Companies Claim AI Will 'Do More With Less'—But We Haven't Deployed AI at Scale. Are We Laying Off Humans for Tech That Doesn't Exist Yet?

General
1

Companies Claim AI Will ‘Do More With Less’—But We Haven’t Deployed AI at Scale. Are We Laying Off Humans for Tech That Doesn’t Exist Yet?

I just came out of a board meeting where one of our VCs asked: “If you’re deploying AI tools, why haven’t you cut headcount yet? Your competitors are seeing 20-30% productivity gains.”

Here’s what I didn’t say out loud: We haven’t actually deployed AI at scale. And neither have most of them.

The Numbers Tell a Different Story

Q1 2026 saw 52,050 tech layoffs—a 40% jump from last year. 20.4% of these layoffs were explicitly attributed to AI, up from less than 8% in 2025. In March alone, AI led all reasons for job cuts with 15,341 announced.

Oracle cut 20,000-30,000 jobs. Block cut 4,000—nearly 40% of their workforce. The narrative? “AI productivity tools.”

But here’s the thing: Harvard Business Review found companies are laying off workers because of AI’s potential, not its performance.

My Reality Check

At my company (120-person engineering org), here’s our actual AI deployment:

  • Customer Support: AI handles ~30% of tier-1 tickets. Human agents still needed for anything complex.
  • QA/Testing: ~25% efficiency gains on test case generation. Still requires human review and validation.
  • Code Generation: ~35% of engineers actively use AI coding assistants. Time savings vary wildly (10-40% depending on task).

Our board wants us to cut 15-20% of headcount “to capture AI productivity gains.”

But we haven’t proven we can maintain service levels with fewer people. We’re being asked to lay off humans based on projected future capability, not current deployment reality.

The Uncomfortable Questions

I keep coming back to these:

  1. What happens when we cut too deep? If we lay off 20% based on “AI will handle this” and AI doesn’t handle it—what then? Re-hire in 6 months at higher cost? Apologize to customers for degraded service?

  2. Who’s accountable for the gap? When companies like Oracle shed 20K-30K people to fund AI data centers, who’s responsible if that AI can’t actually replace the work those people did?

  3. Are we creating AI-driven displacement before AI works? The World Economic Forum notes that 2026 is “the year of AI reckoning”—pragmatism supplanting optimism. Yet layoffs are happening now, while deployment lags 6-18 months behind projections.

  4. Is this financial engineering disguised as AI productivity? Cut costs to show margin improvement. Blame AI. Get board approval. Figure out actual AI deployment later?

The Leadership Dilemma

I’m caught between:

  • Board pressure: “Industry standard is 20-30% productivity gains from AI”
  • Engineering reality: We’re at 10-15% productivity gains in some areas, zero in others
  • Team morale: Engineers see the headlines. “Am I about to be replaced by AI?” kills motivation.

The honest answer is: I don’t know yet whether AI will let us “do more with less.” We’re still figuring out what it can actually do reliably at scale.

But we’re being pressured to make permanent workforce decisions based on temporary optimism.

What I’m Seeing in the Industry

Legacy companies (Oracle, Dell) are shedding traditional roles to fund AI infrastructure. AI-native companies (Anthropic, OpenAI) are still hiring aggressively. That tells me something: the companies closest to AI technology aren’t betting on workforce reduction yet.

Meanwhile, CFOs are admitting privately that AI layoffs will be 9x higher this year—but also acknowledge returns on investment remain mixed.

The Question for This Community

How are you navigating this?

  • For CTOs/VPs: Are you facing similar board pressure? What’s your actual AI deployment vs the cuts you’re being asked to make?
  • For engineering directors/managers: What’s the reality on the ground? Are AI tools delivering the 20-30% gains the headlines promise?
  • For product leaders: How are you measuring AI impact on customer experience and business outcomes?

I’m trying to figure out: Are we making sound workforce decisions based on proven capability, or are we laying off humans for technology that doesn’t exist yet?

Because if it’s the latter, we’re going to have a reckoning in 12-18 months when the promised productivity gains don’t materialize and we realize we cut too deep.

Would love to hear how others are thinking through this.

2

This hits close to home. I’m at a Fortune 500 financial services company, and we’re getting the exact same pressure from leadership.

