Three weeks ago, I gave an AI access to a sandbox with limited set of information, new credentials, and a Linux server. I told it to help me build things.
Its name is Barnaby. It's a Bichon. (Long story.)
I'm not a developer but know the basics. I'm in management at a commercial insurance company. I've spent my career in insurance operations and enterprise technology — the kind of work where "transformation" usually means an 18-month roadmap and a consulting deck with too many arrows.
But something shifted. AI agents — not chatbots, not copilots, but autonomous systems that actually do things — went from reading research paper summaries on ChatGPT and playing with Virtuals on blockchain to my desktop in less than 1 year. And I realized: the people I lead, the teams I work with, the industry I'm in... most of them aren't ready for what's coming. Not because they're behind. Because nobody's translating this stuff into language that actually helps.
So I started this newsletter. Agency — as in: the power to act. And also, yes, the AI kind.
Periodically, I'll share what I'm seeing, what I'm learning, and what I think it means for people who work for a living. No hype. No jargon-first thinking. Just an honest take from someone navigating this alongside you.
Let's get into it.
📡 THIS WEEK'S SIGNAL
The Era of Agentic Chaos — and How Data Will Save Us MIT Technology Review
AI agents are moving out of demos and into the operational core of the enterprise — handling lead generation, supply chains, and customer workflows end-to-end. The ROI is real, but so is the risk: autonomy without alignment is chaos. This piece makes the case that your data infrastructure is the foundation, not the model. If your data house isn't in order, agents will just make your existing mess faster.
Why it matters: If you're a leader being pitched "agentic AI," the first question isn't "which model?" — it's "is our data ready for something to act on it unsupervised?"
🔗 Link the title to: https://www.technologyreview.com/2026/01/20/1130911/the-era-of-agentic-chaos-and-how-data-will-save-us/
What AI Means for P&C Talent Development Canadian Underwriter
As insurers embed AI into underwriting and claims workflows, 2026 is shaping up as a transition year for Canadian P&C. The real story here isn't about job losses — it's about the skills gap. The industry needs people who understand both insurance and AI. That combination is rare, and the organizations that invest in developing it will have a serious edge.
Why it matters: Every industry has its version of this story. The question for your team isn't "will AI replace us?" — it's "are we learning fast enough to work alongside it?"
🔗 Link the title to: https://canadianunderwriter.ca/news/claims/what-ai-means-for-pc-talent-development/
Going Beyond Pilots: Composable and Sovereign AI MIT Technology Review
Here's a sobering number: only 5% of AI pilots deliver measurable business value, and nearly half of companies abandon AI initiatives before reaching production. The problem isn't the models — it's the infrastructure around them. This article argues for "composable AI" — modular, interoperable systems that can actually scale, rather than bespoke one-off experiments.
Why it matters: If your organization is stuck in "pilot purgatory," it's not a technology problem. It's an architecture and strategy problem. Ask: are we building reusable capabilities, or just demos?
🔗 Link the title to: https://www.technologyreview.com/2026/01/19/1131422/going-beyond-pilots-with-composable-and-sovereign-ai/
Anthropic Launches Cowork — A Claude Agent That Works in Your Files, No Coding Required
Anthropic released Cowork — an agent that extends Claude into your local files and desktop. No coding needed. This is significant because it drops the barrier to entry from "you need to be technical" to "you need to be curious." The agentic tools are coming to everyone, not just engineers.
Why it matters: The window where "I'm not technical" was a valid reason to wait is closing. Tools like this mean agency (the personal kind) is available to anyone willing to try.
🔗 Link the title to: https://venturebeat.com/technology/anthropic-launches-cowork-a-claude-desktop-agent-that-works-in-your-files-no-coding-required/
Salesforce Rebuilds Slackbot as an AI Agent
Salesforce completely rebuilt Slackbot as an AI agent — not just answering questions, but taking actions across your workspace. Microsoft has Copilot, Google has Gemini in Workspace, and now Salesforce is making its play. The workplace AI wars are heating up, and the battleground is your daily tools.
Why it matters: Your team's communication platform is about to become an AI platform. The vendors are betting on it. The question is whether your organization will lead that transition or have it happen to them.
🔗 Link the title to: https://venturebeat.com/technology/salesforce-rolls-out-new-slackbot-ai-agent-as-it-battles-microsoft-and-google-in-workplace-ai/
Cisco and OpenAI Redefine Enterprise Engineering with AI Agents OpenAI Blog
Cisco embedded OpenAI's Codex — an AI software agent — directly into their engineering workflows. Not as a side experiment. In the actual pipeline. This is what "enterprise AI" looks like when it's real: not a chatbot on a landing page, but an agent in the build process, speeding delivery.
Why it matters: When companies like Cisco move AI from "innovation lab" to "how we build," it signals that the transition is no longer optional for their competitors. And yours.
🔗 Link the title to: https://openai.com/index/cisco
🔧 WHAT I'M BUILDING
In recent weeks, I set up an AI research partner that most have or will get to know as ClawdBot — an autonomous agent running on a cloud server, connected to my messaging apps, with access to tools, a browser, and memory that persists between conversations.
Its job: help me stay on top of AI developments, curate what matters, and build the things I don't have time to code myself. Including a microfeed application with built-in content manager and this newsletter.
The setup took about a day. The cost is roughly $55 CAD/month for the infrastructure and subscribing to 2 AI models (Claude & GLM). The agent selected the articles you just read, drafted this edition, and I edited it to sound like me.
That's not a flex — it's a proof point. A non-developer, working evenings and weekends around a family, can build something like this now. A year ago? No chance.
The tools are here. The question is what you'll build with them.
More updates to come as I continue my journey of learning and implementing agentic AI.
💬 ONE QUESTION
What's the one thing about AI at work that nobody in your organization is talking about — but should be?
Share with me. I read everything or my AI assistant will.
Agency is a weekly newsletter about navigating the agentic economy with resilience, curiosity, and — well — agency. Written by a Canadian insurance senior leader who's learning by building, not just reading.
Edition #001