AI Builder - Simply Explained

0:00 The AI everyone keeps telling you to learn is already hiding inside the tools you open every morning. That's AI builder and it lives inside power Apps and Power Automate, the low code tools you're probably already using. You don't need to write code, and you don't need a data science team to put it to work. By the end of this, you'll know what AI builder actually is, what it can do with your documents, and how you pay for it. Welcome to the M365 FM podcast, where we explore the people, ideas, and technologies shaping the future of work. And by the time we're done, the jargon will be gone and the whole thing will feel a lot smaller than it sounds. What AI builder actually is AI builder is a feature inside Microsoft Power Platform. It lets you create and use AI models to take over business tasks you'd normally do by hand. The word model is the one that scares people, so it helps to shrink it down to size.

0:43 A model is a pattern finder. You show it a pile of examples and it learns what tends to go with what. Then it applies that lesson to something new and that is the whole trick. The reason beginners freeze up is rarely the model itself. It's the story they tell themselves about what using AI requires. They picture a team of engineers, a data warehouse, and six months of setup before anything works. That picture is what makes AI feel out of reach. AI builder is designed to skip every one of those steps. Microsoft's own recipe is for moves. You pick a model type, you connect your data, the training happens on its own, and then you use the result. There's no coding, no data science degree, and no secret handshake. You also don't have to go find a separate tool to use it. AI builder is built straight into Power Apps and Power Automate, which are the same low code platforms.

1:25 You're likely already learning that placement matters more than it sounds, because it means the AI shows up exactly where the work already happens. It sits inside the app, your team fills out or inside the flow that moves your files around at night. You're not learning a second career. You're adding one new capability to something you already touched. There's a reason Microsoft built it this way. Most AI headlines are about giant models that answer questions in a chat box. Those are impressive, but they're hard to point at one specific business chore. AI builder flips that around. It starts from the task instead of the chat. So you begin with a job that needs doing and then pick the model that does it. That's why a beginner can use it without ever training a network by hand. So keep this picture in mind. AI builder is AI. You can actually use sitting quietly inside tools you already know.

2:07 And everything else in this video hangs off that one idea. Ready made models versus models you build. Once you accept that this is real and reachable, the next question is which kind of model you actually reach for. Microsoft gives you two paths, and the difference comes down to whose data does the learning. The first path is pre-built models. These are models Microsoft already trained for problems that show up in almost every business, so you skip straight to using them point one at an invoice and it pulls out the totals and the dates. Point one at a picture of a receipt and it reads the line items. There are pre-built models for reading text inside an image for judging whether feedback sounds positive or negative, and for pulling names and numbers out of messy text. Others read ID documents. Grab the key phrases from a paragraph, scan a business card, or detect what language something is written in.

2:47 You didn't train any of them and you didn't have to. The second path is custom models, and this is where your own data does the work. These models are trained on your documents, your text, your images, and your history, which makes them fit the way your business actually looks. That list includes document processing, which reads your specific forms and category classification, which sorts text into categories you define. It also includes entity extraction, which finds the exact details you care about, prediction, which guesses future outcomes from your past data, and object detection, which spots your products in a photo. There's even a way to bring your own model over from Azure if you already built one. So the rule of thumb is simple if the problem is common and generic, like reading any receipt that lands in your inbox, a pre-built model has you covered on day one when the problem is uniquely yours.

3:32 Like reading invoices from one specific supplier. You train a custom model on your own examples, and it learns your quirks. The difference is easier to deal with. An example. A pre-built model can read the total of a standard invoice from almost any vendor because Microsoft trained it on thousands of them. But if your invoices come in a layout nobody else uses, with the account number buried inside a table, that pre-built model will miss it. That is exactly the job you hand to a custom model. Most beginners assume custom always means better, and that assumption is wrong. Custom means more specific, and more specific is only better when your problem is specific to. One more thing before you pick. A few of these models are still in preview, which means they're not meant for production work yet, and some aren't available in every region. So if you go looking for a model and can't find it, that's usually why.

