Takeoff Requires Your Knowledge
It shouldn’t consume your life.
Electrical takeoff isn’t just counting symbols.
You look at a drawing and see a receptacle.
But you don’t just see a receptacle.
You see where it is.
What kind it is.
What it’s being installed in.
How it’s going to be installed.
What materials that installation requires.
And how your company normally builds it.
The symbol is only the beginning.
The electrician supplies the knowledge.
AI found 83 receptacles.
Great.
How many different ways do you have to build them?
That’s the problem with treating symbol recognition as electrical takeoff.
AI may become extremely good at finding every receptacle symbol on a drawing.
It may find all 83.
It may even identify them with extraordinary accuracy.
But you aren’t bidding 83 symbols.
You’re bidding 83 installations.
And those aren’t necessarily the same thing.
The symbol didn’t change.
The assembly did.
Two identical receptacle symbols can represent completely different work.
One is installed in drywall on metal studs.
Another is in a concrete block wall.
Another is on a tiled wall and requires a different box or installation method.
Another is outdoors.
Another is surface-mounted.
Another is in a location where information somewhere else in the drawing set changes what has to be installed.
The electrical symbol can be identical.
The work isn’t.
That’s what an electrician sees.
A symbol can tell you what the designer wants there.
It doesn’t necessarily tell you how you’re going to build it.
The drawing may tell you it’s a receptacle.
It may even say GFCI.
Useful information.
But the person who drew the symbol isn’t necessarily telling you every box, fitting, raceway, support, material choice or labor requirement involved in actually installing it.
Some of that information may be elsewhere in the plans.
Some may be in the specifications.
Some comes from the installation condition.
And some comes from knowing electrical construction.
That’s electrician knowledge.
AI can get the count right.
And the takeoff wrong.
Suppose AI finds all 83 receptacles.
83 out of 83.
Perfect symbol-recognition accuracy.
But suppose those 83 receptacles actually represent six different installation conditions requiring different assemblies.
If the software treats them all the same, what does 100% symbol accuracy really tell you?
Recognition accuracy isn’t assembly accuracy.
And ultimately, you’re not buying symbols.
You’re buying material and labor to build the work those symbols represent.
AI can recognize a switch.
An electrician understands what it means.
The same problem exists throughout an electrical drawing.
A switch isn’t necessarily just a switch.
A fixture isn’t necessarily just a fixture.
A receptacle isn’t necessarily just a receptacle.
The symbol is one piece of evidence.
The electrician interprets that evidence in the context of the job.
That’s why electrical takeoff still requires electrical knowledge.
So what’s wrong with AI takeoff?
Nothing — if you ask AI to do what it actually knows how to do.
Finding similar symbols?
Useful.
Recognizing patterns?
Useful.
Searching large drawings?
Useful.
Remembering previous decisions?
Very useful.
Suggesting that something looks like something you’ve seen before?
Potentially extremely useful.
But there’s a big difference between:
“I think I found 83 receptacles.”
and:
“I know exactly how your company will install all 83 of them.”
The first can be useful.
The second requires evidence.
MEPTrax isn’t being built around pretending AI already knows your trade.
We’re building toward AI that learns from it.
That starts with the electrician.
You identify the work.
You determine what it represents.
You choose the appropriate assembly.
And MEPTrax can begin retaining the relationship between what was on the drawing and the decision you made.
Not just:
This picture looks like a receptacle.
But eventually:
This symbol
in this location
under these conditions
was identified by an electrician
as this assembly.
Now the software has evidence.
One decision doesn’t teach very much.
Thousands of decisions create patterns.
Every takeoff creates information.
An electrician sees a condition.
Makes a decision.
Chooses an assembly.
Then does it again.
And again.
Across drawings.
Across projects.
Across different conditions.
Across different electricians.
Patterns begin to emerge.
That’s when shallow AI starts becoming useful.
It doesn’t suddenly become an electrician.
It gets better at predicting what an electrician is likely to decide.
That’s a very different proposition.
First, let the electrician teach the software.
Then stop making the electrician teach it the same thing over and over again.
Imagine you’ve identified the same type of condition repeatedly.
Eventually, why should the software continue acting as though it has never seen it before?
Instead of waiting for you to make every identification from scratch, MEPTrax could begin asking:
“These locations appear to match this assembly. Review?”
Now the electrician isn’t removed from the process.
His role becomes more valuable.
