The most important news pieces from the AI world are here for you, whether you’re an AI enthusiast or just wish to keep an eye on what’s going on.
Focus Topic: AI ROI reality check - what KPMG found and why it matters for Finance
For months, the narrative has been simple – invest in AI or fall behind. But KPMG’s latest global survey of 2,500 technology leaders brings a more nuanced picture.
Yes, 74% of companies report that AI is already creating value. But only 24% say they have actually achieved a return on their AI investments. That gap is the real story. Adoption is widespread. Profitable adoption is not.
On average, organizations that see returns report about 2x ROI. The top performers reach up to 4.5x return. What separates them is not budget size. According to KPMG, success correlates with clear focus, strong data governance, and tight integration of AI into everyday processes. AI as a side experiment does not scale. AI embedded into workflows does.
There is also a risk signal. 69% of leaders admit that moving fast with AI has led to security compromises or growing technical debt. For finance teams, this is critical. Rapid deployment without governance may inflate short-term innovation metrics while quietly increasing long-term operational risk and compliance exposure.
For CFOs and finance leaders, this study reframes AI from a tech trend into a capital allocation question. The real competitive edge is not who spends more on AI – it is who connects AI to measurable business outcomes, manages risk discipline, and aligns it with core processes.
LLMs & AI Models
- Google launched Gemini 3 Deep Think, scoring 84.6% on ARC-AGI-2, beating Opus 4.6 and GPT-5.2, plus gold-level results in physics, chemistry and Codeforces.
- OpenAI introduced GPT-5.3-Codex-Spark on Cerebras chips, generating 1000+ tokens per second, prioritizing speed over benchmark dominance.
- MiniMax released open-source M2.5, matching top coding models at a fraction of the cost, priced from $1.20 per million output tokens.
- Z.ai unveiled GLM-5 (744B parameters) with open weights, scoring 50 on the Intelligence Index and competing closely with leading closed models.
- Anthropic published a Sabotage Risk Report, placing Claude Opus 4.6 in a “gray zone” due to elevated misuse and manipulation risk.
New Tools
- ByteDance launched Seedance 2.0, a 2K AI video model with built-in audio and multi-input support, now in beta via Jimeng AI.
- Waymo introduced a World Model based on DeepMind’s Genie 3, generating hyperreal driving simulations for rare and extreme scenarios.
- ElevenLabs released Audiobooks by ElevenCreative, enabling authors to create, edit and publish AI-narrated audiobooks end-to-end.
👉 Explore these tools: Seedance 2.0 | Waymo World Model | Elevenlabs
Other Quick Picks
- Microsoft AI chief Mustafa Suleyman predicts AI will automate most white-collar tasks within 12–18 months, targeting “professional-level AGI”.
- xAI faces pressure as two more co-founders exit after the SpaceX merger, raising concerns amid model delays and rising competition.
- Former GitHub CEO Thomas Dohmke raised $60M for Entire, building open-source tools to track and manage AI-generated code.
- A Harvard Business Review study found AI increases productivity but expands roles and working hours instead of reducing workload.
- AI companies dominated Super Bowl LX ads, with Anthropic, OpenAI, Google, Amazon and Meta pushing AI into mainstream visibility. Check out my post here.
- Crypto.com CEO bought AI.com for $70M, launching a consumer AI agent platform in the largest domain deal on record.
- Goldman Sachs is using Claude-based agents to automate accounting and compliance, aiming to streamline rule-based processes.
- The US Department of Defense added ChatGPT to GenAI.mil, giving 3 million staff access in a secure government cloud.
- Microsoft plans to build its own frontier AI models by 2026, reducing long-term dependency on OpenAI.
🇪🇪 AI News from Estonia
- Janika Aan became LHV’s new Head of Data & AI, leading group-wide data governance, analytics and AI integration across 20+ teams.
- Cyber expert Jaana Metsamaa warns AI democratizes coding but increases security risks, urging use of dedicated AI security agents and minimal data exposure.
- Estonia’s Foreign Intelligence Service reports that DeepSeek spreads Chinese propaganda, calling biased AI outputs a national security risk.
- Kirke MAAR, head of the Eesti.ai initiative, emphasizes that AI is a structural shift for society, meaning the program must remain agile and avoid overlapping with existing AI councils and networks. Read here.
🎙️ AIPowerment Podcast dropped a new episode featuring Urmo Keskel, co-founder of Phishbite, discussing phishing attacks, Smart-ID risks and AI-driven cyber threats.
🎧 Listen to AIPowerment Podcast on Spotify, Apple Podcasts, and YouTube.
Want to stay in the loop?
Straight to your inbox: practical AI updates, finance use cases, tools to try, and upcoming trainings.
I cut the noise and send only what’s actually useful.
💼 Exploring how to make AI work in finance? Let’s talk practical use cases – connect with Gerlyn Tiigemäe for expert guidance.
📚 If you would like to participate in one of my trainings or listen to speaking engagements, here are the upcoming ones:
- Training “AI võimalused finantsvaldkonnas” @ Äripäeva Akadeemia, next trainings in Q1 2026 – information
- Training “AI võimalused projektijuhtimises” @ Äripäeva Akadeemia, next trainings in Q1 2026 – information
- Training “Tehisintellekti (AI) kasutamine finantsvaldkonna igapäevatöös”, 19.02.2026 – information