4:14 Why this is not just another chatbot. Before we go further, there's a confusion worth clearing up because it trips up almost every beginner. When people hear AI, they now think of a chat window that answers questions and they assume AI builder is the same thing wearing a Microsoft badge. That assumption is wrong, and the difference is the whole reason it's useful. A chatbot is general. It will talk about anything, which is impressive, but it's also vague because it has no idea what your specific job looks like. AI builder takes the opposite approach. It starts with a task like reading an invoice or scoring a record, and it brings a model that's focused on exactly that. You're not asking it to be clever about everything. You're asking it to be reliable about one thing over and over without getting bored. That focus is why the results are easier to trust.

4:55 A focused model does the same job the same way every time, which is what you want when the output feeds a business process. It's also why AI builder fits inside power apps and Power Automate. So naturally, those tools are already about doing a defined job well. So a model that does one thing well slides right in. Now AI builder does include prompts, which are the more conversational generative side of things. So the line isn't perfectly clean. When people talk about AI builder for beginners though, they usually mean the models, the focused pattern finders. And that's where the value shows up fastest. What document processing really does document processing is the one people ask about most, because it attacks a job that quietly eats hours every single week. In plain terms, it reads and saves the information out of standard documents and invoices and tax forms are the classic example.

5:37 Instead of a person opening a PDF, squinting at a total, and retyping it into a system, the model does that. Reading for you. The way you get there is three steps, and none of them are scary. First, you train the model, which just means you feed it examples and show it which pieces of information you want back. Then you publish it, which makes it available to your apps and your flows. Then you use it. And from that point on, the reading takes care of itself. Every time a new document arrives. Once it is published, the model keeps doing that same job quietly in the background. The part that surprises almost everyone comes next. You only need five documents to get started. Not five thousand, and not some labeled data set. You'd have to hire someone to build first. Microsoft designed it so a business user can point it at a handful of real documents and still get accurate results tailored to their own content.

6:21 That single detail is usually the moment the idea stops feeling like an enterprise project and starts feeling like something you could try this afternoon. Notice how this fits the pattern from the last section. Document processing is a custom model because it learns the shape of your documents. Specifically, a generic invoice reader can't know that your supplier hides the due date down in the footer. Yours can because you taught it. And that's the difference between a tool that sort of works and one that actually fits. The payoff isn't just speed either. When a machine reads the document, the data lands in your system clean and consistent in the right fields. At the moment the document shows up, the errors that come from tired eyes and copy paste disappear. And the person who used to do that reading gets their afternoon back. That is the bargain. Document processing makes how prediction works.

7:02 If document processing is about reading what already exists. Prediction is about guessing what comes next, and it works from a completely different angle. A prediction model studies the patterns in your past data, figures out what tends to lead to what, and then uses that on new records as they arrive. You're not feeding it documents this time. You're feeding it history. The answers it gives come in three shapes, and that changes what you can ask. The first is a simple yes or no, like whether a transaction looks fraudulent or whether an account is likely to pay its invoice on time. The second is a choice between several outcomes, like whether a shipment will arrive early on time, late or very late. The third is a number like how many items you should keep in inventory, or how many days a delivery will take notice. How ordinary those questions are. None of them require a research lab. They're the exact questions a manager already asks in a meeting, and prediction just gives a data backed answer instead of a hunch.

7:48 It takes a guess that used to live in someone's head and turns it into something you can see and act on, and that is why it feels useful. Keep one honest caveat in mind, though, because it matters. Prediction is a pattern. Guess not a crystal ball. It's only as good as the history you feed it. And if your past data is messy or thin, the predictions will be messy too. So before you trust a prediction, look at the history behind it, because that history is doing most of the work. The model isn't inventing new wisdom, it's reflecting what already happened and projecting it forward. Used well. That's powerful. Used carelessly. It just automates an old bias where you'd actually use it. So far, we've talked about what AI builder is and what it does. And now it's worth asking the practical question, where does this actually plug in? The answer is in the two tools you're already using.

