AI Shortened Project Cycles from Two Years to Six Months at Deutsche Bank, Says CIO in Bengaluru
Denis Roux, CIO of Deutsche Bank's investment bank, stated at an event in Bengaluru that AI has reduced project cycles from two years to three months. India accounts for 45% of the bank's global technical workforce.
"From Two Years to Three Months": What the CIO of Deutsche Bank Said in Bengaluru
Denis Roux, Chief Information Officer (CIO) of Deutsche Bank's investment bank, chose a number to describe what AI has produced within the bank: technology projects that previously took two years to develop are now completed in three to six months. "We are seeing things that used to take two years being delivered in three to six months," Roux said, according to Reuters, which covered the Bank on Tech, an annual event held on June 18 in Bengaluru, India. Backlogs that historically took months to resolve are dismantled in weeks.
The statement is more specific than the banking standard. Banks generally communicate AI gains in terms of abstract efficiency or the number of pilots in operation. Roux provided a timeline metric: compressing project development cycles from four to one. Reuters did not obtain methodological details of the calculation, but Roux described the pattern as consistent across different types of projects.
The bank is developing two AI systems that the CIO detailed: one automates the extraction and analysis of financial data from large-volume documents and the other connects external events to the bank's portfolio to calculate market exposure in real time. Neither requires the replacement of legacy systems: both function as layers over the already installed infrastructure.
India as the Bank's Technical Gravity Center
Roux spoke in Bengaluru for a structural reason. Deutsche Bank has approximately 9,000 employees in its technology division in India, representing around 45% of its global technical workforce, according to data released by the bank. This percentage positions India, rather than Frankfurt or London, as the epicenter of the institution's technological development.
The Bank on Tech has been held in Bengaluru since the bank recognized that most engineering decisions originate there. Deutsche Bank has operated a Global Capability Centre (GCC) in the country for over a decade, with functions ranging from software development to financial engineering, risk analysis, and applied research. The GCC is no longer a support structure: it is where the bank develops its critical systems.
The difference that AI introduces into this model is the multiplier per engineer. The delivery capacity of each team has increased without a corresponding growth in headcount, and the bank is using this surplus to address a historical backlog of pending work before translating the gain into new services.
Token Costs as a Declared Management Variable
Productivity has advanced, but costs worry Roux. Deutsche Bank has adopted a containment policy: engineers receive a quota of tokens for AI use and need to justify requests for expanding that quota. The justifications are shared among teams to create an internal repository of verified use cases.
The bank deliberately uses simpler models for routine tasks, reserving more capable models for applications where the output quality justifies the cost per token. This decision contrasts with the trend among large corporations that, when launching AI initiatives, indiscriminately adopt cutting-edge models for any use case. For Roux, each use needs to pay for itself.
The policy reflects a problem that European banks are beginning to publicly acknowledge: token spending increases as AI agents become part of the regular workflow. Without consumption controls, the ROI that justified the initial investment can dissipate in unmanaged use.
What This Cycle Anticipates for the Sector
Deutsche Bank has not announced headcount cuts related to AI. The productivity gain is being used to deliver work that was in the queue. It is a phase that other European banks have already experienced before converting efficiency into formal team cuts.
HSBC and Standard Chartered have linked their job reduction plans to the automation of back-office functions: the former is studying cutting up to 20,000 positions, according to a Bloomberg report from March; the latter announced, at an investor event in Hong Kong on May 19, the elimination of 7,800 back-office roles by 2030. Both operate Indian GCCs as a structural part of their transformation plans. Deutsche Bank is at an earlier stage of this same cycle, still in the phase of absorbing productivity before announcing structural headcount impacts.
For CIOs of consulting firms serving European banks, the progression has practical implications: the client that today requests more delivery capacity with the same team may, in two or three quarters, request the same capacity with fewer people.