GTM Strategy: 5 Insights from 500 B2B SaaS Orgs (Jeremey Donovan, EVP Sales + CS @ Insight Partners)
AI’s Reality Check: Why Disciplined Execution Still Beats Tooling
Everyone’s chasing AI tools. The top-performing CROs are chasing something else entirely.
Jeremy Donovan has spent four years advising 500 B2B SaaS companies at Insight Partners. He just surveyed 150 CROs on what’s actually working with AI in sales. The answer surprised me: it has almost nothing to do with which tools you’re using.
The differentiator is disciplined execution. Deal reviews. Pipeline rigor. ICP focus. The boring stuff.
Jeremy’s background makes him worth listening to on this. He started as a semiconductor engineer, got a master’s in data science from UVA, an MBA from Booth, picked up a CFA. Then he went operator side: nearly four years at SalesLoft running solutions engineering, rev ops, and revenue strategy. Before that, almost two decades at Gartner and GLG.
Engineer’s rigor. Operator’s scars. Investor’s pattern recognition. He’s watching AI play out across hundreds of companies in real time, and he’s got the data to back up his takes.
Here’s what the data says: productivity gains are running 5-15%. Modest. Full automation keeps failing. But augmentation is working quietly in companies that were already executing well. AI amplifies what’s there. It doesn’t create discipline from nothing.
We got into it all. Where AI actually delivers. Why the SDR role is getting absorbed, not eliminated. What “build vs. buy” looks like when vibe coding changed the math. And why 2026 might be the year this moves from bottoms-up experiments to top-down systems transformation.
Jeremy also dropped what I think is the most important insight of the conversation: we’ve been optimizing existing processes with AI. The real unlock is rethinking the processes entirely. Unbundle the job. Figure out what’s an AI task and what’s a human task. Then rebuild.
That’s a bigger swing. And nobody’s really doing it yet.
Disciplined Execution Beats AI Tooling
I asked Jeremy what separates companies getting value from AI versus the ones spinning their wheels. His answer was blunt: there’s no silver bullet.
Top performers and average performers are adopting the same use cases in roughly the same order. The difference? The top performers were already disciplined before AI showed up.
“If I were to go back into an operating role and could only run one play, it would be incredibly disciplined weekly deal reviews.”
He’s not wrong. When I talk to struggling sales orgs, the pattern is obvious. They want the new tool to fix a process problem. But tools don’t fix process. They expose it.
Jeremy does a lot of CRO interviewing for Insight’s portfolio companies. The number one thing he screens for: operating rhythm. How disciplined is this person? Do they stick to it? He doesn’t just ask the candidate. He back-channels their former teams to find out what it was actually like to work for them.
This tracks with research on expertise. K. Anders Ericsson’s work on deliberate practice shows that feedback-rich environments with tight iteration loops produce mastery faster than anything else. The CROs who run tight deal reviews, who hold reps accountable to pipeline generation time, who actually inspect what they expect? They’re building that feedback loop into the system.
AI accelerates the loop. But you need the loop first.
One CRO Jeremy interviewed writes his own prompts. You could argue that’s not the best use of a CRO’s time at a bigger company. But at his scale, it means he deeply understands the system. He can train his team because he’s trained himself. That kind of hands-on ownership compounds.
The takeaway for me: stop asking “which AI tool should I buy?” Start asking “how tight is my operating rhythm?” If your deal reviews are sloppy, if your pipeline inspections are sporadic, if your reps don’t know whether they’re hitting their activity targets until month-end, AI won’t save you. It’ll just make the mess visible faster.
2026 Is the Year of Top-Down AI
For the past two years, AI adoption in sales has been mostly bottoms-up. Reps tinkering with ChatGPT. Managers building prompt libraries. RevOps experimenting with enrichment tools.
Jeremy thinks that era is ending.
“There’s only so much you can do in the foundation LLMs,” he told me. The real value requires orchestration across systems. Snowflake. Salesforce. Your data warehouse. Your security layer. And the moment you need that kind of access, a random AE can’t just point an agent at those systems and start mining data.
Security gets involved. Governance gets involved. IT gets involved.
The companies moving fastest have appointed what Jeremy calls an “AI czar.” Someone with organizational influence who owns internal AI optimization. These people are coming out of RevOps, PMOs, finance. They showed passion early, and now they’re project-managing the complex integrations that bottoms-up tinkering can’t touch.
This matches what I’ve seen at Owner. We made aggressive early investments in AI across GTM, but the work is being run out of our data team. We now have applied AI folks embedded in that org. Because the real leverage comes from combining Salesforce data with Snowflake with external research with internal playbooks. You can’t do that in a chat window with limited MCP access. You need infrastructure.
Gartner’s hype cycle framework is useful here. Early 2025 was peak inflated expectations. Everyone was promising astronomical improvements in win rates. By late 2025, we landed in the trough of disillusionment. LinkedIn was full of “AI isn’t working in sales yet” takes.
The climb out of that trough requires systems-level thinking. Top-down investment. And probably a lot less hype about individual tools.
For CROs, this means: if you don’t have someone who owns AI orchestration internally, you’re already behind. Find that person. Give them authority. Let them partner with security and IT rather than fighting them. The bottoms-up era gave us useful experiments. The top-down era is where the real productivity gains live.
The Build vs. Buy Pendulum Is Swinging
Jeremy dropped a stat that surprised me: roughly 50% of companies are now building their own RFP response tools instead of buying off-the-shelf solutions.
