Deutsche Telekom Puts a €2.5B Price on AI Savings by 2030

At its AI Investor Day on October 5, Deutsche Telekom said AI and automation should cut about €2.5B of indirect costs by 2030 against 2023, with roughly €1.1B of gross savings outside the U.S. due in 2027. The release names no job numbers.

Deutsche Telekom Puts a €2.5B Price on AI Savings by 2030

At its AI Investor Day in Bonn on October 5, Deutsche Telekom named a number. In its press release the group said it aims for savings of about €2.5B in indirect costs by 2030, compared with 2023, from AI and automation. For 2027 it expects gross savings of about €1.1B outside the U.S. Reuters, via Yahoo Finance, converted the 2030 figure to about $2.8B and reported that Telekom reaffirmed its 2026 outlook and medium-term targets.

The release gives no headcount, no FTE target and no restructuring charge. That is the gap this piece works inside.

What the release counts

The savings are in indirect costs, and the group says they come from operations with large volumes of repetitive work: network operations, customer service, software development and administrative processes, as AI News summarized it. The release adds a revenue side. It expects about €250M of AI-related revenue in its non-U.S. business in 2026 and targets €800M by 2030.

Do the unit math on the cost side. If 2027 delivers €1.1B against 2023, that is four years for the first €1.1B. The path to €2.5B then asks for another €1.4B in three. The plan accelerates after 2027, and the release does not say what drives the acceleration. It also does not repeat the 「outside the U.S.」 qualifier on the 2030 figure in the sentence we read, so whether T-Mobile US sits inside the €2.5B is not something the release settles.

The one labor-shaped number is American

The release’s operating evidence is a short list. In the U.S., it says, the number of customer service calls has dropped 55%, and AI agents there now handle 40% of customer contacts. Its chatbot Frag Magenta handled about 2.6M customer service calls in the first half of 2026 alone. A network management use case cut response time from several hours to about a minute, and early use cases show a 30% drop in complaints. More than 100,000 employees have been trained to use AI.

Read the 55% carefully. It is a call volume, not a staffing figure. Fewer calls reaching people is the mechanism by which a service floor shrinks, but the release never says that any role was cut, and neither do we. What it does establish is that the U.S. customer-care operation is the proof point Telekom chose to show investors first.

Cohort: a euro figure with no heads in it

Telekom is now the second large employer in a week to price AI in money and leave people out of the sentence. The same day, C.H. Robinson priced about $300M of RXO synergies at 80% AI-driven and gave no RXO headcount, while its own count fell 22% in two years. BMW tied a 20% management cut to agentic AI on September 30, also without a number of people.

The telecom peers show what the headcount version looks like. Verizon’s layoffs have crossed 16,600 alongside record EBITDA. Amdocs, which builds software for carriers, cut 3,000, 10% of its workforce, in its first reorganization labelled AI-era. Block went from 10,000 employees to under 6,000 after its AI resolved most support tickets. Telekom has not published a number in that family. Its €2.5B target is the same kind of claim, a cost line that only moves if the work does.

What it means for the jobs

The release names the workflows. It does not name the roles. Our read of who sits in those workflows: customer service agents in contact centers, network operations staff who watch traffic and fault queues, software developers on maintenance and testing, and back-office administrators handling billing and records. That is our mapping, not something Telekom said.

The 55% and 40% figures say the customer-service line moves first, because it has the deepest data, the shortest feedback loop and a U.S. operation already running at that rate. Network operations is the second line to watch, since a several-hours-to-one-minute response time is a claim about shift coverage as much as speed. Software development and administration are listed but carry no operating number at all.

Timing is the plannable part. 2027 is the first checkpoint, with €1.1B due outside the U.S. Anyone in a Telekom contact center or operations center in Europe should treat 2027 as the year to see whether the savings show up as attrition without backfill or as an announced program. The next filing to check is the 2026 annual report, which carries the group’s year-end employee count for comparison with whatever Telekom says at its next results date.

Sources

Keep reading