"Now that your teams use generative AI tools, why isn't your financial proposal divided by 4?"
The "magic expectation" effect
During budget framing for the core system overhaul of a large financial institution, the CIO challenged us on the supposed gains of AI. In most executive committees, generative AI is seen as a budget steamroller able to cut delivery costs by a factor of 4 overnight.
Result after the trade-offs:
- Expectation of a 75% drop in delivery costs
- Legacy integration, testing and data governance costs left out
- Increased friction between finance, IT and suppliers
The 4 drivers of the budget gap
- Development is only 25% of the iceberg. Cutting coding time by 3 saves only 10% to 15% of the overall program budget.
- The workload shifts. Coding faster produces more code to review, test, secure and integrate.
- Underestimated legacy and data integration. Connecting to legacy systems offers very little room for AI automation (5% to 15% gain).
- Decision latency remains. What slows projects down is not writing speed, but blocked trade-offs between business, IT and procurement.
faster coding, but development accounts for only 25% of the total workload of an IT project.
real net gain across the IT program lifecycle, far from the hoped-for -75%.
"Generative AI speeds up writing code, but it solves neither decision latency nor the complexity of integrating with legacy systems."
Over to you: in your organization, what real net gain have AI tools delivered?
The series "IT & AI: the reality on the ground"
Why 70% of IT projects driftEpisode 2
Generative AI and IT budgets: the "divided by 4" illusionEpisode 3
The "Fake Agile" trap and weak governanceEpisode 4
The POC syndrome: why 80% of AI projects failEpisode 5
The new execution contract: governance, capabilities and P&L
Digital, Data & AI Strategy
Data & AI roadmap, architecture choices, data governance and high-ROI use cases.
Fakhri Essedik
Founder of EDX Advisory
For more than 25 years, Fakhri Essedik has helped executive teams and CIOs steer major Digital, Data & AI programs across Europe, Africa and the Middle East.