AI pilots are done. Now build the operating system.

Hi Futurist,

In the last few weeks, it’s become clearer and clearer: AI labs are now coming for the application layer too. With OpenAI’s new AgentKit, users can build AI-driven workflows. And with Anthropic’s new feature, Skills, you can create an early form of subagents, designed around your custom workflow. The way we work is shifting from doing the task ourselves to overseeing and instructing AI tools. The first glimpse of that future was laid out in the past few weeks. Therefore, I am sending you insights, inspiration, and innovation straight to your inbox. Let’s dive into the depths of the digital future together and discover the waves of change shaping our industry.

💡 In this post, we're dipping in:

• 📣 Byte-Sized Breakthroughs: AI is no longer just helping you write, it’s starting to help you work. With OpenAI’s AgentKit and Claude’s new Skills, we’re seeing a new operating model emerge: one built on agents, not apps. Deloitte’s recent scandal shows what happens without proper AI guardrails. But their same-day Claude rollout? A signal. The future of work won’t be about job titles. It’ll be about tasks, teams, and tech that knows what to do.

• 🤖 Digital Toolbox: Everyday turns plain prompts into real work, handling emails, meetings, and CRM updates while you focus. Snapdeck turns rough ideas into polished slides in seconds, no design skills needed. And Mosaic rewires video editing with a smart canvas that understands your story. No fluff. Just output.

• 🧐 In Case You Missed It: Anthropic deepens its Claude integrations while Google floods the field with Gemini updates, touching video, code, and search. OpenAI streamlines performance and hints at a new device. Microsoft doubles down on agents, and Meta preps for a more personalized web.

Do you have tips, feedback or ideas? Or just want to give your opinion? Feel free to share it at the bottom of this dip. That's what I'm looking for.

No time to read? Listen to this episode of Digital Dips on Spotify and stay updated while you’re on the move. The link to the podcast is only available to subscribers. If you haven’t subscribe already, I recommend to do so.

Quick highlights of the latest technological developments.

Headstory: AI pilots are done. Now build the operating system.

Deloitte made headlines this October, but for all the wrong reasons. A $440,000 government report, commissioned by Australia’s Department of Employment, collapsed under scrutiny. It cited fake court rulings. Quoted judges who never spoke. Referenced academic work that doesn’t exist. All of it written by AI. All of it published as fact. The result? A refund. An apology. And a reminder to us all: AI without governance is a liability masquerading as productivity.

But AI wasn’t the issue here. The infrastructure around it was. No oversight. No verification. No disclosure. No process. Just hallucinations dressed up as truth. The real failure was organizational. A failure to understand that AI is not a magic wand, it’s a system. And systems demand design, control, and care. Deloitte didn’t just forget to build a safety net. They forgot to think like system architects.

And yet, on the same day this scandal broke, Deloitte made its next big move. A full deployment of Claude across 500,000 employees. The contrast was striking. Public backlash in the morning. AI expansion in the afternoon. Call it bold. Call it naive. But don’t miss the signal: Deloitte, even bruised, sees the future. And they know it demands not just AI access, but AI fluency, embedded, governed, and scaled across every function.

This move showed something very important: a recognition that the future of work is actively being rewritten, and that leaders must equip everyone in the organization with the AI infrastructure and mindset to thrive. The deal with Anthropic signals long‑term commitment, not just to using AI, but to redesigning workflows, creating personas for agents across functions, and embedding agents as collaborators inside the business.

This isn't about one bad report. It's about rethinking how work gets done. And that matters more than ever. AI is not an add-on. It’s not a tool sitting quietly in the corner. It’s a shift in the structure of work itself. It’s changing what we do, how we do it, and, perhaps most critically, who does it. Traditional org‑charts become less relevant. Gen Z and Gen Alpha won’t just be niche experts, they’ll become professional generalists. Roles will collapse; job‑titles will fragment; businesses will spawn faster. Jobs will become tasks, careers will climb walls, not ladders, and org‑charts will give way to work‑charts built around projects.

And the proof is here.

In todays Digital Dips we zoom in on OpenAI's Agent Builder and Claude’s new Skills. These aren’t toys. They’re signals. They’re rewiring how tasks are owned, assigned, and executed. Agents that act like coworkers. Skills that behave like team members. Soon, your “head of operations” won’t be a person. It’ll be a smart mesh of AI and human coordination that executes workflows on demand, across boundaries.

So, what should leaders do?

They should stop pretending AI is a side project. They should stop leaving it to the CTO. And they should ask harder questions. Now.

Ask this:

• Have we designed governance that matches our ambition?

• Do we know which tasks AI is quietly reshaping behind the scenes?

