I Know, You Know Me for ERP. But Let's Talk About the Thing That's Eating Every Industry Alive.

Hello readers.

You know me as the person who gets genuinely excited about D365 Finance, CE, Power Automate, Azure integrations, and the kind of transformation work that makes ERPs actually sing. And I'll keep writing about all of that - that's not going anywhere.

But today I want to step back from the D365 Finance for a minute and talk about something bigger. Something that's quietly reshaping the world that ERP sits inside of.

AI.. The real, already-here, already-affecting-your-job version.

 

For years, AI was the thing on the horizon. The thing keynotes were built around. The thing your LinkedIn feed was full of opinions about but nobody had actually touched.

And then somewhere between 2023 and now, it stopped being a horizon and became the ground we're standing on.

Every tool I work with daily has changed. D365 Finance now surfaces anomalies I used to hunt for manually. Copilot in Power Automate drafts flows I used to build from scratch. Azure OpenAI is sitting inside enterprise architectures that two years ago had none of this. Logic Apps are now orchestrating intelligent document processing that used to need a team. Dynamics CE is feeding AI-driven sales insights that no human could have assembled at that speed.

This isn't a future roadmap slide. This is a Tuesday morning now.

 

But First - What Kind of AI Are We Even Talking About?

Because "AI" has become one of those words that means everything and nothing at the same time. And if we don't get clear on this, we end up having ten different conversations under one label.

Generative AI is the one everyone's heard of - ChatGPT, Copilot, Claude. It writes, drafts, summarises, codes. Powerful. Also very easy to misuse if there's no brain steering it.

Predictive AI has been quietly running inside enterprise tools for years. Your ERP flagging a duplicate invoice. Your CRM scoring a lead. Your supply chain tool whispering "you're going to run out of stock in 11 days." You've been using AI longer than you thought - just nobody put a label on it.

Intelligent Automation is what happens when you combine process automation with AI judgment. Think Power Automate reading an unstructured email, extracting the key data, routing it to the right approval workflow - without a human touching it.

Agentic AI is the frontier - AI that doesn't just respond, it acts. Give it a goal, it breaks it into steps, executes them, reports back. We're early here but it's moving fast. And in the D365/Azure ecosystem, the infrastructure for this is already being laid.

Here's what matters: each of these needs a different kind of trust, oversight, and skill to use well. Treating them all the same is like having one driving policy for a bicycle, a car, and a plane. Same word - vehicle - wildly different responsibility.

 

Leadership Has Entered the Chat

Let me tell you something I've observed sitting in steering committees and boardrooms across multiple transformation programmes.

There are two types of leaders in the room right now.

The first type nods when AI comes up, says something like "yes, absolutely critical - we're keeping a close eye on it" and then goes back to making decisions exactly the way they did four years ago. They're not bad leaders. They're just standing still while everything moves.

The second type got personally uncomfortable, personally curious, and personally involved. They started using the tools themselves - not having someone demo it for them, actually using it. They broke things. They were surprised. And now they have something no PowerPoint can give you: instinct. They know what AI can do, what it can't, and where their own thirty years of judgment is irreplaceable.

I know a Senior Partner at a consulting firm - sharp, experienced, the kind of person clients pay serious money to simply be in the room. Last year one of his junior analysts, two years out of university, delivered a strategic options paper in an afternoon that would have taken his team a week in 2021. He wasn't angry at her. He was unsettled. Because he didn't know how she did it.

He called me. "Is this what it felt like when spreadsheets replaced the ledger clerks?"

Yes. Except spreadsheets took twenty years to fully land. This is taking twenty months.

Three months later, I saw him present to a client. He'd clearly used AI in his preparation -the data was richer, the scenarios sharper, the analysis faster. But what was also there - what the junior analyst will need twenty years to develop - was the read on the room. The political instinct. The knowledge of which risk the CFO would actually lose sleep over versus which one they'd wave away.

AI gave him his time back. He spent it on the things only he could do.

That's the whole story. Everything else is commentary.

 

Operations, Discipline, and the People Nobody Writes About

Everyone writes about AI in strategy and leadership. Almost nobody writes about what's happening in the operational engine room - which is honestly where the most significant change is occurring right now.