Instead of spending his time repeating obvious decisions, he can spend more of it reviewing the places where judgment is actually required.
The electrician makes the decision.
The software carries more of the repetition.
MEPTrax doesn’t need to be right every time to become useful.
It needs to know enough to help.
High-confidence match?
Show it.
Unusual condition?
Flag it.
Not sure?
Ask.
Electrician disagrees?
Correct it.
Then retain the correction.
Recognize. Suggest. Review. Correct. Learn.
And do a little better next time.
The goal isn’t for AI to make the electrical decision.
The goal is to stop making the electrician repeat decisions the software has enough evidence to help with.
That’s an important difference.
Electrical knowledge remains valuable.
Judgment remains valuable.
Experience remains valuable.
Repetition isn’t.
If your people have already made the same decision hundreds of times, the software should eventually be able to help with decision number 101.
And there’s something else AI doesn’t know.
How YOUR company does the work.
Even electricians don’t necessarily make identical choices.
Companies have their own:
Materials.
Methods.
Assemblies.
Suppliers.
Labor expectations.
Preferred products.
Installation practices.
Standards.
Experience.
What one electrical contractor considers the normal assembly for a condition may not be what another contractor uses.
So the really interesting question isn’t only:
What do electricians usually do here?
It’s:
What does YOUR company usually do here?
Industry knowledge is useful.
Your company’s knowledge may be more useful.
With enough evidence, MEPTrax can potentially learn broad patterns from electrical contractors.
But your company’s own history can provide another layer.
Perhaps electricians generally choose one of three assemblies for a particular condition.
But your company almost always chooses one.
Now the question can become:
“This looks like a condition where your company normally uses Assembly 27. Is that correct?”
That’s a much more useful assistant.
It isn’t telling you how to be an electrician.
It’s remembering how you already decided to be one.
What if every takeoff made the next takeoff a little easier?
Today, most completed takeoffs eventually become history.
The project is bid.
You win it or lose it.
The plans get filed.
Then another drawing arrives.
And much of the thinking starts over.
But those old takeoffs contain decisions.
What did you identify?
Which things looked alike but weren’t?
Which conditions changed the assembly?
What did your estimator choose?
What did your company consider normal?
That’s knowledge.
Why throw it away?
Takeoff requires your knowledge.
It shouldn’t consume your life.
Your value isn’t your ability to click a receptacle symbol 83 times.
Your value is knowing that those 83 receptacles don’t necessarily represent the same work.
That’s the knowledge MEPTrax should preserve.
That’s the judgment MEPTrax should respect.
And eventually, that’s the evidence MEPTrax can use to carry more of the repetitive workload.
Let electricians do the thinking.
Let software remember what they’ve already taught it.
What about AI accuracy?
We’ll tell you when we’ve earned the right to.
We could put a percentage on a website.
90%.
95%.
98%.
It would sound impressive.
But 98% accurate at recognizing symbols is not the same thing as 98% accurate at understanding the work.
For an electrical contractor bidding real money, that distinction matters.
So MEPTrax’s approach is simpler:
Gather evidence.
Learn from real decisions.
Measure the results.
Test the predictions.
Publish accuracy when we can demonstrate what that accuracy actually means.
Until then, the electrician remains the authority.
This is where we think AI belongs.
Not replacing electrical knowledge.
Not pretending a PDF contains information that isn’t there.
Not confidently turning every matching symbol into the same assembly.
Instead:
Find what machines are good at finding.
Remember what people have already taught them.
Recognize recurring patterns.
Suggest likely answers.
Surface exceptions.
Ask when uncertain.
Learn from corrections.
Then give the electrician his time back.
MEPTrax Takeoff
AVAILABLE NOW
The learning and AI capabilities described above are the direction MEPTrax is being built toward.
They are not what we’re selling you today.
Today, MEPTrax Takeoff gives electricians a straightforward way to work from PDF electrical plans:
Count electrical work.
Use assemblies.
Organize work by room or area.
Review quantities.
Export the takeoff.
And it gives us the right place to start.
Takeoff requires your knowledge.
Start by putting that knowledge somewhere it can eventually learn from it.
Try MEPTrax Takeoff Free
30 days free. No credit card required.
AI-assisted symbol recognition, learned assembly suggestions, company-specific predictions and other future capabilities described on this page are under development and are not included with MEPTrax Takeoff today unless explicitly stated.