buying committee might want to talk:

• The support team is about to get crushed by holiday volume: aimed at brands going into peak season (Black Friday, Cyber Monday, Christmas)
• The chatbot they tried didn't work: aimed at brands that bought an AI chatbot, hated it, and turned it off
• The tools they use don't talk to each other well: aimed at brands whose CRM and payment systems have known gaps
• They didn't reply to the first email, so try a different angle: aimed at contacts who'd already been emailed but hadn't responded

Each target contact got multiple emails over a couple of weeks. When one campaign started outperforming the rest, we copied what was working into the others. The campaigns that fell behind got rewritten or killed.

The gap between the best and the worst campaign is what makes this part of the work matter:

• Best campaign: 9.68% reply rate, 50 booked demos from about 21,000 emails
• Worst campaign: 0.61% reply rate, 1 booked demo from about 15,000 emails

That's a 16x difference between the best and the worst. You need to regularly update the copy to maintain and increase the performance of your outbound campaigns and react to new angles (e.g. AI, Claude Code).

‍

Use case 3

Continuous GTM operation

Booking meetings 120 days in doesn't happen by itself. There's a lot of work that goes on between campaigns:

• Improving data accuracy across target accounts: adding new e-commerce brands that fit the ICP, removing the ones that don't
• Finding new contacts at target accounts: people change jobs, new heads of customer experience get hired, founders leave (the buying committee is a moving target)
• Monitoring deliverability: making sure emails land in primary inboxes, adjusting send volume per address based on inbox performance signals
• Multi-stakeholder buying-committee coverage: when an account has multiple decision-makers, each one gets a different conversation

‍

Outbound Performance

How the months went

• Months 3 and 4 (Jan and Feb): about 12.5 demos a month while we were still building the target account universe and launching the first campaigns
• Month 2 (March): 53 demos booked. By then, the first round of campaigns was working, and we'd already learned what was landing and what wasn't.
• Month 1 (April): 33 demos booked. April was a transition month. We were closing out the older campaigns and getting new ones live. The new campaigns from late April booked 5 demos in their first 7 days.

Why you need a GTM operator, not just "build engine" and leave it as is:

Reply rate decreased in the middle, then came back up in April after we replaced the older campaigns with new ones. Just like any GTM motion, the message can be cooled down, and deliverability fundamentals can change. You need a team to maintain, systemize, improve, and orchestrate these regularly; a GTM motion is never a set-and-forget thing.

‍

Why Cavalry Chose Growth Today

They wanted a GTM orchestration partner

Cavalry didn't want someone to ship one campaign and disappear. They wanted an AI-native partner who'd build the data infrastructure, find the right contacts at each target account, run the messaging, orchestrate deliverability, and keep iterating week over week. Their team needed to stay on the product.

They were also thinking past single-channel outbound. They were running ads to the same buying committee in parallel, treating each channel as one piece of how they were getting in front of customers. They wanted the outbound to fit into that picture.

Getting in front of target accounts happens through a mix of ads, content, events, and outbound. Most of the time, all of them work together as one bound motion. The Roundtable Play is the outbound piece. The same target account data and buying-committee mapping feeds into ads, retargeting, content, and anywhere else Cavalry layers on next.

Results & Impact

• 111 demos in 120 days, in a market where everyone else is asking for the same demo
• 53 demos in the best month, with the engine producing through April, even while we were switching campaigns over
• Increased brand awareness through engaged contacts at the target accounts
• The work doesn't stay in a silo. The target account data and buying-committee mapping we built for outbound carry over to ads, retargeting, and anything else Cavalry layers on top.
• The engine keeps improving because we keep the data accurate, the copy moving, and the deliverability clean month to month.

‍