There’s still one more day to register.

Your most valuable data lives inside the tools you use every day — your CRM, email, shared drives, and calendars — but most AI can’t see any of it. In this free, practical session, Ross Hartmann (Founder & CEO, Kiingo AI) demos how to connect AI directly to those systems so it answers with your real data and runs real tasks across them, live. No coding required.

Tuesday, June 30, 2026 · 1:30 PM ET / 12:30 PM CT / 10:30 AM PT · 60-minute live session with a live demo.
Save your spot →

• OpenAI previewed GPT-5.6 in three priced tiers — but you can’t use it yet. The model splits into Sol (top-end, $5/$30 per million tokens in/out), Terra (balanced, $2.50/$15 and about 2x cheaper than GPT-5.5 at similar quality), and Luna (cheap, high-volume, $1/$6), so teams can match price to task instead of paying premium rates for routine work. Initial access is limited to roughly 20 vetted partners at the U.S. government’s request, with general availability promised “in the coming weeks.” (OpenAI, CNBC)

• The U.S. partially lifted its ban on Anthropic’s top models. After a June 12 export-control suspension of Fable 5 and Mythos 5, the Commerce Department cleared the cybersecurity-focused Mythos 5 for about 100 critical-infrastructure organizations and federal agencies on June 27 — but most Pro, Team, Enterprise, and API customers still need an export license, and Fable 5 stays dark. Anthropic spent the same stretch opening a Seoul office and putting Samsung, LG, and NAVER on Claude, while Austria reportedly began lobbying the EU to host Anthropic’s models on European soil. The expansion continued on the talent side: Nobel laureate John Jumper, who shared the 2024 Chemistry prize for the AlphaFold protein-folding system, left Google DeepMind after nearly nine years to join Anthropic’s life-sciences push. (CNBC, Anthropic, Bloomberg via WHBL)

• “Tokenmaxxing” is over — companies are reining in AI spend. CNBC reports Uber capped employee AI spend at $1,500 a month after burning its annual budget in four months, and startup Lindy moved its traffic off Claude to cheaper DeepSeek models. The emerging playbook is “model routing”: send routine tasks to cheap models and reserve expensive frontier models for the hard subset. OpenAI, meanwhile, rolled out usage analytics and spend controls for enterprise admins. (CNBC, OpenAI)

• Samsung Electronics rolled out ChatGPT Enterprise and Codex to its workforce. It’s a large, named, company-wide deployment rather than a pilot — and a sharp reversal for a company that banned generative AI internally three years ago. (OpenAI, UPI)

• AI pushed deeper into the tools teams already work in. Microsoft’s “Copilot Cowork” agent — which runs longer, multi-step jobs in a secure cloud environment so work continues even when your laptop is off — hit general availability, and Copilot Chat now lets users pick Anthropic’s Claude alongside OpenAI’s models. (Microsoft, Redmondmag)

• Google gave Gemini a hand on the keyboard. Its Gemini 3.5 Flash model added built-in “computer use,” meaning the AI can see a screen and operate a browser, mobile app, or desktop on its own — roughly on par with the best from OpenAI on independent agent tests. The pitch is agents that finish multi-step software tasks without someone babysitting every click. (Google, The Next Web)

• OpenAI and Broadcom unveiled “Jalapeño,” a custom chip built to run AI models more cheaply. With first deployment targeted for late 2026, it’s another move to cut per-use AI costs and reduce the industry’s dependence on Nvidia. (OpenAI, TechCrunch, CNBC)

• Meta is building “Arena,” an AI-powered prediction-market app. Users would forecast real-world events (sports, politics, entertainment) with a daily allotment of play money, with AI generating questions from trending topics. The report knocked shares of DraftKings and Robinhood; Meta hasn’t ruled out real-money betting later. (NYT, CNBC)

• Anthropic’s sixth Economic Index, pairing a survey of roughly 9,700 workers with their real Claude usage, found workplace AI adoption expectations climbing fast: more than a third expect AI to handle most or nearly all of their tasks within twelve months, and close to 6 in 10 put it in a higher band for next year than for today. (Anthropic Economic Index, June 2026)

• McKinsey found only about 6% of organizations are “high performers” linking 5% or more of profit to AI — and what sets them apart is operational: redesigning workflows, embedding AI into core processes, tracking KPIs, and committed senior leadership. (McKinsey, via The Next Web, May 2026)

• PwC’s 2026 Global AI Jobs Barometer, drawn from more than a billion job ads across 27 countries, found the wage premium for workers with AI skills climbed to 62%, up from 57% a year earlier. Jobs requiring specific AI skills grew roughly eight times faster than the overall job market. (PwC 2026 Global AI Jobs Barometer, June 2026)

Every ops team runs on processes that work beautifully — until reality stops cooperating. The SOP assumes the file arrives on time, the vendor confirms, the form gets filled in correctly. Then one input shows up late or wrong and the whole thing quietly stalls, usually noticed three steps too late. This prompt takes a process you already run and pressure-tests it against the messy real world: where it breaks, what it silently assumes, and the exact exception-handling steps to add so it survives a bad Tuesday.

The Break Points
Here’s a process my team runs regularly: [paste the SOP, checklist, or just describe the steps in order].

Pressure-test it against reality. Walk each step and identify: the unstated assumptions it depends on (an input arriving on time, a field filled in correctly, a person being available); the specific points where it silently breaks if an input is late, missing, wrong, or ambiguous; and which break would do the most damage before anyone notices. Then write the missing “if this goes wrong, do this” branches needed to make it hold up.

Return as: a list of hidden assumptions, the top three break points ranked by damage, the single failure most likely to go unnoticed, and the exact exception-handling steps to add to the process.

Run this on the process that burned you last month — the shipment that slipped, the onboarding that stalled, the report that went out wrong. The goal is a process that doesn’t depend on everything going right. Most ops failures are the quiet ones: a happy-path procedure meeting a normal-but-unplanned-for day.

I took a few days off before the weekend, and then Monday rolled around with that familiar weight to it. The inbox, the backlog, the “what did I miss” feeling that usually eats the whole first day back.

This time it didn’t. A few months ago I started setting up small AI workflows for the boring parts of getting up to speed: catch me up on this thread, summarize what changed in this doc, pull the three things I actually need to act on out of forty emails. Nothing fancy. But this morning all of it was just sitting there, ready, and the day I’d braced for turned into a couple of hours.

The funny part is I didn’t set any of this up for coming back from time off. I built it for normal weeks. The few days away just made it obvious how much it had been quietly doing all along.

The work you put into a workflow pays you back loudest on the day you’re furthest behind.

Most companies don’t have an AI problem. They have a “we’ve talked about it for six months” problem. The decks are sharp, the intentions are real, and somehow the work still gets done the same way it did last year.

We help mid-market teams close that gap — finding the two or three places where AI earns its keep, and actually building them in.

Want to see what that looks like for your team? Get in touch.