Calling Bull$hit on IBM's AI Impact Claims

IBM's CEO Arvind Krishna Needs to Have a Talk with Anthropic's CEO Dario Amodei

tl;dr

• IBM’s CEO says AI replaced “a couple hundred” jobs, but created more jobs; our data doesn’t support this claim.

• Anthropic’s CEO thinks half of white-collar jobs are on the chopping block, and our data backs him.

• If you're guessing about AI's impact, you're already behind. We're measuring it. In real time.

• Meet the Live Data AI Replacement Index (AIR), a framework to measure the impact of AI on any company’s headcount.

Introduction

In May, two statements about AI’s impact on the workforce caught our attention. The first was a yawn from IBM CEO Arvind Krishna, who said AI has replaced only "a couple hundred" jobs at IBM. The second was a gasp from Anthropic CEO Dario Amodei, who predicted that AI could replace up to 50% of white-collar jobs within two years. Krishna, presumably, spoke from IBM’s experience with AI substitution. Amodei was speculating, albeit from a very informed position.

Can the two perspectives be reconciled?

Here’s Live Data’s attempt to do just that.

Calling Bull$hit on IBM’s Claim

In the same announcement where Krishna said only 200 jobs had been eliminated by AI, he flexed that IBM’s consulting division had earned $6 billion helping clients implement AI. That doesn’t add up. A company generating $6 billion from AI transformation engagements can only identify 200 internal roles affected by AI? Really?

We set out to challenge that claim, not with opinion, but with data.

Introducing the Live Data AI Replacement Index (AIR)

Our AIR Index was built to quantify how vulnerable a company’s headcount is AI-driven substitution. We launched it with IBM as the test case to evaluate whether the true risk extended far beyond the couple hundred roles their CEO referenced.

At the core of this analysis is Live Data’s workforce intelligence platform. We track employment data for 95M professionals in near real time, including over 100,000 IBMers, across multiple roles, departments, and seniority levels. That gives us a unique and structured view into how IBM’s workforce is shifting based on observable events, not speculation.

From our IBM panel, we identified 38,000 unique roles, intentionally going beyond job titles to account for the actual functions and tasks performed by each role within each department. This distinction is important as a software engineer handling internal IT operations (maintenance) faces greater AI-displacement risk than one creating core product features (innovation). Treating these roles identically simply because they share a title would obscure meaningful differences that would affect AI displacement.

This role-level resolution is what powers the AIR Index. We assess each position based on what it does, not what it's called. We then score roles using three key metrics:

• Salary Band: Informed by public compensation data from Pave, Radford, Levels.fyi, Carta, and others.

• AI Exposure Score: Based on task-level analysis from O*NET, job postings, and skillsets generally associated with certain roles.

• AI Cost-Savings Score: A composite measure estimating financial savings from AI replacement.

And the results?

IBM Jobs Are Disappearing Into Thin Air

IBM’s CEO recently stated that AI has replaced only a few hundred jobs and is creating new ones in programming and sales. Our data shows no evidence of either. IBM’s headcount continues to decline, and even under our most conservative scenario (AI-Assisted), we anticipate an increasing rate of headcount reductions over the next three years due to AI adoption.

To estimate the impact, we extrapolated from insights drawn from our panel of 100,000 IBMers, incorporating role types, departments, and tenure patterns to model AI susceptibility across IBM’s 260,000-person workforce. In the AI-Assisted scenario, more than 51,000 roles—roughly 20 percent—are impacted. In the more aggressive AI-Led scenario, that number exceeds 90,000.

One surprise: the seemingly low displacement risk in Sales and Support. We expected higher exposure, but high-compensation, client-facing sellers masked the department’s overall vulnerability. When we isolated frontline support roles, the susceptibility looked similar to what we see at peer tech companies.

Show Me the Money

The financial impact? Under the conservative AI-Assisted scenario and a 3 year rollout period, we estimate $5.8 billion in potential labor cost savings, which amounted to about 11% of IBM’s 2024 operating expenses.

Of course the story doesn’t end with cost savings. Our analysis doesn’t factor in the upside IBM could generate by helping other enterprises replicate its internal AI efficiencies. That’s the quiet multiplier. The very dynamic shrinking IBM’s expense base may also fuel its future revenue growth.

What Comes Next

Where in that phase where everyone knows AI is a disruptive force, and everyone’s trying to guess what it means for jobs, companies, and the overall economy. The speculation can be interesting, but we believe the impact of the coming-much-sooner-than-you-think Great AI Displacement can be measured.

But to be able to do that, you have to see what Live Data sees.

You have to be able to see the dynamics of a company’s workforce; the arrivals, departures, the roles that are expanding, and those being phased out. And you have to see those shifts as they happen, not quarters later. Too much of today’s commentary is based on imagined futures, shaped by hunches or bias. Ours isn’t. We’re watching the data, every day, and the data indicates companies are being reshaped in a manner and at a rate faster than ever before.

We started with IBM because its CEO made a throwaway claim that made no sense to us. But this isn’t just about IBM. This is about a framework. The AIR Index is designed to objectively evaluate any company’s exposure to AI-driven workforce change, role by role, department by department.

IBM is just the first callout. Others are coming.

Watch this space.