A year ago, that would have been crazy. RFP software was a mature category. Vendors had spent years building integrations, compliance features, knowledge bases.
What changed? Vibe coding.
“It’s just so freakishly fast,” Jeremy said. He told me a story about needing to convert a calculator from JavaScript to a Microsoft Power App. His engineering team quoted a month with outsourced developers. Jeremy opened Cursor in agentic mode, described what he needed, and had a working app in 60 seconds.
The math on build vs. buy has shifted. When you can prototype something in an afternoon that used to take a quarter, the calculus changes. You build exactly what you need. Less friction. Better fit.
But Jeremy was careful to note the other side. Historically, most internal builds stop working within a year. Nobody maintains them. The world moves on. We’re going to see a lot of shelfware from this era of rapid building.
The smart framework here: prototype internally, validate the use case, then decide whether to productize or buy. Jeremy did this at SalesLoft. He’d build something scrappy, run it for six months, and if it stuck, he’d go find a vendor who could do it better at scale.
There’s also a complexity threshold to watch. Simple use cases favor building. But when you need predictive ML combined with generative AI combined with workflow automation combined with security compliance? That’s where platforms start winning again.
Jeremy gave a good example from customer success. Predicting churn is a traditional ML problem. But once you identify an at-risk account, you need to generate a workflow, draft messages, maybe trigger an escalation. That heterogeneity of AI types gets complicated fast.
My take: build for speed and learning, but don’t fall in love with your prototypes. The companies trying to run production workloads on duct-taped internal tools are going to hit a wall. Know when to graduate to real infrastructure.
AI Makes Humans More Valuable, Not Less
Here’s a stat that should reframe how you think about AI in sales: 70% of outbound-sourced opportunities are still booked via phone.
Not email. Not LinkedIn. Not AI-generated sequences. Phone.
Jeremy’s been in the SDR game for 15+ years. He’s watched cold outbound go from shockingly effective to genuinely difficult. Response rates cratered. Touches per opportunity went from 200-400 five years ago to 1,000-1,400 today. Email is essentially destroyed as a channel for cold outreach.
And yet. The phone endures.
Why? Robert Cialdini’s work on reciprocity gives us the answer. We respond when another human puts genuine effort into engaging us. The reverse is also true. If you phone it in with obvious AI-generated outreach, people don’t just ignore you. They flag you as spam and hope your emails never arrive again.
This is why Jeremy sees augmentation working where automation fails. The winning formula is human + AI, not AI replacing humans.
The SDR role isn’t dying. It’s shape-shifting. Jeremy expects outbound SDR responsibilities to get absorbed back into AEs, especially as AI handles research and personalization. Inbound SDRs are disappearing faster because routing and qualification can be automated when prospects have already opted in.
But here’s the part that gave me pause: one CRO Jeremy works with ripped out an AI tool that auto-extracted MEDDIC fields from call transcripts. The tool worked perfectly. The problem? “The AEs were getting stupid.” They stopped thinking critically about their deals because the AI was doing the synthesis for them.
She removed the tool to protect her team’s judgment.
That story is a warning. AI that removes cognitive load can also remove cognitive development. The best implementations keep humans in the loop not because the AI can’t do the task, but because the human needs to stay sharp.
The Unlock Is Rethinking Roles, Not Optimizing Tasks
This was the insight that stuck with me most.
Jeremy put it simply: “So far, we’ve tried to optimize the jobs to be done within the existing role structure and within existing processes. The unlock is to say, maybe we don’t have the roles right.”
We’ve been using AI to make the current system faster. Write emails quicker. Research prospects in less time. Summarize calls automatically. Those are real gains. But they’re incremental.
The bigger swing is unbundling roles entirely.
I talked about this with Jordan Crawford on a recent episode. His frame: look at a role as a basket of tasks. Unbundle it completely. Which tasks are AI tasks? Which tasks are human tasks? Once you’ve sorted that across your SDR, AE, and SE roles, you might find the optimal grouping looks nothing like what you have today.
Jeremy sees this happening with sales engineers already. The low-end SE work is getting absorbed into AEs as AI handles more technical Q&A in real-time. The high-end is evolving toward forward-deployed engineers who embed deeply with customers post-sale. The middle is getting squeezed.
This requires real creativity. You’re not optimizing a process. You’re redesigning how work gets done. Most companies aren’t there yet. They’re still in the “add AI to existing workflows” phase.
But the productivity gains everyone expected from AI? They’re probably hiding on the other side of that redesign. The 5-15% lift CROs report today is the optimizing-existing-tasks version. The 30%+ lift requires rethinking the tasks themselves.
The Clock Is Running
Jeremy and I covered a lot of ground. The through-line: AI rewards discipline and punishes sloppiness.
If your operating rhythm is tight, your deal reviews rigorous, your pipeline inspection consistent, AI will amplify that. You’ll move faster. Your reps will be better prepared. Your forecasts will be more accurate.
If your operating rhythm is loose, AI will expose every crack. The tools will sit unused. The data will be garbage. The productivity gains will never materialize.
The question for every CRO is simple: are you building the infrastructure for AI to accelerate, or are you hoping AI will fix problems you haven’t solved?
One path leads to compounding advantage. The other leads to expensive shelfware and frustrated teams.
And the companies that figure out how to truly rethink their roles and processes, not just optimize existing ones? They’re going to create a gap that’s very hard to close.
The clock is already running.
If this episode gave you something to think about, I’d appreciate a rating or review on Apple Podcasts or Spotify. It helps other revenue leaders find the show!