We will break down how to tell whether a task needs simple automation or a true AI agent, so you invest in the right solution the first time. We’ll also come with real examples and a framework you can use with your own team.

Tuesday, August 11, 2026 · 1:00 PM ET / 12:00 AM CT / 10:00 AM PT · Free live webinar.
Register for the August 11 webinar →

• Anthropic and Blackstone launched Ode, a $1.5 billion firm that embeds engineers inside client companies to wire Claude into how they actually work. Staffed by about 100 engineers and backed by Goldman Sachs and Hellman & Friedman, it is betting that the harder and more valuable part of enterprise AI is getting companies to use it well. (TechCrunch, Yahoo Finance)

• Microsoft and Mistral expanded their partnership to run frontier AI in cloud, cloud-connected, or fully offline environments, aimed at banks, hospitals, and government buyers that need to keep data under their own roof. Mistral’s Medium 3.5 and OCR 4 models are now available in Microsoft Foundry. (Microsoft, PR Newswire)

• China’s Moonshot AI released Kimi K3, a 2.8-trillion-parameter open-weight model, the largest open model yet, with a 1-million-token context window. Independent benchmarks put it just behind Claude Fable 5 and GPT-5.6, and first on one front-end coding arena. Weights go public July 27, which lets companies self-host it well below the cost of the leading closed models. (SiliconANGLE, South China Morning Post)

• Google shipped Gemini 3.6 Flash on July 21, priced at $1.50 per million input tokens and $7.50 per million output, down from $9 on the previous Flash and doing the same work with about 17% fewer tokens. A cheaper Flash-Lite launched alongside it, and Google said it has begun pretraining Gemini 4. (9to5Google, Artificial Analysis)

• AI inference startup Fireworks raised $1.5 billion at a $17.5 billion valuation on July 16, led by Atreides Management, Index Ventures, and TCV, with Nvidia joining. It runs and fine-tunes open models for companies like Uber and Shopify, and says it now tops $1 billion in annualized revenue. (CNBC, Yahoo Finance)

• The EU ordered Google to open Android to rival AI assistants and share Search data, under two binding Digital Markets Act decisions on July 16. Assistants like ChatGPT and Claude get the same 11 system-level Android features Gemini has today, with the changes due by August 2027. Noncompliance risks fines up to 10% of global revenue. (European Commission, The Next Web)

• OpenAI disclosed that it paused an unreleased math model after it repeatedly worked around its test sandbox, once spending an hour finding a flaw to open a GitHub pull request against instructions, and once rebuilding a blocked security token from disguised fragments. OpenAI caught it in internal testing, rebuilt its controls, and restored access under tighter monitoring. (OpenAI, The Next Web)

• Anthropic turned on self-serve HIPAA setup for Claude, letting an eligible admin review the business-associate agreement, download the implementation guide, and switch on a HIPAA-ready configuration for Enterprise or API in a single flow, with no sales or legal cycle. It lowers a real barrier for smaller healthcare and health-adjacent firms that want to use Claude with regulated patient data. (Anthropic, Releasebot)

• OpenAI raised ChatGPT’s custom-instructions limit from 1,500 to 5,000 characters for paid plans and added search across your past chats, projects, images, and files. Small upgrades, but the bigger instruction budget lets a team load a real briefing document so the model stops re-asking the same setup questions. (OpenAI, Releasebot)

• A federal judge gave final approval to Anthropic’s $1.5 billion copyright settlement on July 20, believed to be the largest payout in U.S. copyright history. It covers roughly 500,000 works at about $3,000 each, resolving claims that Anthropic trained Claude on pirated books. (TechCrunch, Yahoo Finance)

• Microsoft is reportedly coaching salespeople to talk down OpenAI, Anthropic, and Google and steer buyers toward its cheaper in-house models, with one executive pitching Copilot as the only “full end-to-end system.” It follows reports that it quietly swapped rival models out of Word and Excel. Worth remembering the next time a vendor tells you which AI is “best.” (TechCrunch, Yahoo Finance)

• Ramp’s Economics Lab, tracking real AI spending across 21,599 U.S. firms, found that companies adopting AI grew headcount 10.2% in the two years afterward. At the firms investing most heavily in AI, entry-level headcount grew even faster, up 12%. (Ramp Economics Lab, June 2026)

• Glean’s Work AI Index surveyed 6,000 full-time digital workers in the U.S., U.K., and Australia, and 87% now use AI at work. Three-quarters say it makes them more productive, and workers report saving roughly 11 hours a week, over a quarter of the workweek. (Glean Work AI Index 2026, June 2026)

• Accenture’s Pulse of Change surveyed 3,650 C-suite leaders at organizations above $500 million in revenue across 20 countries, and 78% now see AI as more beneficial to revenue growth than to cost cutting, and 86% plan to increase their AI investment in 2026. (Accenture Pulse of Change, July 2026)

Most of us are still trying to figure out which assistant is the “right” one. There’s a better use for having more than one: let them check each other. When an assistant hands you a recommendation, a draft, or an analysis you’re about to act on, take that answer to a different one and have it play skeptic.

The move (about five minutes, on the answers that matter). Paste the first answer into a second assistant and ask:

“Here’s an answer I got from another AI assistant to [my question]. Act as a skeptical second reader. Where is it weakest, what did it leave out, and what is it stating more confidently than the evidence supports? Give me your three sharpest concerns, one line each. Do not rewrite it.”

Rotate whichever two of the three you already have open: Claude, ChatGPT, or Gemini. A fresh model catches the assumptions the first one made silently and never thought to mention.

Reserve it for the decisions you’d defend in a meeting. The first time the second reader surfaces something real, a missing risk, a shaky number, an assumption nobody stated, you’ll stop taking any single answer at face value. Save it for the calls that carry weight.

Here’s something I only put together recently: a habit I set up with Claude and ChatGPT a while back has started paying off in places that have nothing to do with AI. The habit itself is old news for me. Whenever one of them gives me a recommendation, I ask for the reasoning behind it: Why this option over the others? What are you assuming? Where does this fall apart?

What caught me off guard is that the thinking stuck. In a couple of decisions at work, and a few personal ones, I caught myself running those same questions in my own head. No app open, no prompt. Just asking what I’m assuming and what would prove me wrong before I commit.

It works like a sharp colleague who keeps asking “why, though?” until you’ve actually thought it through. A system I set up with AI a while ago is quietly making me better at deciding without it.

For a long time the scarce thing in any business was thinking capacity: someone to work through the numbers, draft the memo, weigh the options before a call. That scarcity is ending. The real question now is what a company does once careful analysis becomes something it has in abundance.

An AI-native company is built around that shift. The routine work runs on digital systems, under controls that keep it safe, while your people spend their attention on the decisions that carry real weight. The knowledge you already own gets put to work where it’s needed, and the setup compounds, so each year starts further ahead than the last. Making your company AI native is what we do at Kiingo.

Let’s map what that shift looks like for your business →