top of page

Is Your Business Ready for AI Agents?

18 hours ago
5 min read

Genna Barbara Zimmel is an AI implementation strategist, Technical Project Manager, and founder of Torus Solutions, a company that builds intelligent AI systems, CRM automation, and conversational AI products. She is also the creator of Sibyl, an AI companion designed to enhance human self-understanding and relationships.

Executive Contributor Genna Barbara Zimmel Brainz Magazine

AI agents don't arrive with good judgment built in. They work with the information you give them and the access you hand over. Before you launch one, check your data and your doors. AI agents are about to become a normal part of work. Gartner predicts that by 2028, the average global Fortune 500 company will have more than 150,000 agents in use, up from fewer than 15 in 2025. Yet only 13% of organizations believe they have the right governance in place.


Vintage blue car beside a silver UFO-like pod in an open carport, warm sunset light; torussolutions.io watermark.

That gap matters for every business, not just the giants. Whether you run a team of five or five thousand, you'll soon be asked, should we let an AI agent handle this? Before you answer, it helps to understand what an agent actually is.


An AI agent is a new team member with no history


A chatbot answers questions. An agent takes actions. It can update a customer record, send an email, book a meeting, move a deal forward, or process a request, often without anyone clicking "approve."


Think of it as a new employee who starts work instantly, never sleeps, and has no memory of how your business really works. It knows only two things.


  1. The information you give it (your data).

  2. The doors you open for it (your access).


In more than a decade of building and running business systems, from established customer relationship management (CRM) platforms to conversational AI built from scratch, I've seen that when agent projects stall or go wrong, the cause is more often one of these two than the AI itself.


Your data: The agent believes everything it reads


If your customer records are full of duplicates, outdated details, or fields nobody maintains, a person learns to work around them. An agent won't. It treats every record as true and acts on it, faster and at a larger scale than any person could.


That's why so many AI projects never deliver. Gartner predicted that through 2026, organizations will abandon 60% of AI projects that aren't supported by AI-ready data.


The problem is rarely dramatic. It looks like three teams using different definitions of a "qualified lead" or customer notes that haven't been updated in two years. An agent doesn't fix that confusion. It automates it.


Your doors: The agent can go wherever you let it


The second risk is access. When you connect an agent to your systems, you're handing it a set of keys. Most businesses hand over too many.


Real examples show why this matters. In 2025, security researchers demonstrated that hidden instructions placed in an ordinary lead form could trick a CRM's AI agent into sending customer data to an outside website. The agent wasn't "evil." It simply followed instructions it should never have trusted, using access it should never have had.


Many companies can't even see what their agents are doing. In a 2026 survey of 157 sales and marketing leaders by LeanData, a software vendor, about a third couldn't say how many agents touch their customer records.


The rules are arriving, and they're about how AI is used


Earlier this year, Israel's Supreme Court ruled on a case that should worry every business using AI. A city department denied transport to a child with a disability, citing an official government circular that didn't exist. The AI tool used to draft the response had invented it.


The court was clear. Using AI wasn't the problem. Using it without oversight, verification, and validation was.


Around the world, the rules are arriving piece by piece, and almost none of them is one big "AI law." China became the first country to regulate AI agents as their own category this July. Israel and Canada are strengthening accountability through privacy law. The UK is preparing for incidents involving AI agents, and the UAE has created a federal authority for AI and data.


The approaches differ, but they point the same way. Responsibility is moving toward the businesses that use AI, not just the companies that build it. When I built Sibyl, a conversational AI, I applied the same principle. A separate safety layer runs before the model responds, so protection doesn't depend on the model alone.


Start with the workflow, not the vendor


Much of the conversation about AI agents is shaped by whichever platform a company already pays for. That's understandable, but it's backward. The right starting point is the work itself, what needs to happen, who is responsible, and where things slow down.


Sometimes the best answer is to build on the systems you already have. Sometimes it's to build something new from scratch, designed around how your team actually works. I've done both, and the lesson is the same either way. The tool matters less than the foundation underneath it.


6 checks before you launch an AI agent


You don't need to be technical to lead this well. Ask these questions before any agent goes live.


