AI news & training for business owners & operators. One email. One clear next step.

THIS WEEK IN AI

This was the week the AI race stopped being about chatbots. Google launched a personal AI agent that runs 24/7 in the background. Meta moved 7,000 employees into AI roles and laid off 8,000 more on the same day. And OpenAI partnered with Dell to put AI inside the buildings where regulated businesses keep their data. The common thread: AI is moving from a tool you open to a system that runs in the background. That changes what you need to plan for.

What happened:

At Google I/O on May 19, Google announced Gemini Spark, the first persistent, cloud-based AI agent from a major platform that runs 24/7, accesses your apps, and takes actions on your behalf. Spark runs on Google Cloud virtual machines so it works in the background on your phone or laptop even while they're turned off, and is designed to check with you before taking major actions on your behalf. Google also released Gemini 3.5 Flash and dropped the AI Ultra plan to $100 a month, half its prior price.

Why it matters to your business:

This is the first AI tool from a major platform that does not require you to open an app and type a prompt. It runs on its own. For a small business, that sounds useful and is also the moment to slow down. An agent that can act on its own is only as safe as the workflow you point it at. If your process is messy, the agent will do messy work faster. Fix the workflow first. Then let the agent run it.

Source: blog.google/innovation-and-ai

What happened:

On Monday May 18, Meta announced it is reassigning 7,000 employees to focus on AI initiatives, just two days before the company plans to lay off roughly 10 percent of its workers. In an internal memo, Meta Chief People Officer Janelle Gale said the company plans to transfer roughly 7,000 employees into new AI initiatives while eliminating layers of management and flattening organizational structures. Workers reportedly refer to the reassignment process as being "drafted," and the changes will affect roughly 20 percent of Meta's approximately 78,000 employees.

Why it matters to your business:

Meta is doing in public what a lot of mid-sized businesses are about to do quietly. They are flattening management layers, closing open roles, and moving people into AI work instead of hiring outside. If you have a hiring plan for the back half of the year, this is the pattern to study. The question is not "should I replace people with AI." It is "which open roles on my desk right now should not be filled the way I planned to fill them."

Source: nbcnews.com/tech

What happened:

On May 18, OpenAI and Dell Technologies announced a partnership to help more enterprises deploy Codex in the environments where their most important data, systems, and workflows already live. Codex is becoming one of OpenAI's fastest-growing enterprise products, with more than 4 million developers using it every week across the software development lifecycle.

Why it matters to your business:

Most small businesses do not run on-premise servers, so the headline is not for you. But the signal underneath it is. AI is moving into places that have hard rules about where data can go. Healthcare practices, law firms, accounting firms, anyone who deals with regulated data. If you have been told "we can't use AI because of compliance," that excuse has a shorter shelf life every month. Start asking your IT person what would actually need to be true for AI to be allowed.

Source: openai.com

The word of the week is agentic. Google announced one. Meta is restructuring around them. OpenAI sells them. You will hear it again next week, and the week after.

Most coverage treats agentic AI like the next ChatGPT moment. A new toy. Try it out. See what it does.

That framing is wrong. Agentic AI is not a new chatbot. It is a different category of software. The difference matters because it changes what can go wrong.

A chatbot answers when you ask. An agent acts when you do not.

Here is the practical version. Right now, when you use ChatGPT or Claude or Gemini, the loop is simple. You type a prompt. The model responds. You read it. You decide what to do. You are in the loop on every step.

An agent removes you from that loop. You give it a goal. It picks the steps. It runs the steps. It might check with you before something big. It might not. Spark operates autonomously, under your direction, and is designed to check with you before taking major actions. The keyword is "major." Everything else, it just does.

That is fine when the workflow is clean. It is dangerous when the workflow is broken.

Last month I sat with an operations manager at a 40-person services company. She wanted to add AI to their proposal process. The process looked like this on paper: pricing team builds a quote, sales reviews, owner approves, proposal goes out. In reality, the pricing team was guessing at numbers because the cost data lived in three spreadsheets that nobody trusted. Sales was "reviewing" by skimming. The owner was approving because someone had to. The proposal went out with errors about half the time, and the errors got fixed in revision rounds with the client.

If you pointed an AI agent at that workflow and told it "send out proposals automatically," it would do exactly what the humans were doing. It would guess at pricing. It would skip the review. It would send out errors. Faster.

That is the actual risk for a small business this year. Not that AI will replace your team. That you will put an agent on top of a process that does not work, and the agent will scale the broken parts.

So here is the question to sit with before the agent wave hits your inbox. If the work in your business has a person sitting in the middle making sure it does not go off the rails, what is that person actually fixing? Bad data. Missing steps. Unclear ownership. Old policies. Whatever they are catching, that is what the agent will miss.

The good news. You have a few months. Spark is rolling out to Ultra subscribers first. Most agent tools for SMBs are still in beta or coming soon. The window between "agents are a press release" and "agents are running real work in your business" is short, but it is not zero.

Use the window. Fix the work first.

This week's move is a framework, not a prompt. It is one question to ask in your next ops meeting.

Pick one process in your business that runs every week. Then ask: "If we took the person out of the middle, what would break first?"

Pick something real. Invoicing. Onboarding. Quote generation. Lead follow-up. Inventory reorders. Whatever has a human in the middle making sure it does not fall apart.

Then walk through it out loud with whoever owns the process. Not "how would AI help here." That is the wrong question. The right question is what the person is actually doing when they sit in the middle.

You are looking for one of three answers:

If it is the first one, that is a data problem. AI will not fix it. If it is the second one, an agent might actually help once you map the connection. If it is the third one, you need to write down the judgment call before any AI can do it.

Do this once this week. One process. One conversation. Twenty minutes. You will know more about where AI fits in your business than ninety percent of the businesses chasing the agent wave.

Quick note before the sign-off. This week's offer is for home service businesses specifically. HVAC, plumbing, electrical, roofing, anyone whose phone is the front door of the business. If that's not you, forward this to someone it is.

Introducing the Missed Call Recovery System

If you run a home service business, you already know this feeling. The phone rings while you're 20 feet up a ladder. By the time you climb down, the call's gone. No voicemail. They called the next shop.

Per Invoca's research on the home services industry, 27% of inbound calls to home service businesses go unanswered. For an HVAC shop, every missed call could be a $30,000 install. As Tyson Chen, co-founder of Avoca, put it in Fortune earlier this year: "When a home service business misses a phone call, that could be a $30,000, $40,000 HVAC install they're missing."

The Missed Call Recovery System catches those calls before they turn into lost jobs.

Here's how it works. When a customer calls and you can't answer, an AI voice receptionist picks up. It has a real conversation with the caller, asks about the issue, captures their name, address, urgency level, and what they need. The information lands in your CRM or calendar within seconds. If the caller doesn't want to talk to an automated system, they get a text back within 30 seconds asking what's going on. Every Monday morning, you get a one-page report showing exactly how many calls came in, how many got recovered, and what those recovered jobs were worth.

You keep your existing phone number. No software for your team to learn. The system runs in the background and feeds you booked jobs.

Want to see what you're losing before deciding anything?

Book a free 30-minute Lost Revenue Check at THIS LINK. I'll walk through your call volume, your average job value, and your current process, and you'll walk away with a one-page scorecard showing exactly how much money is on the table. No pitch. Just the math.

Thanks for reading,

Andrew Keener

Operations & AI Strategist

Keen Alliance Consulting