• Are our workflows designed for AI-enhanced teams or for 20th-century structures?

• Are we teaching employees to build agents, not just use apps?

• Have we stopped asking “What’s your job title?” and started asking “What’s your impact?”

• Are we measuring performance in outputs, or outcomes?

Because here’s the truth: AI is challenging how we think about work. The real risk isn’t just automation. I believe it’s stagnation. The belief that the org chart we have today is the one we’ll need tomorrow.

It won’t be.

A recent study by Massachusetts Institute of Technology (MIT) found that a staggering 95% of enterprise AI pilots fail to deliver measurable return. Only 5% proceed into production with real business value. Yes, you read that right. The report pinpoints the reason: it isn’t that the models are weak. It’s that leadership, culture and operations are weak. If your underlying systems are weak, adding AI just creates a faster train wreck. It won’t solve your problem, it just accelerates the crash. You can’t “pilot” structural change. You have to build for it. From the ground up.

The winners in this next era? They won’t be the fastest adopters. They’ll be the most intentional builders. The ones who move from AI hype to human capability. From pilots to practices. From chaos to clarity.

The org chart won’t save you. The right operating model will. Let’s build the workplace of the future.

OpenAI launches AgentKit and Apps SDK

TL;DR

OpenAI dropped two big updates: AgentKit for building AI workflows and Apps SDK for integrating interactive apps into ChatGPT. AgentKit lets teams design agents with drag-and-drop logic. Apps SDK introduces app experiences directly in the chat. Both signal one thing: ChatGPT is not just a chatbot, it’s becoming a platform.

Read it yourself?

Sentiment

This is big news. AgentKit makes it easier to build AI agents that can talk to your tools and work with your data. It’s not a Zapier-killer, but it does open the door for companies to create serious agent-based workflows. And the Apps SDK? Smart move. It turns ChatGPT into the front door of your digital workspace, where you can talk to your tools, take action, and get results. Just like an operating system.

My thoughts

OpenAI no longer just makes AI models. It builds software layers on top of them. And AgentKit proves it. Yes, you can now build AI agents with a visual editor. Yes, it’s all integrated with OpenAI’s tools. But no, it’s not simple. It doesn’t replace Zapier, or Make, or Lindy. It’s not for non-tech users. The complexity hasn’t disappeared, it’s just been hidden behind better design. That’s what makes this moment interesting. It’s for teams who want control, speed, and structure, all within OpenAI’s ecosystem. The real shift? Agents move from theory to production. From a playground to the workplace. From "cool demo" to "actually saves time." It doesn’t lower the bar, it just moves it. You still need someone who knows what they’re doing. But if you do, the power is wild. Apps SDK brings another layer. It puts real apps inside the ChatGPT window, so you’re no longer just typing. You’re clicking, dragging, building, buying. Want to trigger campaigns, pull CRM data, or build a Canva deck? Now you can. The chat becomes a workspace. We’ve been stuck in the “text-only” phase of AI for too long. That’s about to end. Because ChatGPT becomes the place where real work happens. It’s not the future of tools. It’s the future of software. And it just got closer. ChatGPT is becoming the browser.

Claude introduces Skills

TL;DR

Claude now supports "Skills", custom instruction folders it can use when needed. These skills include everything from brand guidelines to code, and they allow Claude to automatically apply your workflows across all apps, code environments and APIs. It only loads the context it needs, exactly when it’s relevant. One skill can make Claude your Excel wizard. Another, your brand police. It’s like plug-ins, but smarter.

Read it yourself?

Sentiment

People online are excited about this new feature, which they see as a more accessible step toward true agent capabilities. Not quite sub-agents, but close. Most of the buzz is around how Skills are invoked by Claude, based purely on context. Some even say this is bigger than MCPs, where those eat up the context window, Skills stay lean and modular.

My thoughts

This is the first AI feature that actually fixes the thing everyone complains about, and no, it’s not memory. It’s context. Real, useful context. Skills let you package up how you work, once, and Claude just… gets it. You don’t just “teach” Claude something, it applies that knowledge automatically, across tools, based on what you're doing. No more repeating yourself. No extra prompts. No uploading brand decks for the 14th time. But here’s the bigger shift: Skills are modular. They stack. They scale. This mirrors how we should think about AI workflows: composable, reusable, and context-aware. That’s the real win. These skills can also run code. Which means they can talk to your tools. That’s wild. So yes, this makes Skills way more usable than MCPs. It’s one Markdown file. Break your task into chunks, turn each into a skill, drop them into a folder, and boom. You’ve just built your own AI ops team.

