AI in Finance: News Brief April 2026

Key Takeaways

• Claude Opus 4.7 outperforms GPT-5.5 on long‑form finance tasks
• DeepSeek V4 offers top‑tier AI at roughly 1/35th the cost
• Microsoft Copilot agents now default in Excel, reshaping finance modeling
• KPMG cuts 100 audit partners; EY deploys AI across 160k engagements
• EU AI Act high‑risk deadline August 2, costing ~$54k per system

Pulse Analysis

The latest generation of foundation models is redefining what finance teams can automate. Anthropic’s Claude Opus 4.7 excels at multi‑step, narrative‑heavy tasks such as month‑end close reporting, while OpenAI’s GPT‑5.5 continues to lead raw reasoning power. China’s DeepSeek V4 narrows the performance gap at a fraction of the price—about one‑thirty‑fifth of proprietary alternatives—making open‑source options viable for high‑volume, routine workloads. CFOs should treat model selection as a strategic procurement decision, building modular architectures that allow rapid swaps as capabilities evolve.

Embedded AI agents are now standard features in the enterprise stack. Microsoft’s Copilot is default in Excel, turning the spreadsheet into a live AI assistant; SAP’s Joule suite offers over 30 finance‑specific agents, though adoption remains under 5 %. Workday’s pre‑payment test suite and BlackLine’s control‑layer agents illustrate a shift from experimental pilots to production‑grade, auditable tools. At the same time, the industry is abandoning per‑user pricing in favor of consumption‑based fees, prompting finance leaders to lock in predictable terms while usage patterns solidify. Early renegotiations can capture 15‑20 % discounts before vendors fully transition to usage models.

Audit and regulatory landscapes are moving in lockstep with technology. KPMG’s reduction of 100 audit partners signals a broader labor compression, while EY’s deployment of AI across 160,000 engagements demonstrates that AI‑enabled audit is already mainstream. The EU AI Act’s impending August 2 deadline—without a deferral—means high‑risk finance AI systems must be documented and compliant, incurring roughly $54,000 per system in certification costs. CFOs should inventory all AI assets, map data lineage, and embed audit trails now to avoid costly retrofits. Balancing cost reduction, governance, and compliance will be the defining challenge for finance executives in the AI‑driven era.