  1. What one job is it doing? Start with a single task, and measure how it works today so you can prove the agent improved it.

  2. Is the data it will use clean? Look at the actual records it will read. Remove duplicates, update stale information, and give every important field an owner.

  3. Does it have only the access it needs? Give the agent its own identity and start with no permissions, then add only what the job requires.

  4. Can it act without a human? Let it read and suggest freely, but require a person to approve anything that sends, deletes, or changes records in bulk.

  5. What could trick it? Treat any outside text, such as web forms, emails, and uploaded files, as untrusted. These can carry hidden instructions.

  6. Who is watching? Name the person who gets alerted when something looks wrong and who has the authority to pause the agent.


If you can't answer these yet, that's your starting point, not your launch date.


Build the foundation, then add intelligence


In my recent Brainz interview, I described what I call the Implementation Gap, the distance between having access to AI and being able to use it reliably in real work. AI agents make that gap bigger because now the system doesn't just advise. It acts.


My prediction is that over the next year, the agent projects that succeed will be the ones that started with a data clean-up and an access review, not a flashy demo.


The question isn't whether AI agents are coming to your business. It's whether your data and your doors are ready when they arrive.


Find out where your business stands



Follow me on LinkedIn and visit my website for more info!

Read more from Genna Barbara Zimmel

Genna Barbara Zimmel, Founder of Torus Solutions

Genna Zimmel is the founder of Torus Solutions and creator of Sibyl, an AI companion built on the principle of Support Over Dependency. She designs and implements conversational AI systems, CRM automation, and multi-agent workflows for organizations navigating the next generation of AI adoption. Genna builds in public, documents the real work behind intelligent systems, and is currently completing formal training in Cloud Computing and Python for AI.

References:


Tags:

 
 

This article is published in collaboration with Brainz Magazine’s network of global experts, carefully selected to share real, valuable insights.

Article Image

The Invisible Ceiling and the Beliefs Quietly Running Your Business

There is an old story about a baby elephant tied to a small wooden stake. Every day it pulls against the rope, and every day the rope holds. One day, it stops pulling. Years later, that elephant weighs...

Article Image

Pre-Builder or In-Seat Builder? How to Build Your 90-Day Plan for a New Leadership Role

Some leaders have a 90-day plan drafted before the promotion is even announced. Others won't write a word of it until they're in the chair and can see what they've inherited. Every time I describe...

Article Image

Spiritual Intelligence vs. Emotional Intelligence – Why Emotional Capacity Matters

I have spent much of my life paying attention to what sits beneath the surface: the feeling in a room, the shifts in someone’s energy, the things people communicate without saying them out loud. I have had powerful...

Article Image

Closer to Hell on Earth and What an Ordinary Office Afternoon Costs a Neurodivergent Mind

It is mid-afternoon. You forgot to have lunch because you were too busy working on that presentation. You do not eat breakfast, so it has been more than 16 hours since the last time you ate. “Maybe...

Article Image

The Erosion of Values Through Scale, Speed, and Automation

Modern organisations have long celebrated speed: faster delivery, faster decisions, faster innovation, faster customer response, and faster transformation. For decades, technological progress has been closely...

Article Image

I Am, Therefore I Think – Reconsidering Descartes

A woman I will call Anna sits across from me. She is a composite of many people I have worked with over four decades, and her words will be familiar to anyone who has ever struggled. She is accomplished...

The Invisible Ceiling and the Beliefs Quietly Running Your Business

Pre-Builder or In-Seat Builder? How to Build Your 90-Day Plan for a New Leadership Role

Spiritual Intelligence vs. Emotional Intelligence – Why Emotional Capacity Matters

Closer to Hell on Earth and What an Ordinary Office Afternoon Costs a Neurodivergent Mind

The Erosion of Values Through Scale, Speed, and Automation

I Am, Therefore I Think – Reconsidering Descartes

I Stopped Chasing the Number and Started Studying What Created It

You Call It Incompatibility, Your Nervous System Calls It a Trauma Response

Beyond Toxic or Supportive and Understanding the People We Turn To

bottom of page