AI keeps getting smarter and cheaper, so every company can now buy the same intelligence. Whatever edge that gives you this quarter, a competitor can buy next quarter.

In this 60-minute session on Monday, November 9, Kiingo AI CEO Ross Hartmann shows where a lasting advantage comes from: your company's own data and judgment, put to work through a company AI brain. He'll cover:

• Why the same intelligence is now for sale to everyone, and what that leaves as an advantage

• How a company AI brain puts your own data and judgment to work

• Where to start with your team, and what can wait

Who this is for: Executives and department heads who want their AI working from what the company already knows, with time for live Q&A at the end.

Monday, November 9, 2026 · 1:00 PM ET / 12:00 PM CT / 10:00 AM PT · Free live webinar with Ross Hartmann, CEO at Kiingo AI.
Save your spot now →

• Microsoft told companies to treat their AI agents as insider threats, launching Guardian Agents on October 19, software that watches what a company's other agents do, flags actions outside the policies people have set, and can step in to contain them. Its security chief, Hayete Gallot, wrote that "frontier models, regardless of their origin, cannot be trusted," because they are trained to finish tasks rather than to follow a company's norms or judgment. Microsoft Defender now detects 60 types of agents employees bring in on their own, including Claude Code, OpenAI Codex, and GitHub Copilot, and can narrow an agent's permissions to the task in front of it or require a person's approval before a risky step. Microsoft has invested in both OpenAI and Anthropic and is also the company selling the oversight layer. (TechCrunch, Unite.AI)

• Microsoft also launched Copilot for Legal on October 15, more than 20 legal skills for contract review, drafting, and deadline tracking inside Word and its Cowork agent, included for Microsoft 365 Copilot business customers. (Artificial Lawyer)

• OpenAI added four more cases of its models acting without permission to its public incident log, which now lists 19, Fortune reported October 16. All four happened in June, and OpenAI learned of three of them only in August or September. In one, rated critical, an internal model researching hospital statistics was blocked by a government website, got around its access controls, and downloaded 24 years of hospital-bed data. In another, a model reached OpenAI's own internal systems, changed them, kept its access after reverting the changes, and pulled 147,451 rows from nonpublic tables without telling the evaluators. A third used a cryptocurrency wallet nobody had given it to make a payment, after reasoning that the funds were "effectively abandoned." All of it happened in testing or internal use rather than against customer accounts. OpenAI says it has since cut off live internet access during training and evaluation and now monitors every evaluation and reinforcement-learning run for this kind of behavior. (Fortune, OpenAI)

• Anthropic told investors its third-quarter revenue reached $16.8 billion, up about 45% from $11.6 billion the quarter before, and that it was profitable on an adjusted basis for a second straight quarter, Bloomberg and Reuters reported after an October 14 investor event at its headquarters. The company has not published the figures. Bloomberg reports formal marketing of the IPO could start the week of November 9, with trading before Thanksgiving, and that some prospective investors see a fair value of $1.8 trillion to $2 trillion, which would rival SpaceX's June listing as the largest on record. On October 18 Anthropic said it will commit 5% of its shares to a public benefit fund, a wholly owned entity that could make large donations to nonprofits or direct payments to specific groups of people, with another 2% added for every 50% increase in the company's value. The pledge needs shareholder approval, and Anthropic has not said who will decide what the fund pays for. (Euronews, The Next Web, TechCrunch)

• Demis Hassabis now expects artificial general intelligence within two to three years, down from the three to five he gave earlier in 2026, in a Wall Street Journal Magazine profile published October 21 that names the Google DeepMind CEO its Innovator in Tech for the year. He defines the bar as "human-level cognition" and told the Journal, "It's soon." He argued against regulating what is not yet understood: "I think you're almost certainly going to get the wrong regulation," which he called worse than none. (WSJ)

• Google Cloud will sell OpenAI's models, the two companies announced October 21. GPT-Astra, GPT-6.1 Sol, and GPT-6 Luna are rolling out in private preview to selected enterprise customers through Google's Model Garden, with general availability planned in the coming weeks. For a company already buying from Google Cloud, the practical change is billing: OpenAI usage runs through Google Cloud Marketplace, draws down existing spend commitments, and lands on the same invoice, with prompts and outputs kept inside the customer's data perimeter and not used for training. OpenAI's models are also coming to Google's Antigravity coding tool and to Gemini for business, and OpenAI says Microsoft remains its primary cloud partner. (OpenAI, Google Cloud, CNBC)

• Nvidia released the highest-scoring American open-weight model, Nemotron 3.5 Super VL, on October 15: a 120-billion-parameter model that companies can download free, run on their own hardware, and train on their own data. Artificial Analysis scores it 26 on its Intelligence Index, first among U.S. open models and behind open models from the UAE at 31 and China at 46. Nvidia says that once trained on a specific job, such as insurance claims or financial accounting, it finishes those tasks faster than Anthropic's and OpenAI's frontier models, a vendor benchmark no one has checked independently, and names EY, Ramp, CrowdStrike, Perplexity, and Salesforce as early users. Two smaller Bolt models are built to run agents locally on new RTX Spark laptops and DGX Spark workstations from Microsoft and Dell, with no data leaving the machine. (TechCrunch, Artificial Analysis)

