We’re past the hype cycle on AI replacement. After working with 40+ B2B companies over the past 18 months and deploying 20+ AI agents across our own operations at SaaStr, I can tell you exactly where AI actually works today—and where it doesn’t.
Here’s the honest breakdown:
Outbound Email-Based SDRs: 90%+ Replaceable (But It’s Not Easy)
You can replace most—possibly 100%—of your outbound SDRs with AI today. But here’s the catch everyone misses: you still need humans managing it and training it. Every. Single. Day.
The companies getting this right aren’t just “turning on AI.” They’re going deep. They’re training platforms daily. They’re orchestrating responses in real-time. If you truly commit to this—if you have someone spending 2-3 hours per day refining prompts, analyzing responses, and adjusting targeting—you can replace the vast majority of classic email-based SDR work with AI today.
But you absolutely cannot just push a button. The companies that try the “set it and forget it” approach? They’re the ones complaining that “AI doesn’t work for sales.”
And AI certainly can’t replace any in-person cold calling.
Inbound BDRs: 95%+ Replaceable (Right Now)
This one’s actually easier than outbound. A well-trained AI should be able to qualify an inbound lead faster and better than 95% of human BDRs can today.
Why? Because qualification is pattern matching. It’s asking the right questions, recognizing signals, and routing appropriately. AI excels at this. It never has a bad day. It never forgets to ask about budget. It responds in seconds, not hours.
The companies still using human BDRs for first-touch qualification are leaving money on the table. Full stop.
Customer Support: 50% Replaceable (With Investment)
You can replace 50% of your support team with AI today—IF you invest. And you can deflect even more issues if the AI agent is properly trained.
But again, this only works if you go deep. You need to train the platform continuously. You need daily orchestration. And most importantly, you need a very responsive human-in-the-loop escalation strategy.
The magic number I’m seeing? Companies that get this right are handling 50-60% of tier 1 support tickets entirely with AI, escalating the complex stuff to humans who can actually solve problems. The result? Better response times, happier customers, and lower support costs.
But it requires real investment. You can’t half-ass this.
Marketing Managers: You Probably Don’t Need Half of Them
This is the one that makes people uncomfortable, but it’s true. Claude and ChatGPT—as they exist today, no special tools required—can handle a ton of marketing operations. Visual tools can create incredible collateral. Gamma can arm your sales team with better materials than most marketing teams were producing 18 months ago.
But here’s what hasn’t changed: you still need orchestration. A lot of it. The companies cutting 50% of their marketing team aren’t just “using AI.” They have someone—usually a very talented marketer—orchestrating multiple AI tools, reviewing outputs, and maintaining quality standards.
You don’t need five people doing campaign management anymore. But you need one really good person managing AI that does the work of five.
Customer Success: 60%+ Not Needed (But We Need Better Platforms)
This is where it gets controversial. CSMs that don’t solve problems—that just do QBRs without adding true value—are not necessary in the age of AI. You can probably move on from them today.
We need better AI platforms for CS, but even with current tools, you don’t need 60%+ of the CSMs you have today. The ones who are just “checking in”? The ones whose main value is “relationship”? That’s not enough anymore.
The CSMs who will survive—and thrive—are the ones who solve real problems, drive product adoption, and identify expansion opportunities. Everyone else? At risk. At the margin, put the money into FDEs here. Get customers successfully onboarded, especially with AI Agents that require training.
Account Executives: Still Need 70% (For Now)
You still need most of your AEs. This should change in the next 24 months for inbound sales, but we’re not there yet.
For field sales and in-person selling? Who knows. The human element still matters tremendously for complex enterprise deals. The relationship still matters. The ability to read a room, adjust on the fly, navigate complex political dynamics—AI isn’t there yet.
Give it two years for inbound. Give it five to ten years for field. Maybe longer.
Engineering: Zero Net Headcount Reduction (But 20-40% Productivity Boost)
Here’s the surprise: I’m not seeing AI let companies reduce net engineering headcount at all.
Instead, what I’m seeing is a 20-40% productivity boost across the board. And far more—sometimes 10x—for prototypes and proof-of-concept production.
But here’s what’s happening: it becomes an arms race. Everyone is more productive. Software is shipping faster. Features that took months now take weeks. So you need more great engineers than ever to keep up with competitors who are also moving faster.
It’s a cracked arms race in product development today. The winners aren’t the ones cutting engineering teams. They’re the ones doubling down on great engineers armed with AI tools.
The Real Lesson: AI Requires Sustained Commitment To Work. Upfront and Ongoing.
The pattern across all of this? AI doesn’t replace humans by magic. It replaces humans when companies commit to:
- Daily orchestration. Someone needs to be managing, training, and refining AI systems every single day.
- Deep integration. You can’t just bolt on an AI tool and expect results. You need to rebuild workflows around AI capabilities.
- Human-in-the-loop escalation. The best AI implementations know exactly when to hand off to humans—and do it seamlessly.
- Continuous training. AI systems degrade without constant attention. The companies winning with AI treat it like a junior employee that needs daily coaching.
The companies that “try AI” and declare it doesn’t work? They’re the ones that aren’t doing any of the above.
The companies quietly replacing 50%+ of certain roles with AI? They’re the ones treating it like the serious operational initiative it is—with dedicated resources, daily attention, and real investment.
That’s the difference between AI hype and AI results.
