Eugene Winter

09/08/2026

big 4 accounting firms use of ai

The Big 4 accounting firms—Deloitte, EY, KPMG, and PwC—are investing billions of dollars to integrate generative and agentic artificial intelligence into their core audit, tax, and advisory workflows.

Core AI Deployments by Firm

Deloitte: Rewired its Omnia platform with end-to-end agentic AI and launched automated finance agents targeting a 25% reduction in operating costs. [1, 2]

EY (Ernst & Young): Deployed thousands of specialized AI agents built with NVIDIA to support hundreds of thousands of professionals across tax, risk, and finance. [1]

KPMG: Leverages platforms like KPMG Ignite and Clara to scan millions of accounting entries and surface operational anomalies, though the firm has also faced industry scrutiny over AI-generated citations in reports. [1, 2]

PwC: Invested heavily in a Microsoft-backed, AI-native audit platform and Agent OS to run automated audit guidance and general ledger analysis tools like GL.ai. [1, 2]

Impact on Operations and Workforce

Full-Population Audits: Instead of examining traditional data samples, AI allows firms to analyze 100% of client journal entries and transactions for anomalies. [1]

Managed Services Pivot: Firms are using AI efficiency to scale multi-year, recurring revenue contracts to run corporate back-office suites. [1]

Restructuring the Pyramid: Routine data entry, document review, and junior-level tasks face high automation rates (70%–95%), altering traditional hiring pipelines and graduate intakes. [1, 2, 3]

adapted

09/08/2026

Nvidia CEO Jensen Huang made the comment on Sunday, days after OpenAI released Astra, its newest and most powerful AI model.

09/05/2026

09/05/2026

Git, GitHub, and GitLab serve different roles in the version control ecosystem, and knowing how they relate helps teams choose the right tools for their development workflow.

Understanding how Git differs from the platforms built on top of it helps clarify where version control ends and where collaboration, automation, and project management begin.



Follow the link to read more, and save this post so you can easily refer back to the full blog explanation and the infographic that breaks down the differences: https://hubs.la/Q04g5ptq0

09/05/2026

09/05/2026

How people in different parts of the world spend their time.

v/Our World in Data

09/05/2026

Over 70 years ago today, the term “artificial intelligence” was coined in a conference proposal: https://stanford.io/2WJJJGN

09/05/2026

In honor of Git's 21st birthday, a poem written in Git: https://shorturl.at/fp80F

09/05/2026

AI Engineering in 2026 can feel overwhelming.

There’s Python, ML, LLMs, RAG, agents, multimodal AI, MLOps, security… and somehow you’re expected to learn all of it.

The good news is: you don’t need to learn everything at once.

Build it layer by layer.

→ Start with 𝐦𝐚𝐭𝐡, 𝐬𝐭𝐚𝐭𝐢𝐬𝐭𝐢𝐜𝐬, and probability so the fundamentals make sense.

→ Get comfortable with 𝐏𝐲𝐭𝐡𝐨𝐧, APIs, Git, testing, debugging, and clean code.

→ Learn 𝐌𝐚𝐜𝐡𝐢𝐧𝐞 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠: supervised learning, feature engineering, evaluation, and tuning.

→ Move into 𝐃𝐞𝐞𝐩 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠: neural networks, CNNs, RNNs, Transformers, and optimization.

Then the modern AI stack starts opening up:

→ 𝐋𝐋𝐌𝐬: prompting, fine-tuning, evaluation, quantization
→ 𝐑𝐀𝐆: embeddings, chunking, vector DBs, hybrid search, reranking
→ 𝐀𝐈 𝐀𝐠𝐞𝐧𝐭𝐬: tools, memory, planning, orchestration, multi-agent systems
→ 𝐌𝐮𝐥𝐭𝐢𝐦𝐨𝐝𝐚𝐥 𝐀𝐈: text, image, audio, video, OCR, vision models
→ 𝐋𝐋𝐌𝐎𝐩𝐬/𝐌𝐋𝐎𝐩𝐬: deployment, monitoring, CI/CD, scaling, and cost control

And once you start building real systems, security and governance become just as important.

Privacy. Guardrails. Compliance. Reliability. Observability.

One thing I’d add: don’t wait until you “finish the roadmap” to build.

Pick a project at every stage.

That’s where the learning actually sticks.

Which part of this roadmap are you on right now?