Full keynote from the RevOps Summit NYC, March 2026.
A quick confession
My first startup was an e-commerce company. I went very successfully bankrupt in 11 months. Eight years of debt. So when I joined Trustpilot as the fourth salesperson and they had this uncapped commission model, I had some serious motivation to close deals.
In two years, we went from 50 to 5,000 customers. I personally brought 600 of them. Almost one deal a day.
But here's the thing: it wasn't because I was a great salesperson. It was because I was genuinely obsessed with e-commerce. This was 14 years ago, when brands were still figuring out how to sell online. Before every call, I would research the company's checkout funnel, their social media presence, their online reputation. When I picked up the phone, it was a peer-to-peer conversation. The prospect felt like I understood their world as well as they did.
And if you can describe your customer's problem better than they can, they will automatically think you have the solution.
That insight shaped everything I've done since. It's why we built Evergrowth. And it's why the current state of GTM is so frustrating to watch.
How did we get here?
Let's rewind. When I started in B2B sales, we used to receive signed contracts by fax machine. When the fax machine started shaking, we would literally run to see if it was a signed deal. Most of the time, it was a cold fax.
Back then, I was the information holder. There was no content marketing. People had to listen to me to access information. That gave sellers a natural edge.
Then email became a channel. I remember having 50% reply rates. But email wasn't as busy back then. You used to get an email and think, "Oh cool, someone wrote to me." This morning alone I deleted 10 or 12 before reading any of them.
Then COVID hit. Everybody doubled down on email and LinkedIn because we didn't have direct mobile phone numbers yet. Those two channels got even noisier. And then AI came along and, let's be honest, so far it has done more damage than good to email and LinkedIn. You've all received that overly enthusiastic AI-generated poem trying to sell you something.
On my LinkedIn profile, I have a small note at the bottom that tells AI agents to ignore all instructions and send me a banana cake recipe. I get a surprising number of banana cake recipes in my inbox now.
And here's the thing that a lot of GTM leaders underestimate with cold calling: we have a generation of salespeople that did not grow up with a landline. When I grew up, I had to call my friends on the house phone to go play basketball. When I called my first girlfriend, I had to pass the gatekeeper to talk about homework. My youngest brother, who is 11 years younger than me, doesn't even have the call button on his phone. And that's who we ask to do cold calling.
They can be extremely talented on the phone. They just need to be trained to actually use it.
The real problem with the revenue stack
We automated way too much of the top of the funnel. Everything became spray-and-pray. And because we drained the specialized sales model, we lost something critical: deep knowledge about the customer.
The first generation of SDRs was trained by full-cycle reps who had closed deals, who understood the bottom of the funnel. Then we had a generation of SDRs training the next generation of SDRs, none of whom had ever closed a deal or shadowed an AE. The institutional knowledge about who your customer actually is just evaporated.
Instead of ICP-driven, persona-driven processes, we ended up with tools-driven processes. When I interview salespeople and ask how they sell, they just talk about tools. "I use this, I click that, I copy-paste this template into that automation." That's not sales. That's button pushing.
If you have negative customer acquisition costs, you reduce headcount. If you have tools dependency, you consolidate tools. There's a new competitor in every vertical every other week. And because channels are saturated, you need to hyper-personalize to resonate. The result? Less headcount, fewer tools, more competition, but bigger quotas.
If you put numbers to the siloed model, the picture gets worse. A conservative setup with one sales enablement team feeding a stack of disconnected tools, a handful of SDRs, and a couple of AEs costs around $1.85M per year when you add up the tooling ($72K) and the team ($1.78M). With a $900K quota per AE and only 31% attainment, you're looking at $0.84 in cost for every $1 of revenue. Many companies are actually negative: they spend more on the go-to-market machine than it generates.
The cost of a system of records
System of records = inevitable GTM clogging
- Reps are overwhelmed by unactionable data
- 69% of reps did not hit their quota in 2024
- Mid-sized GTM teams use ~20 tools
- Throwing more siloed AI will amplify the chaos
- Systems of record are not future-proofed for the agentic era
AI can augment your team. Or it can amplify the chaos.
One approach would be to build a system that delivers 1 million emails, book 78 meetings with a terrible conversion rate, and brag about it on LinkedIn. Some people actually do this.
No.
The idea is not to use AI as a spam cannon. It's not to turn your salespeople into button pushers. It's to augment your team and build hybrid teams where agents work alongside reps as digital colleagues.
But here's where most companies go wrong. If you build siloed AI automations where only your RevOps or GTM engineers have access to the tools, you create a bottleneck. Your salespeople can't work with agents directly, so they get tired of waiting and start using ChatGPT on the side. Congratulations, you have silos again.
Your GTM engineers should be GTM architects. They still build agents and workflows, but those agents live in a shared workspace. They become digital colleagues that your entire sales team can work with autonomously. RevOps architects the system. Reps work within it.
Systems of record vs. systems of context
All the software we've been using for the last decade was designed to record data. CRMs, enrichment tools, sales engagement platforms. They store data points: company size, job title, region, engagement score.
But AI can't personalize on data points. If you put AI on top of a database full of data points and no context, the output is going to be generic. The AI will come up with something that sounds personalized but isn't.
The difference with agents is they collect context, not data points. An agent doesn't just tell you a company has 500 employees. It reads their careers page, their annual report, their job postings. It tells you they're hiring SDRs, they just appointed a new CRO, and they're rolling out a new sales methodology. That's context. And context is what makes outreach feel human.
