Where could AI make a difference in your business?

This is where things get more technical. Business applications can communicate through technologies such as APIs, connectors and automation platforms. These effectively provide controlled ways for one system to request information or perform an action in another. AI can sit across those connections.

For example, an AI application could potentially interact with your:

Microsoft 365 > SharePoint > CRM > finance software > databases > service desk > other business applications

Exactly what’s possible depends on the systems you’re using and what integrations they support.

AI becomes considerably more useful when it can work with the technology your business already uses.

An AI Agent takes the idea another step. Rather than simply asking:

“What’s in this document?”

you might give an agent an objective:

“Prepare me for my meeting with this customer.”

With the appropriate access, an agent could potentially find the customer in your CRM, review previous communications, retrieve relevant documents, identify outstanding actions and produce a meeting briefing.

The difference is subtle but important.

Traditional AI: Answer this question.

AI Agent: Help achieve this objective.

Potentially, yes. Imagine telling an AI system:

“Follow up our outstanding sales opportunities.”

Depending on how it had been configured, it could potentially:

Check CRM > identify opportunities > review communications > prepare follow-ups > create tasks > update records

Some steps might happen automatically. Others might deliberately require somebody to approve them.

And that’s an important distinction. Automation doesn’t have to mean removing the human. Sometimes the best workflow is AI doing 80% of the preparation and a person making the final decision.

For many of the more useful business applications, yes. A public AI model doesn’t automatically understand your customers, processes, projects or internal knowledge. Giving an AI system controlled access to relevant business information can make its responses significantly more useful.

But that introduces an important question:

What should AI actually be allowed to see?

This is where identity, permissions and data governance become extremely important. If someone shouldn’t have access to sensitive HR information, financial documents or confidential customer data, an AI assistant working on their behalf shouldn’t suddenly give them access either.

Before connecting AI deeply into a business, organisations need to understand: Who has access to what? Where is sensitive information stored? Are permissions correct? Which information can AI use? Which actions can it perform?

Giving AI intelligence is one thing. Giving it access is another.

AI systems can make mistakes. That’s why the level of autonomy should reflect the potential consequence of an action.

Having AI draft an internal meeting summary presents a very different level of risk from allowing it to approve a payment, delete information or communicate with a customer automatically. Businesses need to decide where AI can act independently and where human approval remains part of the process.

Good AI adoption isn’t simply about asking: “Can we automate this?” it’s also: “Should we?”

Probably not by buying an AI product. Start with your people and processes.

• Where does your team repeatedly lose time?
• Where is information difficult to find?
• What gets copied between systems?
• Which administrative tasks happen every day?
• Where are people manually checking, sorting, categorising or summarising information?

Those are potentially much more useful questions. Then you can begin exploring whether AI, automation or better integration could help.

This might ultimately be the biggest question. AI doesn’t exist separately from the rest of your technology. But instead, relies on the systems underneath it. Data. Identity. Permissions. Security. Integrations. Applications. Governance.

If those foundations aren’t right, introducing more powerful AI won’t necessarily solve the problem. It could simply make an existing problem move faster. AI readiness isn’t just about AI. It’s about the technology underneath it.