Happy Friday, FME'ers! Welcome back to TGIF: Thank Goodness It's FME.
Last week, Sam Walker walked us through what FME Education looks like today — the FME Academy, the learning paths, and the certification program. This week we're back with Part 2, and Sam gets a little more opinionated: the one habit every new FME user should build, why the transformer list is less scary than it looks, and where AI might be taking workspace authoring. Let's go!
Sam's Number One Tip: Use the Run Button
A workspace is a series of instructions, but you have to run those instructions to actually carry out the transformation. Sam's number one tip for new users is to use the Run button — early and often.
New users are often a little nervous about running a workspace. His reassurance: as long as you haven't hooked up a writer yet, you're not writing any data anywhere, so go ahead and run it.
The pattern his team teaches is to make one change — add a transformer, configure a transformer, add a reader feature type, whatever it is — then run the workspace, then inspect the data as it appears at that step. That way you can see immediately whether the workspace did what you wanted. It matters because it catches mistakes early; make a dozen changes before you run anything and it gets much harder to work out where you went wrong.
Why Transformers Feel Overwhelming (And Why They Shouldn't)
The most common challenge Sam sees in new users is the overwhelming feeling that comes from the sheer number of transformers. If you're new, a transformer is a tool in FME Workbench that you add to the canvas to change your data in some way. There are hundreds of them in the product, and more again if you count the custom transformers available on the FME Hub.
You don't need to learn them all to be capable with FME. Sam suggests a couple of dozen to get started, beginning with the top transformers — the ones you'll find useful regardless of what kind of workspace you're building. Once those are under your belt, move on to the more format- and use-case-specific ones, which you'll usually pick up from a colleague's workspaces or from example workspaces in Knowledge Base tutorials.
Break the seemingly insurmountable task into smaller chunks, he says, and people get up to speed very quickly.
Where AI Is Taking Workspace Authoring
Sam is careful to say this is early days and changing quickly, but he sketches out three ways people might author FME workflows.
The first is how nearly everyone works today: manually in FME Workbench, adding objects to the canvas and clicking and dragging to connect them. It's been the main way for a very long time because it's fun, it's easy, and once you're proficient it's very fast. The catch is the learning curve for someone brand new.
The second is conversational. You chat with an agent, give it your requirements and samples of your data, and iterate on the design together until the workspace works the way you expect. A basic version already exists — FME can tell you which transformer to use and offer to add it to the canvas for you — and the team is working on what that looks like at a much bigger scale.
The third is longer term and, in Sam's words, a little crazy: an agent working directly against the FME platform, finding what it needs in the FME engine and running it, with no visual workspace at all.
All of which creates new things to teach. How do you have an effective conversation with an AI? How do you write an effective statement of what you need? How do you build in test and verification loops to make sure it's building the correct thing? Sam's team is planning training content to support exactly that, while keeping quality high for everyone's workflows.
What's Next for FME Education
The project Sam is most excited about is making the FME Academy work for people who don't have a geographic information systems background. FME has been used widely in that industry for a very long time and most customers still come from it, but increasingly the team is seeing people working with databases, web data and addresses who don't think of themselves as spatial users. When they arrive at a course and hear about coordinate systems and geodesy, those terms can alienate them.
Fixing that is, as Sam admits, “a major overhaul of our approach to training” — many of the team have geography backgrounds, and most of their examples use spatial data. It's the challenge he's looking forward to tackling in the coming year.
Sam's FME Wishlist
We couldn't let him go without asking. Top of the list is MCP on FME Flow, so he could turn his team's workflows into something he just asks for in chat and have an agent trigger them. Second is authoring a workspace with an agent working away in the background while you have a conversation about it. Third is quick preview, or auto-caching — where making a change in Workbench automatically runs the workspace up to that point and updates your caches for you. He got to test it internally and thinks it would be incredible, though he's candid about the tricky parts, not least making sure nothing gets written out by accident along the way.
Watch the full conversation with Sam here!
That's a wrap on our Q&A with Sam! Missed Part 1? It covers the FME Academy, the learning paths, and the certification program — catch up here. [LINK TO PART ONE]
And we'd love to know: what's on YOUR FME wishlist? Drop it in the comments. Happy FMEing, everyone!