Summary

• Société Générale is targeting approximately €1.9 billion of gross savings between 2026 and 2029.
• The bank expects around €500 million less IT spending as simplification, system retirement, and AI-enabled productivity contribute to the programme.
• AI is being treated as one component of a broader operating-model change rather than a standalone technology initiative.

Société Générale has made artificial intelligence part of the financial arithmetic of its next strategic plan, putting AI-enabled productivity alongside IT simplification, procurement savings, and workforce change as it tries to lower its cost base through 2029.

The French banking group is targeting approximately €1.9 billion of gross savings between its estimated 2026 position and 2029, enough to absorb inflation and additional investment while still producing a net reduction in costs. Its 2029 cost base is expected to fall below €16.3 billion.

Technology forms a substantial part of the programme. Société Générale expects IT spending to decline by roughly €500 million, with simplification, removal of redundant and obsolete systems, process redesign, and productivity gains from AI contributing to the reduction.

The targets move enterprise AI into territory investors can eventually measure against the accounts. Rather than merely reporting the number of employees using new tools, management will have to show that technology is contributing to a financial plan with deadlines, ratios, and profitability targets.

AI moves into the operating model

Banks have used machine learning for years in fraud, credit, trading, customer service, and compliance, but generative and agentic systems have widened the range of work that can be automated. Software development, document processing, internal research, operations, and service workflows now sit alongside more established analytical applications.

Société Générale’s strategy places that technology inside process re-engineering rather than separating it into an innovation programme. An assistant that saves an employee several minutes creates a different economic effect from redesigning a workflow around automation, because the latter can change staffing, systems, approval stages, and the amount of legacy technology still required.

The bank’s €1.9 billion savings target should not be treated as an AI number. Procurement, IT simplification, workforce changes, and broader organisational measures all contribute, while even the expected IT reduction includes system retirement and conventional efficiency work alongside AI.

Legacy technology sets the pace

Large banks often operate overlapping applications accumulated through acquisitions, local requirements, regulation, and years of incremental development. Generative systems can make those estates easier to navigate, but they can also become another layer sitting above applications that remain expensive to maintain.

Durable savings require work or systems to disappear rather than simply becoming easier to use. If staff gain copilots while the same applications, controls, reporting requirements, and approval structures remain underneath, the technology bill can increase before meaningful savings arrive.

Financial regulation also limits how quickly autonomy can spread. Models operating around customers, financial decisions, compliance, risk, or regulatory reporting need controls covering data, explainability, security, oversight, and accountability.

That favours lower-risk processes where productivity can be measured without immediately handing consequential decisions to autonomous systems. Software engineering, knowledge retrieval, document handling, and administrative workflows can provide material gains while retaining established approval structures.

Investors now have numbers to test

The wider roadmap targets a cost-to-income ratio below 55% by 2029 and return on tangible equity of between 13% and 14%, rising above 15% thereafter. AI investment will consequently be judged within a broader profitability programme rather than through pilot counts.

Separating AI’s individual contribution will remain difficult because system simplification, procurement, attrition, process redesign, and automation will operate together. Management will nevertheless have to demonstrate that the combined programme is reducing expenditure without weakening controls or customer service.

By 2029, Société Générale should provide a useful case study in whether enterprise AI becomes economically visible inside a large regulated organisation. The bank has moved beyond promising productivity and placed the technology inside a cost plan that can eventually be tested against reported performance.