You bring your own client's requirements document. You leave with the automated check that runs on every project after this one.
A three-workshop system for BIM coordinators and BIM managers.
6 hours building + 3 Q&A · recordings, templates & context files included · The first 10 enrolled get a one to one consulting hour worth €100, included.
Enrolment closes in
The door is shut for this cohort. Enrolment closed on 10 September and the group is settled. Put your name on the waitlist and you hear about the next edition before the page does.
Downloading models. Updating them. Waiting for the software. Adjusting rulesets. Running the check. Waiting again. Sorting the issues, grouping them into packages, building the report, creating the tasks.
I know, because I did it for years. On the $1 billion hospital project this cycle took me twenty hours every second week after model delivery. Mindless clicking, most of the time.
None of that is engineering.
But that's how BIM Coordination has been done for years.
But BIM coordination is not only clash detection.
Ask what "model checking" means in your company. On most projects it means clash detection. Run the test, get three thousand issues, click through them, argue about them, close the ones that matter.
Clash detection is easy to find, every software sees it. Nobody has to define what a clash is, and everyone understands it.
Data is invisible.
Data has no natural definition of wrong. Wrong compared to what? Somebody has to have written down what right looks like.
That written-down thing is the requirements document. On my last hospital project it was six documents, 146 pages, over two hundred properties demanded across all disciplines.
Be honest about your current project. Who has read past page ten?
But even if nobody read it, the contractor still had to order those doors. If a BIM coordinator missed the fire rating requirement on just 1% of the doors, we are talking about at least 100 doors.
I built these checks by hand and ran them for years. Rule by rule. It worked, but it was so slow and so inaccurate that nobody in the project would touch it. That is the honest reason data validation stayed a slogan for a decade.
Then two things happened.
IDS describes data requirements in a form a machine reads and acts on, and most IFC checkers already support it.
One problem stayed. IDS is XML: brackets, namespaces, rigid semantics. A standard project has at least a hundred requirements you have to again click through in an editor. That takes ages.
AI is still bad at modelling. It cannot design for you, it cannot QA the design, and it obviously cannot build it.
AI is good at two boring things: understanding text and structuring data.
Which is exactly where translating the EIR problem is.
You did. So did I.
I asked ChatGPT to translate and create a requirements file. It did.
The syntax came back broken, which was easy to catch if you know it. Then came the part that was not easy. It invented things that do not exist and handed them over with total confidence.
That one was particularly uncomfortable, because it does not validate what the client asked.
AI does not guess randomly. It completes patterns.
That is why we need a method, not a prompt. And I tested a three step method I want to teach you.
Draft. You give AI the client's document, a structured prompt and the context files. It turns prose and merged cells into one structured table.
Verify. You check what came back, in three different ways that catch different AI hallucinations.
It is rigid, and boring. That is exactly why you can trust it.
Validate. You load the file into your coordination software and run it against the model. You get the issues and the report, then you automate the whole thing so it runs overnight.
AI drafts it. You verify it. The software validates it.
Picture the same week, four weeks from now.
The new model revision lands on Friday afternoon, the way it always does. You drop it in the folder and go home. Overnight, the check runs against every object in the model.
Monday morning the report is waiting. Not three thousand unsorted issues. The ones that failed your client's actual requirements, with the elements named.
You spend two hours deciding what gets reported and to whom.
You stop being the person who checks models. You become the person who set up the checking.
I reduced my data validation from 20 hours to 2 hours.
| Stage | What you do now | What you do after |
|---|---|---|
| Understanding client requirements | You read hundreds of lawyerly pages, or scan endless Excel rows, and jot down the lines that sound like a requirement. Or you ignore all of it. | A clear path and a set of prompts that turn the requirements document into a structured table. |
| Matching requirements to IFC classes and properties | You guess, or you search the schema documentation one property at a time. | AI builds the table against the IFC schema, with a source quotation so you can spot-check it easily. |
| Creating the checkable file | At least seven clicks (counted!) per requirement in an editor. | One minute, from the table you already verified. |
| Validating the model | 20 hours every model delivery. | You automate the check. You choose which issues get reported - down to 2 hours. |
I was in that left-hand column for years, on a 125 000 m² hospital.
Let me be straight about the nearest future:
Today, you have two options. Click through by hand, or ignore it.
I want to give you a third.
From your client’s EIR to a model you can actually trust. Draft the rules with AI, verify them yourself, validate against the model. Repeatable method, live, hands-on, no coding.
One method split in three live workshops with me. Two hours of teaching and thirty minutes of questions each, spread over three weeks.
This is a workshop and not a lecture. I show the example on my screen, you work in parallel on your own document and your own model. If you do not have one with you, you use the files I share. The chat stays open the whole session.
My name is Konrad Fugas. Ten years in this industry, from three different chairs.
I have run around site in contractor's boots. I have submitted bug tickets and led implementations as a software vendor. And for the last five years I worked owner-side on Nye SUS, the new Stavanger University Hospital. 125 000 square metres, built without drawings, opened in January 2026.
My job there was writing the requirements and checking the IFC models. Data validation in particular, through the whole detailed design cycle, for around three years. I know exactly how bad data can get.