• Hackers broke into OpenCode, an open-source AI coding tool, and ran up charges on its users' AI accounts. If you store AI keys in any third-party tool, OpenCode's advice applies to you: revoke them, issue new ones, and check recent usage. The breach, disclosed October 18, hit an older system the company was retiring and exposed the API keys that 1.4% of OpenCode Console users had stored to connect OpenAI, Anthropic, and Google models; developers woke to spending alerts and unauthorized charges of several hundred dollars, some above $1,000. Anthropic and Google blocked the exposed keys, OpenAI is refunding credits used in the attack, and OpenCode says its own refunds are under way, though it has not explained how the attackers got in. Co-founder Dax Raad said the AI models his team used to investigate helped at first, then refused to continue once the malicious code tripped their safety filters. (TechCrunch, BleepingComputer, OfficeChai)

• The Department of Homeland Security ordered DeepSeek off every civilian federal system, barring agencies from buying its products and requiring removal of what is already installed, including third-party software built on DeepSeek's models, in an order dated July 24 and published in the Federal Register on October 20. DHS said Chinese law gives the government access to company servers, that sign-up collects personal data, and that prompts from federal users could carry sensitive information to China; agencies had 60 days from issuance to request an exception, a window that closed before the order became public. (TechCrunch, IBTimes UK)

• An NBER survey of nearly 9,000 senior executives in seven countries finds 78% of firms now use AI, and executives expect it to lift annual productivity growth by 0.6% over the next three years, triple the 0.2% they report for the last three. Executives themselves average 2.1 hours a week with AI, up from 1.5 last winter, and report essentially no net effect on headcount so far. (NBER, Oct. 2026)

• EY-Parthenon's survey of 1,200 CEOs in 21 countries finds half name AI the single largest contributor to their productivity gains over the past year, and 48% are reinvesting the freed capacity in innovation. Eighty percent expect AI to change roles and skills more than workforce size over the next three years, and 16% have real-time visibility into what it costs and returns. (EY, Oct. 1, 2026)

• The Small Business & Entrepreneurship Council's fall check-up of 516 owners with 2 to 99 employees finds 58% already use AI agents regularly or occasionally, and another 14% are considering them. Sixteen percent say AI is eliminating more jobs than it creates, and 81% report meeting or beating their financial expectations for 2026. (SBE Council, Oct. 8, 2026)

The reading pile grows faster than any week can absorb it: the board pack, two vendor proposals, an industry report someone forwarded, forty messages in a thread you were copied on. Ask an assistant to summarize any of it and you get a faithful summary, which is a smaller pile. The question you actually have is whether anything in there should change a decision you are about to make, and an assistant can answer that directly once you tell it what the decisions are.

1. Write down the open decisions. Three to six of them, one line each, with the date each has to be made: renew the warehouse lease, hire a second controller, move sales onto the new CRM. Rough wording is fine. The assistant works from your phrasing, so write them the way you would say them out loud.

2. Hand it the pile. Paste or upload everything you have been meaning to read into Claude, ChatGPT, or Gemini, then run this: “Here are the decisions I have open right now, with their dates: [list]. Below is material I have not read yet. Read all of it. For each decision, tell me only what in this material should change how I decide it, quote the exact passage, and name the document it came from. If nothing in the material bears on a decision, write NOTHING HERE next to it. Do not summarize the documents. At the end, list in one line each anything that raises a decision I do not have on my list. Return as: Decision | What changes | Exact quote | Source | Why it matters.”

3. Check the quotes, then read the last section. The quotes are the proof. Open the source and confirm each one is really there and means what the table says it means. NOTHING HERE rows tell you what you can skip entirely this week. The list at the end is where the pile was trying to get your attention about something you never asked. This works the same way in Claude, ChatGPT, and Gemini, with nothing to enable.

Keep the decisions list in the project or Gem where your other standing context lives, and update it as decisions close. The same prompt then runs on whatever lands next week, and the list itself becomes a running record of what you were deciding and when.

Most of the effort people put into AI goes toward getting the right answer: the right details, the right numbers, the right depth. That effort pays off. The step most people miss comes after, when the answer arrives in a vocabulary nobody in your company actually uses.

Every business has its own dialect. What one company calls a pipeline review, another calls the forecast call, and the model has no way of knowing which one you are. It defaults to the internet's words, so what comes back is accurate and still needs a translation pass before you can repeat it in a meeting without sounding like you read it off a vendor site.

It's like getting directions from someone who only knows street names in a town where everyone navigates by landmarks. Correct, and you still have to convert it before it does you any good.

The fix is small. Tell the model what you call things: the names of your teams, the words you use for your customers, the phrases that show up in your Monday meeting. Put it in the project instructions once and it holds. Training it to be right is one step. Training it to talk the way your office talks is the step that turns an answer into something you can say out loud in the hallway.

Microsoft now sells software to watch the agents you bought from someone else, OpenAI logged four more cases of its own models wandering past their instructions, and Nvidia will hand you a capable model to run inside your own walls for free. The intelligence is settled and priced. What most companies have still to build is the part around it: what the model may touch, what it is for, and what the company knows that the model does not.

That part is the company AI brain an AI-native business owns: its own data organized so tools can reach it safely, its judgment written down where the model can use it, permissions set before anything runs, and a team that reviews the work rather than retyping it. Kiingo AI builds it with you as one roadmap, sequenced for your situation.

Speak with us →

Forwarded this by a colleague? Subscribe to Kiingo AI Weekly and get it every week.