6 Developments That Changed the Game (While You Were Planning for Christmas)
December 2025 wasn’t just about end-of-year planning. While New Zealand businesses were wrapping up 2025, AI platforms shipped features that fundamentally change how work gets done in 2026. If you’re finalising technology budgets or strategic planning for the year ahead, these six developments matter more than the usual “AI hype cycle” announcements.
1. GPT-5.2: When AI Matches Professional Work Quality
What changed: OpenAI’s GPT-5.2 doesn’t just generate content anymore. It creates work that matches professional standards. On knowledge-work benchmarks, it now beats or ties industry professionals in 70.9% of comparisons across spreadsheets, presentations, coding, and document analysis.
Three deployment options launched: Instant (fast responses), Thinking (deep analysis), and Pro (expert-level quality). Error rates dropped significantly, and the model handles genuinely long documents without losing context.
Business impact: Tasks that previously required professional review—complex spreadsheet analysis, presentation creation, technical documentation—now reach usable quality in first drafts. The productivity gain isn’t marginal; it’s the difference between “AI-assisted” and “AI-delivered” work.
If you’re using ChatGPT Enterprise or the API, evaluate whether tasks currently assigned to junior staff could shift to AI execution with senior review.
2. Google’s Interactions API: Agents That Actually Complete Tasks
What changed: Google introduced the Interactions API—a single endpoint that lets developers deploy agents capable of multi-step workflows, tool use, and background execution. The flagship preview, Gemini Deep Research, handles extended research tasks without constant supervision.
This isn’t another chatbot API. It’s infrastructure for agents that coordinate multiple tools, maintain context across long operations, and run tasks asynchronously.
Business impact: Complex workflows that currently require human coordination, such as gathering data from multiple sources, synthesizing findings, and generating reports, become programmable sequences. The constraint shifts from “can AI do this?” to “have we documented the workflow clearly enough?”
Start identifying processes where the bottleneck is coordination overhead, not expert judgment.
3. The New York Times vs. Perplexity: Licensing Becomes Table Stakes
What changed: The New York Times filed suit against Perplexity AI, alleging unauthorised use of paywalled journalism and outputs that closely mirror Times content. This follows similar actions against OpenAI and represents publishers drawing clear boundaries around content usage.
Business impact: Expect AI providers to accelerate licensing agreements and attribution features. If your organisation generates proprietary content or research, understand how your preferred AI platforms handle data privacy, content rights, and output attribution.
The legal landscape is shifting from “what’s technically possible” to “what’s contractually permitted.” Review your AI usage policies accordingly.
4. Amazon Kindle’s “Ask This Book”: AI Reading Assistance Goes Mainstream
What changed: Amazon deployed “Ask This Book” in the Kindle iOS app for thousands of English-language titles. Readers can highlight passages and ask contextual, spoiler-free questions based on content already read.
Critically: the feature is always-on with no opt-out for authors or publishers. Amazon plans expansion to all devices and Android in 2026.
Business impact: This signals AI features becoming default product experiences, not optional add-ons. If your organisation publishes content, delivers training, or creates documentation, expect similar AI-assisted experiences to become standard expectations.
Consider how your content strategy adapts when readers can interrogate documents directly rather than consuming them linearly.
5. Zoom AI Companion 3.0: From Meeting Notes to Automated Follow-Through
What changed: Zoom launched AI Companion 3.0, moving beyond transcription to automated workflow execution. It drafts documents, manages tasks, and provides a web interface (ai.zoom.us) that connects meetings, chats, and files into actionable outcomes.
The platform runs federated AI—combining Zoom’s models with OpenAI, Anthropic, and NVIDIA—with a $10 standalone option making features accessible without full Zoom subscriptions.
Business impact: Meeting productivity shifts from “capturing what was said” to “executing what was decided.” The coordination overhead of translating discussions into tasks, documents, and follow-up actions becomes automated.
Evaluate whether your organisation has clear decision-making processes that AI can execute reliably. The technology works; the question is whether your workflows are sufficiently structured.
6. Mozilla’s New CEO and Firefox’s Privacy-First AI Approach
What changed: Mozilla appointed Anthony Enzor-DeMeo as CEO and positioned Firefox’s AI features as optional and transparent. Features like “Shake to Summarise” (iOS) and the forthcoming “AI Window” emphasize user control and privacy-respecting design.
Business impact: As Chrome, Edge, and Safari integrate AI features aggressively, Mozilla represents a different approach: AI that users enable rather than disable. For organisations with strict data governance requirements, browser-based AI that doesn’t send content to external servers becomes strategically relevant.
If your team handles sensitive information, evaluate browser-level AI controls alongside application-level AI policies.
Three Realities These Developments Confirm
Frontier capability is reaching mainstream outcomes. GPT-5.2’s improvements target everyday artifacts—spreadsheets, slides, documents—where teams spend real time. AI isn’t just getting “smarter”; it’s getting useful for mundane work.
Agents are becoming the interface. Google’s Interactions API and Zoom’s federated approach shift AI from “assistant that suggests” to “executor that completes.” This requires organizational readiness: documented workflows, clear permissions, rollback procedures.
Licensing and attribution define the next phase. The Times v. Perplexity case accelerates pressure for explicit agreements. Expect more deals—and lawsuits—before norms settle. If you create or consume proprietary content, understand your AI providers’ licensing frameworks.
What This Means for Your 2026 Planning
December’s developments aren’t just technology announcements. They are shifts that affect budget allocation, workflow design, and competitive positioning.
Immediate actions:
- Audit current AI usage: which platforms, what data, what permissions
- Identify high-coordination workflows where agents could deliver immediate value
- Review content licensing implications for both consumption and creation
Strategic planning:
- Budget for AI features becoming default product expectations, not optional add-ons
- Assess organisational readiness: documented processes, API access, governance frameworks
- Develop policies for AI-generated work: quality standards, review requirements, attribution
Risk management:
- Understand content licensing across your AI toolchain
- Evaluate browser-level AI controls for sensitive information handling
- Establish boundaries between AI suggestion and AI execution