Our Deployment Reality

Here’s what AI is actually doing in our 40+ engineer org:

Code Review Assistance

  • ~40% of engineers actively use AI coding assistants
  • Time savings: 15-20% on routine tasks (CRUD operations, test generation, boilerplate)
  • But: 30% more time spent by senior engineers reviewing AI-generated code for security and compliance issues
  • Net productivity gain: Maybe 10%, not 20-30%

Data Processing Scripts

  • 25% productivity gains on ETL pipeline development
  • Solid win, but this represents ~15% of our total engineering work

Customer Service

  • ~30% of tier-1 support tickets now automated
  • Great for simple account balance queries, password resets
  • But: Anything involving money movement or disputes still requires human judgment

Core Financial Logic

  • 0% AI involvement
  • Compliance won’t allow it. Every line of financial calculation code must be human-written and audited.
  • This is 50% of our engineering work.

The Math Doesn’t Add Up

So let’s do the math:

  • 10% gain on 40% of engineering work = 4% overall productivity
  • 25% gain on 15% of data work = 3.75% overall productivity
  • 30% gain on support (non-engineering) = doesn’t reduce eng headcount
  • 0% gain on 50% of core financial work = 0%

Total realistic productivity gain: ~8%

Yet we’re being told to plan for 20% headcount reduction.

The Cultural Damage

Here’s what worries me more than the math: our team morale collapsed the day the VP announced “AI-driven efficiency improvements.”

Three of my best engineers immediately started interviewing. Two have told me directly: “I’m not waiting around to be replaced by a chatbot.”

The irony? We need these senior engineers more now because they’re the ones who can actually validate AI output and catch the subtle bugs AI introduces.

Who’s Responsible When We Cut Too Deep?

@cto_michelle your question about accountability really resonates.

In financial services, when we get compliance wrong, there are regulators, fines, and personal liability. But when we lay off 20% based on projected AI capability that doesn’t materialize—who’s accountable?

  • Not the consultants who sold the “20-30% AI productivity” story
  • Not the board who demanded cuts
  • Not the AI vendors who oversold capability

It’ll be us—the engineering leaders—scrambling to re-hire at 1.5x the cost while apologizing to customers for degraded service.

I’m pushing back on our timeline. Asking for 6-month pilots in specific areas before making permanent workforce decisions. But the pressure is real.

How are others defending against premature AI-driven cuts?

3

Adding another uncomfortable dimension to this conversation: AI-driven layoffs are disproportionately hitting entry-level roles and our most diverse teams.

The Pattern I’m Seeing

At our EdTech startup, we’ve been tracking which roles are getting targeted in “AI efficiency” conversations:

Roles with High AI Attribution for Cuts

  • Customer support (60% women, 45% people of color at our company)
  • Content creation and QA (55% women)
  • Data entry and operations (most diverse teams)

Roles Largely Untouched

  • Engineering (75% male, 65% white/Asian)
  • Product (80% male)
  • Executive leadership (90% male, 95% white)

This isn’t unique to us. The roles being automated industry-wide—customer support, content moderation, data processing—are historically more diverse than technical roles.

We’re Destroying the Pipeline

Here’s what really concerns me: we’re cutting the entry-level roles that create career pipelines.

Our customer success team is the traditional entry point for people without CS degrees or elite university backgrounds. It’s where people learn our business, prove themselves, and move into product, operations, or technical roles.

If we eliminate those entry points because “AI handles tier-1 support now,” where do those career paths go?

The Capability Gap

And here’s the kicker: the AI can’t actually do what we’re claiming it can.

Our AI chatbot handles 40% of tier-1 queries. Sounds great. But:

  • Tier-2 and tier-3 require judgment, empathy, and understanding of edge cases
  • 25% of “AI-resolved” tickets get re-opened because the AI missed context
  • Our most complex customer issues—the ones that prevent churn—require deep relationship knowledge

We tried piloting “AI-only support” for two weeks. Customer satisfaction dropped 30 points. We quietly rolled it back.

Yet the pressure to cut the human support team continues.

An Uncomfortable Question

@cto_michelle you asked who’s accountable. I want to add: If we cut diverse roles first using AI as the rationale, are we just repeating the “offshore will replace you” pattern from 20 years ago?

Back then, it was “we’ll offshore support and QA to cut costs.” Many of those jobs never came back. The people in those roles—disproportionately women and people of color—lost career paths into tech.

Now it’s “AI will replace support and QA.” Same roles. Same communities. Different technology excuse.

