AT&T is a useful company to watch right now. It is large, old, heavily unionized, operating in a regulated industry, and sitting at the center of everything AI needs to function: the pipes that move data. Its CEO, John Stankey, has been inside the company for over 40 years. That makes what he is doing, and saying, worth paying close attention to.
The headline numbers are uncomfortable. AT&T eliminated about 8,000 jobs in 2025, finishing the year with 133,000 employees. In 2017, the company had 280,000. That is a loss of roughly 140,000 jobs over eight years, and the pace has not slowed. Stankey is direct about the cause: technology has made running a communications network dramatically less labor-intensive over time, and AI is the next step in that same process. He frames it honestly rather than defensively.
Internally, AT&T has deployed an AI tool called Ask AT&T across more than 100,000 employees. It handles customer service queries, network diagnostics, software development, legal research, and HR questions. The numbers from the Microsoft partnership that built the system are specific: customer care resolution time is down 33%, and the company has deployed over 70 distinct AI applications. Stankey says software developers are 30% more productive, but the team is the same size; they are finishing more projects, not cutting headcount. That is an important distinction. Efficiency gains are being plowed back into growth, not extracted as cost savings.
The more practical insight for any business operator is Stankey's separation of strategy and culture. Strategy, he says, is relatively straightforward: communicate the rationale clearly, set measurable goals, align pay and incentives to those goals. Culture is far harder because it is not a single thing you can change with a memo. It has to show up in every hiring decision, every promotion, every performance conversation, every day.
At AT&T, the old culture was engineering-led: smart people in a lab decided what the technology could do, then offered it to customers. That worked for decades. It stopped working when the internet arrived and customers could simply walk away. Stankey's replacement model is what he calls market-based: start with what the customer needs, then figure out how to deliver it. That sounds obvious, but making it real inside a 150-year-old institution means changing what gets rewarded, not just what gets said.
On AI and jobs, Stankey pushes back on two narratives that he thinks are too simple. The first is that AI only destroys entry-level work. His observation from AT&T's legal department is that new graduates using AI tools are actually more productive on day one than senior staff who had to adapt to them. The skills gap runs in the opposite direction to what most people assume. The second narrative he challenges is that AI will simply erase job categories wholesale in the short term. His view is that definitions of work change, but the pace of that change is uneven and often misread.
The broader data supports a mixed picture. A 2026 PwC analysis of over a billion job postings found that companies most exposed to AI are actually growing headcount and wages faster than companies least exposed. But that split is becoming starker: the gap between companies that have moved AI from pilot to production and those still experimenting is now measurable in revenue and margins, not just strategy documents.
What AT&T's story offers any senior operator is a concrete model. Stankey's company spent $22 billion on its network in 2025, plans to spend $23 to 24 billion annually through 2028, and is simultaneously running seven major internal transformation programs. The scale is unusual, but the method is not: build the infrastructure, deploy AI tools into the actual workflow of every department, train people in small groups rather than through company-wide mandates, and accept that the workforce will be smaller and differently skilled in five years. The question for every organization is not whether that happens. It is whether you are ahead of it or behind it.