Deutsche Bank says AI cuts tech projects to months
TL;DR:
- Deutsche Bank says AI is compressing technology projects that once took two years into three to six months.
- The bank is using AI to clear internal backlogs while watching rising compute costs as providers shift to usage-based pricing.
- Engineers receive token quotas and must justify extra capacity, mirroring the cost discipline firms learned with cloud.
Deutsche Bank has put a concrete figure on AI’s productivity dividend, with a senior executive claiming the technology is turning multi-year technology projects into ones delivered in months. Denis Roux, chief information officer for the German lender’s investment bank, told Reuters on the sidelines of an event in Bengaluru that work once scoped at two years is “now getting done in three to six months”, though he declined to quantify the overall impact.
Productivity gains, with a cost caveat
Roux said backlogs that used to take months are now cleared in weeks, with AI helping the bank chip away at internal work. But the more interesting signal is the cautious framing around cost. As AI providers such as Anthropic and OpenAI move from subscriptions to token-based pricing that charges by usage, Deutsche Bank is treating consumption as something to be governed rather than left open-ended. Engineers are given token quotas and can request more capacity only if they can demonstrate value, with the lessons shared across teams — a discipline Roux likened to the cost control firms developed when migrating to cloud computing.
That measured tone stands out against louder claims of AI transformation. It also reflects where the bank places its bets: simpler models for routine tasks, traditional software where it still works best, and AI reserved for higher-value applications such as extracting financial data or linking geopolitical events to portfolio exposure. Much of this runs through its Indian technology hub, which employs around 9,000 people — roughly 45% of its global tech workforce.
Looking forward
For UK financial services, where peers from HSBC to Legal & General are scaling AI across operations, Deutsche Bank’s stance is a useful corrective to hype. The productivity story is real, but so is the emerging cost story: usage-based pricing means efficiency gains can be quietly eroded by runaway consumption. The banks that benefit most will likely be those that pair ambition with the kind of token-level cost governance Deutsche Bank is now building — treating AI spend with the same rigour they once brought to the cloud.