I am also co-founder of the BIM Corner blog - one of the most popular education platforms about BIM in the world, with around 12 000 monthly readers. For the last seven years I have written over 50 articles about data management and open standards. I have spoken about digital delivery from the stage at BILT Europe, InfraBIM and BIM Meetup Poland, and I was a co-lecturer on the IFC Masterclass by BIM Corner.
This is the first edition, so nobody has taken this course yet. I will not pretend otherwise. Here is what people said about how I teach.
4.86/ 5
BILT Europe, participant evaluationVery good presentation. Fun breaks in the slides. Well structured, managed to keep our attention also through the more boring stuff.
4.54/ 5
IFC Masterclass by BIM Corner, average ratingLoved the clarity where it is explained how Exchange Information Requirements are translated into IDS.Which is the exact step Workshop 1 is built on.
Very inspiring, and well presented. The team felt grateful for a deep dive to what is mostly in your past but might be in some of our futures.
Jan LeenknegtBIM Director, BIG
I read your content because you are practitioners, not the pseudo-experts who are becoming more and more common in the BIM world. I wrote two diploma theses largely based on your blog, which allowed me to develop further with a new employer and to pursue a doctorate.
Karol ZawadaSenior BIM Specialist, PhD Researcher, CPK
From the offers I have seen land on projects I worked on, an hour of this kind of consulting in Western Europe goes for €100 to €150. This is six hours of teaching plus the Q&A, and that is before the templates and context files you keep.
Now put the number next to what it stands against.
For the first 10 enrolled, as a gratitude for the trust you gave me, I add a one-hour session with me, one to one.
It is the only part of this that does not scale. Six hours of workshops cost me the same whether ten of you turn up or fifty. Ten hours one to one cost me ten hours. That hour is worth €100, and here it is included with the workshops.
Once the ten are gone, they are gone.
You decide what we spend it on. Your own requirements turned into a working IDS file. An audit of a validation setup you already run. The data validation process for a project that is starting. Or anything else inside what I do.
If your documents sit under an NDA, we agree on a way to anonymise them so you are not sharing what you cannot share.
We schedule it after the workshops finish.
Only 10 of 10 spots left
One honest thing before you decide. This workshop series is not for everyone.
Join if you
NOT for you
I would rather you did not join than join for the wrong reason.
Six hours live, spread over three weeks. Two hours a week. Recordings go out the next morning if a session clashes with something, and you can watch them at 1.5x speed. You keep the recordings, the notes, the assignments and the quizzes.
Neither is this. I show you how to build it without writing a single line. The only coding-adjacent skill required is the ability to look at an XML structure on screen without panicking.
That is expected, and it is fine. You get pre-workshop materials on IDS before the Workshop 2, so you arrive with the basics already in your head, and we go through the parts that matter again in the first session. You do not need to know the schema by heart. You need to know what a good requirement looks like, and that is what we build together.
If you run Solibri, usBIM or BIMvision, you are set and you will get the most out of the third session. If Navisworks or ACC Model Coordination is all you have, be honest about what that does today: it finds clashes, not data. Although you will still build the same IDS file and still run the check. Sortdesk validates IDS and it is free.
Recommended, yes. I strongly recommend a paid plan for two reasons: data protection and the context limits you run into on requirements documents. You can do all of the work on the web interface of any LLM, but in the free version you will face a wall very fast.
You send me one sentence before the refund window closes. I refund in full, and you keep the Workshop 1 materials. There is no form, no questionnaire and no follow-up call. If you paid in two instalments, the second one is cancelled and the first is returned.
Correct, and mine were too. In the first session we take apart a document spread across every building stage, with three or more requirements crammed into one cell. It is based on a real project document. If yours is worse than that, the bonus for the first 10 enrolled is exactly what it is for.
We cover data security in the first session: what stays local, what does not, and how to anonymise a document so the process still works. If your organisation has a hard ban on AI tools, this is not the course for you, and I would rather say that now.
Most people who hesitate are not stuck on the price. They are stuck on one sentence - "I need to ask my manager first." Here is how you win that conversation.
Your manager cares about three things:
Hours.How much of your week goes into work that produces nothing new.
Risk.What is sitting in the delivery, and whose problem it becomes.
Reuse.Whether the money buys something that benefits the team.
Say it in those words and you will get a yes far more often than if you open with technicalities.
I wrote a one-page version of that argument, with a budget request template you can forward without editing and the VAT line your finance department will ask for.
I send it to you personally, usually within a few hours. You do not have to buy anything.
If three or more of you are coming from the same company, write to me before anyone pays.
konrad.fugas@bimcorner.com. I answer the same day, and I will tell you straight if a group makes no sense for you.
Here is what is inside:
Three recorded live workshops · every file you build stays yours · none of it expires
Leave your address and you hear when the doors open again, before the page says a word.
And if you decide to do nothing today, that is a fine decision too. The next set of requirements still lands on your desk in three weeks. It will look exactly like the last one.
PS. Enrolment for the first cohort is closed and the group is settled. The door stays shut until the next one, and the waitlist is how you find out when that is.
PPPS. The first 10 enrolled also get an hour with me, one to one, on whatever they need most: their own requirements, an audit of what they already run, or the process for a project that is starting. Bring the messiest document you have.