More byte-sized breakthroughs:

• Figure shows off its latest robot for home and work
  Figure 03 is a humanoid robot that looks and moves more like a person than ever before. Built for Helix, the home, and mass production, it features better hands, eyes, and charging, plus real-time learning from its surroundings. At home it’s soft, safe, and smart. At work, it’s strong and scalable. It’s one step closer to putting a useful robot in every room. And this time, it’s built to scale, not just to show off.

• Elevenlabs Agent Workflows turns anyone into a support engineer
  Elevenlabs introduces Agent Workflows. It lets you visually design smart AI conversations using drag-and-drop logic. Switch models mid-chat, change voice, or trigger tools at the right moment, all without touching a line of code. It’s built for teams who want full control over how their agents respond, act, and escalate. Think of it as your AI call centre, built like Lego. Now anyone can build sophisticated support flows in minutes.

• Microsoft Copilot gets smarter across Excel, Teams and more

  Microsoft’s AI updates are coming in fast. Excel now works like an IDE, just prompt and Copilot builds in your sheet. Teams adds multi-user agents that actually understand your meeting. You can switch between OpenAI and Anthropic models, generate code straight from chats, and tap into enterprise data with Knowledge Agent. Work just got a new teammate, and this one never forgets what was said last week.

Three (new) tools that you can use for your work now or later.

Everyday – Turn your intentions into outcomes

Everyday, from drafting emails to running workflows, so you don’t have to. Just tell it in plain language what you need done, and it runs in the background. It clears your inbox, schedules meetings, updates your CRM, and runs workflows, all from a natural promptIt frees you to focus on what matters most, without juggling tools or tabs. It’s like having a superpower in your browser. is your personal AI that handles tasks

Snapdeck – Build slides by typing, not designing

With Snapdeck, you no longer wrestle with slide layouts or design tools. Just type in your topic and audience, and it generates clean, interactive decks in seconds. No fiddling with fonts, layouts, or colours — it handles all of that for you. Charts, diagrams and themes adapt automatically to your content. You stay in the driver’s seat, making changes where needed, without ever starting from scratch. It's fast, clear, and finally makes slide creation feel like writing, not designing.

Mosaic – Video editing with a mind of its own

Mosaic is an infinite canvas where your video ideas come to life. Just drag, prompt, and connect, your footage becomes stories, your edits become fast. It cuts filler, keeps context, and makes your story flow, without you lifting a finger. From transcripts to timelines, it all syncs automatically. No timelines to scrub or clips to trim manually. Ask it to find emotions, moments, light, or structure. Mosaic sees it all. And when you’re done, export straight to Premiere or Final Cut.

A roundup of updates that are too cheesy to ignore.

• Anthropic’s Claude lands on Slack, ready to assist via DMs, threads, and the AI assistant panel.

• Anthropic’s Claude Code introduces Plugins in public beta, enhancing functionality with slash commands and agents.

• Anthropic’s Claude Haiku 4.5 rivals Sonnet 4's coding power at a fraction of the cost and speed.

• Anthropic’s Claude integrates enterprise search to centralize your company’s knowledge from connected tools.

• Anthropic’s Claude integrates with Microsoft 365, searching SharePoint, OneDrive, Outlook, and Teams seamlessly.

• Salesforce selects Claude as the go-to model for regulated sectors, enhancing Slack integration and empowering engineers with Claude Code.

• Salesforce’s Agentforce 360 unifies humans and data on a single platform with Customer 360 and Data 360.

• Hume’s AI Octave 2 debuts with faster, cheaper multilingual text-to-speech.

• Microsoft debuts Agent Framework in Azure AI, streamlining multi-agent system management for developers.

• Microsoft transforms every PC into an AI powerhouse with Copilot at its core with Windows 11.

• Freepik’s Clip Editor lets you modify, extend, and add audio to clips with simple prompts.

• Synthesia 3.0 debuts Video Agents for interactive roleplays and revamped Avatars with lifelike movements.

• Riverside FM launches ChatEdit, transforming video editing with conversational AI.

• IBM's Granite 4.0 rolls out, offering efficient small language models for enterprise needs.

• Perplexity Comet browser is now globally available, transforming how you navigate the internet.

• Perplexity reinvents language learning with interactive experiences on iOS and web.

• Google's Gemini 2.5 Computer Use model debuts, mastering web tasks with ease for API developers.

• Google's Gemini 2.5 Flash Image (Nano Banana) enriches creativity with 10 aspect ratios and seamless image edits.

• Google Home enters the Gemini era with AI-powered devices and a revamped app.

• Google’s Veo 3.1 boosts video creation with 30-second clips, 1080p quality, and dynamic scene control.