8:28 And each one gives you a slightly different way to reach the models. In Power Automate, you add an AI builder action to a cloud flow. A flow is just a set of instructions that runs on its own. So a new invoice lands in a folder. The flow hands it to the document processing model, and the extracted details get filed into your system without a person touching it. You're not building the AI here. You're wiring an already trained model into a process you set up in PowerApps. You reach the same models through power effects, which is the Excel like formula language behind the app. That means a form can read a document while the user is working or score a record the moment it's saved. The AI isn't a separate screen you switch to. It's a step inside the app, which is exactly why beginners pick it up without learning a new discipline. There's one caveat that trips people up, and it's worth naming plainly.

9:10 Adding an AI builder action turns a flow into a premium flow. And if that flow lives inside an app, it makes the app a premium. App premium means a paid license. So this isn't a free add on to the cheapest plan. It's better to know that before you build because the cost conversation starts the moment a model runs in something real that leads straight to the question everyone asks next, which is what this actually costs. That's where credits come in. What credits are and why they matter. The reassuring part comes first because it's the part beginners never hear. Building, training and testing models in AI builder costs nothing. You can design a model, teach it, run it against samples and confirm that it works all without spending a cent only running a model for real inside a live app or flow spends anything at all.

9:50 So what is a credit? It's a prepaid unit of AI usage and you spend it when a model does work. Microsoft has historically sold an add on of about a million credits for roughly five hundred dollars a month, and some premium licenses also seed a few thousand credits to get you started. Treat those numbers as a rough shape rather than a fixed price. Because licensing moves around and the details change over time. What burns through credits is worth a look, too, because the spread is wide. Reading a page with basic text recognition costs a few credits. Reading a page of a receipt, an invoice, or an ID costs a bit more. Custom document processing where the model is tuned to your own forms costs the most around one hundred credits per page. A couple of tools are simply free, and those include prediction and the business card reader. So the same platform can be almost nothing or expensive, depending on which model you lean on and how often it runs.

10:35 One more shift deserves a sentence because you'll hear about it. Microsoft has been moving AI builder onto copilot credits, the same currency that pays for its other AI tools, and a feature will try your AI builder credits first and then fall back to copilot credits. The practical upshot is that the idea stays the same even as the label on the currency changes. And here's the calm down fact for anyone worried about a surprise bill, going over your monthly credits is treated as a grace period, and it isn't charged. Unused credits don't roll into the next month, so the goal is simply to buy roughly what you use. Because consumption resets every month, your planning window stays short, so you can adjust what you buy as your usage becomes clear. Nothing about this is designed to trap you. Where AI builder fits in real work. Now it's time to bring this down to earth because the concept only lands once you can picture it doing something you recognize.

11:19 AI builder isn't one tool doing one job. It's a set of models, and each one maps to a task that probably already lives on someone's to do list. Think about expense reports, receipt processing, reads the receipts, pulls the amounts and dates, and drops them into a report, which is the kind of chore nobody misses once it's gone. Think about contracts. Contract processing lifts out the key clauses and data points, so a legal team isn't hunting through forty pages for one date. Customer feedback is another easy one. Category classification sorts it into topics, and sentiment analysis tells you whether people are happy. So you stop reading every message just to find the trend. Then there's the work that used to need a specialist prediction can flag transactions that look fraudulent, or customers who might pay late and object detection can take inventory by spotting products in a photo of a shelf.

12:00 Even something as small as snapping a photo of text and saving it into a database is a model away, and text recognition does exactly that. Each of these tasks has the same shape. It's repetitive, it's rules, light, and a person can do it but would rather not do it a thousand times. That is where AI builder earns its keep. The pattern repeats across teams, which is why the same handful of models keeps showing up in expense, legal, sales and support work. It isn't here to think for you. It's here to take the boring half of a task off your desk so you can spend your attention on the part that actually needs a human. And that's really the whole point. AI builder is the bridge that lets you use AI without becoming an AI person. So you keep working in the tools you already know, while the intelligence shows up as one more step in an app or a flow you built yourself.

12:40 Wrapping up. So here's the quick version. AI builder puts AI inside Power apps and Power Automate so you can add it without writing code. You choose between ready made models and custom models trained on your own data. Document processing reads your paperwork prediction forecasts, what happens next, and credits are how you pay once a model runs for real. And here's the question I'll leave you with. If you could hand just one repetitive task to a model this month, which task would you choose? And would a ready made model do the job? Or would you need to teach one your own way of working?