Think about a mid-size company's accounts payable function. Invoices come in, get matched against POs, exceptions queue up, humans review them, payments go out. Unglamorous. Essential. In most organisations this involves a small team doing a lot of mechanical work with occasional bursts of actual judgment.

AI has restructured this. The matching is smarter. The exception queue is now prioritised by risk rather than arrival order. The anomalies that actually need human eyes float to the top. The team now spends its time on real decisions instead of data entry.

But here's what made it work in the organisations that got it right - they didn't impose it top-down. They sat with the people who actually do the work and asked: where does the judgment live in your day? What's mechanical and what isn't? The people doing the job knew exactly. Management listened. That's a discipline and leadership story as much as it's a technology story.

The same pattern plays out in procurement, project delivery, HR, compliance, finance reporting. The mechanical layer compresses. The judgment layer matters more. The organisations that understand this distinction will use it to do more with the same people. The ones that don't will cut headcount, save money once, and destroy trust for a decade.

 

Trust, Ownership, Accountability, Governance - The Four Things That Decide Whether AI Works or Blows Up

I've watched AI initiatives fail. Not because the technology was bad. Because the organisation around it wasn't ready.

Here's how it typically goes: a few enthusiastic people start using AI tools. Then someone pastes confidential client data into a public model without thinking. Legal finds out. Blanket ban. Eighteen months of momentum erased overnight.

Or: an AI-generated output goes to a client without proper review. Something is wrong in it. The client loses confidence. The firm spends six months rebuilding trust that took six years to build.

None of this is the AI's fault. It's a governance failure.

Ownership - someone has to own AI adoption. A real person, with real accountability. Not a committee. Not a shared responsibility that nobody actually holds. One person who sets the direction, makes the calls, and is answerable when things go sideways.

Trust - it has to be earned through visible wins, not assumed because leadership said so. Find a use case that's high effort, low risk. Show people what changes. Let them feel it. "Remember when month-end variance commentary used to take two days?" Trust is built through experience, not announcements.

Accountability - if an AI tool produces an output that influences a decision, a human owns it. Full stop. "The AI said so" is not a defence. Ever. This sounds obvious until you watch it break down in practice - people start treating AI output as fact rather than a draft that needs their brain on it.

Governance - which tools are approved? What data can go into them? What can't? What's the review process before something goes out the door? Not bureaucracy — structure. The difference between AI being an asset and a liability often comes down to whether someone wrote these things down.

 

This Is Not a Programmers-Only Conversation. I Cannot Stress This Enough.

I'll say it plainly because I think it needs saying plainly.

If you are a finance manager, a project lead, a procurement officer, an HR director, a consulting principal, a COO, a CEO - AI is your conversation. Not your IT team's. Not the data science team's. Yours.

The junior analyst who outpaced that Senior Partner didn't have a computer science degree. She just started using the tools. Seriously. That's the gap right now - not technical knowledge, willingness to engage.

I've watched a 57-year-old CFO become sharper with AI tools than most of the graduates joining her team. I've watched a 29-year-old consultant who refuses to touch any of it already starting to look slow compared to his peers. The deciding variable is not age, experience level, or technical background.

It's curiosity. And the willingness to stay a student even when you've earned the right to feel like an expert.

Whether you're a fresher figuring out your first role, a mid-career professional wondering which way to lean, or a senior leader who hasn't had to learn something genuinely new in years -this is for you. All of you. At every level, in every function, the question isn't whether AI is going to affect your work. It already is. The question is whether you're going to understand it well enough to shape how.

 

I'm not going anywhere near the "AI will take all our jobs" conversation today -that deserves its own post and a lot more nuance than a passing line. But what I will say is this:

The people I see thriving right now are not the ones who have the most technical knowledge. They're the ones who have strong judgment, deep experience in their domain, and enough familiarity with the tools to point them in the right direction. That combination is genuinely rare. And if you're building it deliberately, you're building something that matters.

So. That's where I wanted to start. 

Next time: I'm going to slow right down and take you inside the actual mechanics - what a token is, what a vector does, why "model" means something very specific, and how these building blocks connect to the AI you're already using. 

See you there.