Don’t Do What Salesforce Does
Human-in-the-loop Customer Support is a Competitive Advantage for Startup and Growth Companies
Salesforce told the world they cut roughly 4,000 customer-support roles because AI agents can do the work. That headline set us off—not because AI can’t handle a chunk of Tier-1, but because this is being sold as a simple, virtuous efficiency story.
It isn’t.
It’s a mega-platform flexing advantages that most companies don’t have.
Why Salesforce can “get away with it”
At Salesforce’s scale, three forces cushion the blow:
Switching costs + inertia. When you’re the system of record, customers tolerate more automation and longer time-to-human, because ripping you out is expensive and political.
Distribution > delight (for a while). With entrenched enterprise relationships, you can let support quality slide and still hit numbers—until a real challenger shows up with a meaningfully better experience.
Narrative cover. Tying layoffs to “AI breakthroughs” plays well on earnings calls and helps juice up sagging revenue. It reframes a blunt cost / hype move as a visionary transformation.
We’re not saying their tech isn’t real. Agentforce can triage, deflect, and resolve known intents at scale. But holding up a platform outlier as an operating model for everyone else is how companies amputate their learning loop.Why other companies absolutely cannot
If you’re still building and flying the rocketship, support is R&D. We’ve said this in Building Rocketships: the fastest way to hear the truth about your product is to sit inside the messy parts of customer conversations—exceptions, ambiguity, and “this broke my workflow” moments. That’s where loyalty forms and roadmaps sharpen.
When an early-stage company replaces human judgment with wall-to-wall automation, three things happen:
Signal loss. Bots close tickets; humans escalate patterns. Kill the escalation, kill the product truth.
Brand erosion. In a world of synthetic everything, “a competent human who actually owns the problem” is a differentiator.
Bench collapse. Support is a talent farm for PMs and operators. Flatten the entry layer and your org gets slower and thinner next year.
None of that shows up in this quarter’s presentation. All of it shows up in next year’s retention and roadmap.
We’re AI maximalists for the right work - high-leverage moments - novel, cross-functional, politically sensitive, high-dollar, or architecture-shaping problems - where judgment and ownership matter more than speed.
Great support isn’t answering quickly; it’s owning outcomes. AI should bend the cost curve so your best people spend more time on the hairy stuff, not less.Our recommendation
Don’t slavishly follow Wall Street’s 90-day drumbeat. Quarterly optics are not a strategy. If you are growing, AI should expand capacity and let humans handle higher-value work. If you are slowing, tell the truth about demand and fix the product—not hide behind an “AI efficiency” press release.
We believe in AI deeply—and in people even more. We’ll automate the obvious, guard the exception lane, and keep humans as close as possible to customers and product. That’s how we build the kind of company we talk about on Building Rocketships: one that listens hard, learns fast, and earns the right to keep serving.
🎥 Check out our latest podcast on the ProductMind YouTube channel.
🎵 We are excited to announce we have expanded our podcast 🎙️to Spotify. Please give us a listen and if you like what you hear share with a friend, follow us, and (or) rate us 5 stars. ⭐️⭐️⭐️⭐️⭐️
Want to dive deeper?
Check out our book BUILDING ROCKETSHIPS 🚀 and continue this and other conversations in our 💬 ProductMind Slack community and our LinkedIn community.
Loved your points on signal loss and brand erosion