• Google Flow’s introduces scenebuilder 'Save a Frame' feature, helps you capture and reuse standout video moments for future projects.

• Google’s Flow now lets you insert and generate objects directly in your videos for seamless edits.

• Google’s CodeMender AI agent leverages Gemini Deep Think to autonomously patch critical software vulnerabilities.

• Google’s EmbeddingGemma rolls out a 308M parameter model for powerful offline AI tasks like semantic search and RAG.

• Google introduces Gemini Enterprise to streamline AI-powered workflows for every employee.

• Google ADK now supports all major AI protocols, linking agents with external tools, other agents, and users.

• Google launches Analytics Advisor for GA4, offering tailored, transparent insights powered by Conversational AI.

• Google introduces grounding with Google Maps in Gemini API, linking data from 250 million places for richer experiences.

• Genspark AI expands its reach with Meeting Notes now available on web and all devices.

• Genspark AI unveils Custom Super Agent for creating AI agents in minutes with a simple prompt.

• Manus 1.5 supercharges full-stack web app creation with unlimited context and AI features.

• Lindy AI’s Gaia AI phone agents just got faster, shaving over 500ms off response times.

• Lindy AI CMO agents now automate full marketing workflows, with new Sora 2, Veo 3.1, Nano Banana, and GPT Imagen integrations.

• Ming-UniAudio introduces a single model for universal speech understanding, generation, and free-form editing.

• Higgsfield AI unveils WAN Camera Control with 15+ dynamic moves like bullet time and dolly in.

• TildeOpen releases a 30B multilingual LLM that’s EU AI Act-compliant, boosting European AI initiative.

• VEED unveils Fabric for Business, turning images into on-brand AI-talking videos.

• Emergent Labs Voice Mode enables hands-free app creation, transforming everyday moments into productive ones.

• Slack unveils context-aware AI apps connecting directly to your workflow.

• OpenAI ChatGPT’s new Slack app powers your sidebar with real-time search and AI insights.

• OpenAI and Jony Ive craft a palm-sized AI, grappling with compute and privacy hurdles for a 2026 debut.

• OpenAI speeds up GPT-5 API requests by 40% with priority processing and introduces a detailed service health dashboard.

• OpenAI’s Sora 2 and Sora 2 Pro hit the API with new cost-effective image and speech models.

• OpenAI’s Sora 2 gives Pro users web storyboards, letting all users craft 15-second videos anytime.

• OpenAI’s ChatGPT now curates and organizes your memories, rolling out globally to Plus and Pro users.

• Matts Humer’s Sora Extend lets you create infinite-length Sora 2 videos, beating OpenAI's 12-second cap.

• Pietro Schirano’s Sora MCP empowers servers to generate, remix, and download videos with ease.

• AMD partners with OpenAI to power massive AI deployments with 6GW of infrastructure.

• Pika unveils Predictive Video, transforming short thoughts into full-fledged video productions.

• Boba AI Labs Anime 1.4 revolutionizes animation with enhanced motion, dialogue, and character consistency.

• xAI’s Imagine v0.9 elevates video-making with enhanced visuals, motion, and audio, now free on all platforms.

• ElevenLabs unveils open-source UI components for creating customizable AI voice and audio agents.

• NeuroChat reads brain signals to tweak responses based on your engagement.

• Retool's AI AppGen crafts custom enterprise apps instantly, integrating seamlessly with your data and security.

• Browserbase and 1Password team up for secure AI-Agent password autofill in your browser.

• Wabi debuts as the first personal software platform to help you create customized apps, ad-free.

• Exa unveils Exa 2.0: a search engine tailored for AI's rapid, complex queries.

• The European Commission introduces the Apply AI and AI in Science strategies to boost AI in key sectors.

• Runway's new Apps simplify creation with tailored workflows for stunning outputs.

• Vercel now supports shipping ChatGPT apps, streamlining deployment for developers.

• CapCut rolls out AI Design on desktop and web, offering 10 daily freebies for crafting flat images.

• Meta is using AI interactions to personalize ads and content on its platforms, starting December 16, 2025.

• MIT's SEAL AI rewrites its own code, outperforming GPT-4.1 in self-driven learning.

This was it. Our forty-third digital dip together. It might seem like a lot, but remember; this wasn't even everything that happened in the past few weeks. This was just a fraction.

I help business leaders turn AI ambition into action. I design governance that scales. I help teams redesign how they operate, with AI embedded at the core. If you’re ready to build the new operating model of work, one that’s fluid, task-based, and AI-native, I’d love to help you lead it. Hit me up on LinkedIn of reply to this email.

Looking forward to what tomorrow brings! ▽

-Wesley