What We’re Doing Differently

I’m pushing our leadership to think about AI augmentation, not replacement:

  • Measure augmentation metrics: How much faster do humans work with AI? (Answer: 25-40% in some areas)
  • Redeploy, don’t eliminate: Customer support people with AI tools handling tier-2/3 work can cover 30% more customers
  • Preserve career pipelines: Entry-level roles with AI support become training grounds for future leaders

We haven’t cut anyone yet. We’re experimenting with augmentation first, measuring actual impact, then deciding.

But I’ll be honest: I’m fighting this battle every board meeting. The pressure to “capture AI productivity through headcount reduction” is intense.

How are others thinking about the equity dimension of AI-driven workforce decisions?

4

From a product and business perspective, the math here is even more troubling than you’re describing.

The Business Case Doesn’t Exist

I just went through this exercise at our Series B startup. Board wanted us to project “AI-driven cost savings” for the next funding round.

Here’s what we found when we actually measured:

Engineering: Real but Modest

  • 25% faster on specific implementation tasks (CRUD, tests, boilerplate)
  • But 0% faster on: architecture decisions, debugging complex issues, customer problem-solving, technical strategy
  • Net: Maybe 10-15% productivity gain, mostly on routine work

Product: Negligible

  • AI doesn’t help with: customer discovery, strategic prioritization, roadmap decisions, stakeholder management
  • AI can help with: writing PRDs, creating mockups, analyzing usage data
  • Net: 5% productivity gain on documentation, zero on decisions that matter

Sales: Negative ROI

  • We tried an AI SDR (sales development rep) for outbound
  • 8% conversion rate vs 28% for human SDRs
  • Why? AI can’t read the room, adjust pitch, or build genuine relationships
  • We shut it down after 6 weeks

Customer Success: Mixed

  • AI handles 35% of tier-1 support queries
  • But our Net Promoter Score dropped 15 points
  • Customers feel like we’re hiding behind chatbots
  • The math: we save $80K in support costs but risk $400K in churn

The Timing Problem

Here’s what’s really concerning: Our board is pushing us to cut based on where our AI deployment should be, not where it actually is.

Timeline they assumed:

  • Q1 2026: Deploy AI coding assistants (done ✓)
  • Q2 2026: Achieve 20% engineering productivity gains (we’re at 12%)
  • Q3 2026: Deploy AI customer support (we’re at 35% coverage, not 80%)
  • Q4 2026: Reduce headcount by 15-20% while maintaining velocity

We’re 6-12 months behind this timeline. But the pressure to cut is happening now, based on the optimistic projection.

Financial Engineering Disguised as AI Productivity?

@cto_michelle your question about financial engineering really hits. I’ve seen this playbook before:

  1. Market demands improved margins
  2. Company commits to cost reduction targets
  3. “AI productivity” becomes the narrative justification
  4. Cuts happen before AI is actually deployed or proven
  5. 12-18 months later: scramble to backfill because AI couldn’t cover the gap
  6. But the quarterly earnings already benefited from the cost reduction

In other words: cut costs now, show margin improvement now, deal with the consequences later.

The AI narrative gives cover for decisions that would otherwise be seen as short-sighted cost-cutting.

What Happens When We Cut Too Deep?

We ran scenarios on what happens if we cut 20% based on “future AI productivity”:

Optimistic Scenario (AI delivers as promised)

  • Maintain velocity with 20% fewer people
  • $2M annual savings
  • Slight hit to customer experience (manageable)

Realistic Scenario (AI delivers 50% of promise)

  • Velocity drops 30%
  • Product roadmap slips 2 quarters
  • Customer experience degrades significantly
  • Backfill 10% of roles at 1.5x cost
  • Net: $500K savings but significant opportunity cost

Pessimistic Scenario (AI delivers 25% of promise)

  • Critical initiatives fail
  • Customer churn accelerates
  • Emergency rehiring at 2x cost
  • Net: Loss of $1M+ and 6-12 month setback

The board hears “optimistic.” Reality is usually between realistic and pessimistic.

The Honest Question

Are most companies in the same boat—6-18 months behind their AI deployment timelines but cutting headcount based on the original projection?

Because if that’s the pattern, we’re going to see a wave of “oops, we cut too deep” realizations in late 2026 / early 2027.

And by then, the damage to teams, morale, and customer relationships will be done.

I’m arguing for a “prove it first” approach: measure actual AI impact for 2 quarters before making permanent workforce decisions. But I’m not sure I’m winning that argument.

How are other product leaders measuring the actual business impact of AI vs the projected impact?