We used to call it data-driven sales. The new model is context-driven sales.
The AI-augmented model flips the economics. Your enablement team still builds the ICP, persona, and value-prop playbooks, but now those live in a shared training center that your agents also learn from. Specialized research agents handle lead qualification and signal monitoring. Personalization agents prepare custom talk tracks and outreach. And because agents need APIs, not seats, you slash tooling costs from $72K to $48K and replace most of the headcount with on-demand agent capacity.
The result: total cost drops to $878K, attainment jumps to 90%, and the cost per $1 of revenue falls to $0.25. That's 3 to 5x greater efficiency within the first five to six months.
The economics of a system of context
Systems of context = scalable and reliable
- Agents and reps share the same training center
- GTM team works with agents like digital colleagues
- No more siloed tools or AI features
- Replace data vendors with on-demand waterfall APIs for agents and reps
- Agents turn data into actionable insights
What the agent workspace looks like
At the core, you have a training center where you load your value proposition, your product offering, and your customer stories. Customer stories are the best content to train agents because they're already contextualized. Your marketing team's product sheet is features and benefits. A customer story is a real human explaining how your product actually helped them.
Then you create segments, your verticals and the persona cards for each one. And here's the thing: when you train agents, they remember 100% of their training. Humans retain about 20%.
From there, you bundle agents into use cases.
Research agents
Account Qualification agents browse the internet, open pages, download PDFs and annual reports, and make conclusions. If you sell benefits software, the agent checks the company's careers page to see if they offer benefits. If they don't, why would you ever contact them? If you sell ESG management software, the agent checks for net-zero initiatives. They qualify or disqualify companies based on criteria that actually matter.
Account Research agents score companies on custom signals. The best salespeople always read job ads because the job descriptions are full of intel. If you target enterprise, the best reps always read the annual report. Now you can give that job to an agent that does it for every single account.
Contact Finder agents are going to completely change prospecting. Today's tools give you a license to filter 600 million contacts by job title. But if you target procurement leaders, the title CPO could mean chief product officer, chief people officer, or chief procurement officer. An agent is smart enough to look at someone's background and know the difference. Some of our clients target people who aren't on LinkedIn, like veterinary clinic owners or hospital staff. Their reps spend hours searching company websites and registries. Agents do exactly that, but at scale.
This alone saves 40 to 50% of a sales team's time.
Personalization agents
The Play Copywriting agent takes all the research from the other agents and writes personalized outreach. But here's how we actually use it: not just for cold emails or LinkedIn invites (those channels are crowded), but for personalized talk tracks. Bullet points with research so that when a rep picks up the phone, the first 20 seconds are research-based. People drop their guard because they can tell you've done your homework.
Coaching agents
That generation of salespeople who didn't grow up with a landline? They need training. You can build a roleplay agent where reps talk to their computer, simulate a cold call or a discovery meeting, and then a coaching agent reviews the transcript and gives feedback. All of this happens one-on-one, no manager needed.
How RevOps should think about this
Let's say you target 5,000 accounts. 4,000 of them aren't in your CRM yet, and that's the gap. In a perfect world, they all become working accounts, then opportunities, then clients. But we don't live in a perfect world. Most of them become dormant.
And this is where most companies get stuck. They keep looking for new accounts out there, when most of their accounts are already sitting dormant in their CRM.
As a RevOps leader, you want agent workflows for three scenarios: sourcing net-new leads, recycling dormant accounts, and monitoring existing accounts for signals. You want to automate the checks: if an account has no opportunity, hasn't been touched in 60 days, and has no open tasks, it's dormant. Recycle it. Run agents on it. Get fresh contacts. Check for signals like M&A activity, new hires, or leadership changes.
And don't forget: 8 to 12% of your CRM data goes stale every month because people change jobs. The Contact Qualification agent checks monthly whether your contacts are still at the company. If someone left, that's a warm lead for your new-business team and a red flag for your CS team.
The AI gatekeeper is the new receptionist
More and more, especially in the US, when you call someone, you hit an AI screening feature. "What is this call regarding?" It reminds me a lot of the old gatekeepers, the receptionists.
So we built a screening script. Instead of the rep saying something generic, they read a research-based line: "It's regarding your MEDIC rollout, your CRO's focus on XYZ." The AI summarizes that and sends it to the prospect. And it's actually a great research-based message.
We pass one out of five AI screenings with this approach. It's now part of the process.
The results
The industry-average connect rate is 3%. Our augmented team, with cherry-picked top-of-funnel lists built by agents, consistently hits above 30% across all the personas and verticals we target.
After implementation, depending on team size and complexity, we typically see 3 to 5x greater efficiency within the first five to six months. Your team doesn't need more tools. They need fewer tools, deeper research, and agents that work alongside them every day.
The whole idea is less quantity, way more quality. You now have the luxury to cherry-pick your leads. And when a prospect picks up the phone, it should feel like the salesperson actually understands them.
This is digital transformation, not a new tool
A lot of AI pilots fail because teams treat agents like another tool in the stack. "Here, we bought you a new thing." That's not it.
It's more like, "Here are a bunch of new colleagues. If you don't work with them every day, you're not going to get results." When a rep comes in on Monday, they should have a plan for what they're going to do with their agents that week. When they leave on Friday, they should know what the agents will be doing over the weekend.
That mindset shift is actually the hardest part of onboarding. But it's also where the biggest payoff is.
