How to Setup & Use Claude Cowork to Build Agents. A 60-minute walkthrough of Anthropic's new desktop agent. We'll cover what Cowork is and how it differs from chat, how to set up and run your first agent from scratch, real agent examples you can adapt, and practical agent-design patterns to use from day one.

Whether you're new to Cowork or already pushing past one-off setups into real agent workflows, you'll leave with something usable.

Presented by Josh Sullivan, COO at Kiingo AI

This Thursday, May 14, 2026
10:00 AM PT / 12:00 PM CT / 1:00 PM ET
Save your spot here →

Note: requires a paid Claude plan.

• OpenAI launched its Deployment Company with $4B from 19 partners including Bain, Brookfield, and TPG. Engineers will embed inside customer orgs to run AI rollouts. McKinsey, Deloitte, and Accenture now have competition that owns the model layer. (OpenAI, CNBC)

• Anthropic had a major enterprise week. It stood up a parallel $1.5B services arm with Goldman, Blackstone, Apollo, and Sequoia to embed engineers inside customer orgs; shipped ten finance agents (KYC, pitchbooks, month-end close), Claude add-ins for Excel, Word, PowerPoint, and Outlook, and a native Moody's app covering 600M+ companies; made Claude Opus 4.7 generally available; and acquired all 300+ megawatts of compute at SpaceX's Memphis Colossus 1 data center, doubling rate limits for paid plans. If your vendor strategy was "OpenAI for everything, Anthropic for code," it's now out of date. (Anthropic, Fortune, Bloomberg)

• Salesforce went headless. The new Headless 360 product turns every Salesforce capability into an API or command-line call, no browser required. Translation: AI agents can now drive the entire Salesforce stack directly. Expect more SaaS to follow, since "the UI" is increasingly something only humans need. (Salesforce, CIO)

• JPMorgan moved AI from R&D to core infrastructure in its 2026 budget, treating it as non-negotiable alongside cybersecurity and operational resilience. Total tech spend will hit $19.8B this year. The internal LLM Suite now reaches more than 230,000 employees, with roughly half using it daily. Translation: at the largest US bank, AI is no longer a pilot, it's overhead. (Fortune, Banking Exchange)

• Microsoft Agent 365 is generally available, bundled in a new $99/user/month E7 SKU. Agent 365 is Microsoft's control plane for governing AI agents across the enterprise. Copilot Cowork also went mobile with Power BI and Dynamics integrations. For a 1,000-person company, E7 pencils out to roughly $1.2M annually on top of existing licensing. (Microsoft, VentureBeat)

• Perplexity made its Personal Computer agent generally available to all Mac users, after launching it in April at $200/month for Max subscribers. The app turns a Mac (especially a Mac Mini) into an always-on AI agent that works across your files, apps, and the web — and you can send it tasks remotely from your iPhone. Of a piece with Cowork, Agent 365, and Headless 360: agents are leaving the chat window. (Perplexity, TechCrunch)

• GPT-5.5 Instant is the new default ChatGPT model. OpenAI says it produces 52.5% fewer hallucinated claims on high-stakes prompts (medicine, law, finance), uses about 30% fewer words, and dials back the gratuitous emojis. The 52.5% figure is OpenAI's own benchmark; independent evaluations have not yet been published. (TechCrunch)

• Alphabet briefly passed Nvidia in after-hours trading and is closing the market cap gap. Wall Street is paying a premium for vertical integration: Alphabet owns the chips (TPUs), the model (Gemini), and the cloud (GCP). Google Cloud grew 63% to $20B last quarter with a $460B backlog. Takeaway for vendor strategy: full-stack is beating best-of-breed this quarter. (Fortune)

Three findings worth knowing this week.

• Deloitte surveyed 100 U.S. private-company leaders ($100M–$1B+) and found 64% of firms above $500M report moderate-to-significant AI ROI, versus 11% of smaller firms. 52% now say expanding AI is a top-three priority, up from 22% a year ago. The top barrier: data quality (72%). (Deloitte Private Survey, April 2026)

• BCG analyzed 1,500 U.S. roles and forecasts AI will reshape 50–55% of them within three years, with senior workers gaining expanded responsibilities as AI handles routine tasks. Only 10–15% face full replacement over five years. (BCG via CBS News, April 2026)

• The Atlanta Fed's March survey of 750 corporate executives found AI adoption widespread but productivity gains still ahead, expected to strengthen through 2026 in high-skill services and finance. Larger firms expect AI to reshape staffing over time, while smaller firms anticipate modest employment gains. (Atlanta Fed working paper / NBER w34984, March 2026)

OpenAI and Anthropic both launched consulting arms this week to embed engineers in customer offices. That help won't come cheap. This prompt walks you through the same pilot-picking work yourself, in five minutes, on a team you already know.

[Paste: (1) the team you're considering, and (2) what they do in a typical week. 3 to 7 short phrases will do.]

Help me pick the best 30-day AI pilot for my team. Rate each activity 1 to 5 on four things: how often it happens, how repetitive it is, how much of the input is already written down somewhere, and whether you can measure if it worked. Pick the one with the highest total. For that activity, tell me what success looks like at Day 30, what to do Monday, what has to be true first, and the biggest thing that would kill it.

Return as: The pilot activity + score breakdown + Day 30 success + Monday's first step + Prerequisite + Killer risk.

There's a small but satisfying thing that happens when I sign in on Monday morning and realize last week's work is already neatly organized for me. Not by magic. By the "all knowing" assistant I closed out the week with on Friday.

Lately, instead of just ending a session, I've started signing off with my AI models the way I'd sign off with a coworker: "Great work this week. Let's stop here and pick it back up Monday." That one sentence is doing two jobs. It signals the model to save state, commit something to memory, or draft a recap doc before closing. And it sets the rejoining moment. Monday-me gets greeted by Friday-me's summary instead of a blank cursor.

I've started thinking of these as context bookends. The closing line matters almost as much as the opening prompt. The more you treat an AI like an assistant, intern, or even colleague, the more naturally your workflows take shape around it. You stop ending sessions mid-thought. You start writing the kind of close-out a good teammate would write, because in a working sense that's what's on the other side.

None of this is anthropomorphizing for fun. The structure of "Friday goodbye, Monday hello" is exactly what makes any working relationship durable. Try it once and see how Monday lands.

Deloitte reported this week that 64% of private firms over $500M are seeing real ROI on AI, while only 11% of smaller firms are. Same technology in both groups. The difference is whether anyone has actually redesigned the workflow underneath.

That's where Kiingo helps. Training, strategy, and implementation focused on the work that closes that gap. If you want to know whether yours is set up to land in the 64% or the 11%, that's exactly the conversation we have. kiingo.com/contact