[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"content:\u002Fai-takeoff-software":3},{"path":4,"title":5,"description":6,"h1":7,"kind":8,"updated":9,"calculator":10,"downloads":11,"sections":12,"toc":52,"stats":59,"readingMinutes":60,"related":61,"children":80,"parent":10},"\u002Fai-takeoff-software","AI Takeoff Software: What the AI Does and Where It Fails","How AI counts electrical devices on PDF drawings, with results from our tests on three public bid sets, a real miss on grouped switches, and the fix we wrote.","AI takeoff software: what it does well and where it fails","product","2026-11-13",null,[],[13,17,22,26,30,34,38],{"id":14,"title":15,"kind":14,"html":16},"intro","","\u003Cp class=\"my-5 text-lg leading-relaxed text-slate-800 sm:text-xl sm:leading-relaxed\">AI takeoff software uses a language model to read a drawing set and count what is on it. Ours counts electrical devices. It is fast, and it can be wrong in a specific, repeatable way. Below is what our tests on three public bid sets showed, including a real miss and the rule we wrote to fix it. We make no accuracy claim beyond those tests.\u003C\u002Fp>\n",{"id":18,"title":19,"kind":20,"html":21},"what-the-ai-does","What the AI does","body","\u003Ch2 id=\"what-the-ai-does\" class=\"mt-0 mb-6 text-3xl font-extrabold tracking-tight text-brand-950 scroll-mt-28 before:mb-4 before:block before:h-1.5 before:w-14 before:rounded-full before:bg-accent sm:text-4xl\">What the AI does\u003C\u002Fh2>\n\u003Cp class=\"my-5 text-lg leading-relaxed text-slate-800 sm:text-xl sm:leading-relaxed\">You upload a PDF. The AI follows a written method, and the method covers these steps:\u003C\u002Fp>\n\u003Col class=\"my-5 ml-6 list-decimal space-y-2.5 text-lg text-slate-800 marker:font-extrabold marker:text-accent-dark sm:text-xl\">\n\u003Cli class=\"pl-2\">List each sheet and its role: legend, schedules, one-line diagram, power plan, lighting plan, demolition, details.\u003C\u002Fli>\n\u003Cli class=\"pl-2\">Read the legend, notes and schedules before counting anything.\u003C\u002Fli>\n\u003Cli class=\"pl-2\">Pull the text labels out of the PDF where it can, and count tags such as fixture types and circuit numbers.\u003C\u002Fli>\n\u003Cli class=\"pl-2\">Look at each plan sheet in overlapping zoomed tiles and count visually. The tag count and the visual count must agree.\u003C\u002Fli>\n\u003Cli class=\"pl-2\">Separate new, existing, relocated and demolished devices.\u003C\u002Fli>\n\u003Cli class=\"pl-2\">Compare circuit loads with the panel schedule.\u003C\u002Fli>\n\u003Cli class=\"pl-2\">List open questions and items that are noted but not counted.\u003C\u002Fli>\n\u003C\u002Fol>\n\u003Cp class=\"my-5 text-lg leading-relaxed text-slate-800 sm:text-xl sm:leading-relaxed\">Code then adds up the totals. The AI never does the final addition, because in an early test a model made an addition mistake in its own total.\u003C\u002Fp>\n",{"id":23,"title":24,"kind":20,"html":25},"what-our-tests-showed","What our tests showed","\u003Ch2 id=\"what-our-tests-showed\" class=\"mt-0 mb-6 text-3xl font-extrabold tracking-tight text-brand-950 scroll-mt-28 before:mb-4 before:block before:h-1.5 before:w-14 before:rounded-full before:bg-accent sm:text-4xl\">What our tests showed\u003C\u002Fh2>\n\u003Cp class=\"my-5 text-lg leading-relaxed text-slate-800 sm:text-xl sm:leading-relaxed\">Three tests on public bid sets. Three sets is a small sample.\u003C\u002Fp>\n\u003Cp class=\"my-5 text-lg leading-relaxed text-slate-800 sm:text-xl sm:leading-relaxed\">\u003Cstrong class=\"font-bold text-slate-900\">A park restroom, 3 sheets.\u003C\u002Fstrong> In a blind test, a larger Anthropic model (Claude Fable 5), run without our instructions, matched a commercial takeoff service: 82 devices. The product (Claude Sonnet with our instructions) also counted 82 in five runs, but we wrote the instructions with this set known, so that is not an independent test. We counted the main device types by hand as well, and the numbers agreed (one receptacle near the panel was hard to read, 7 or 8). The AI flagged two extra items itself as doubtful and left them out of the 82: a fire alarm panel symbol and a bonding terminal that a note asks for. In an early run with Claude Sonnet and no written method, the sink GFCI rule was applied to one receptacle instead of three, the model's own total had an addition error, and it did not check the panel loads.\u003C\u002Fp>\n\u003Cp class=\"my-5 text-lg leading-relaxed text-slate-800 sm:text-xl sm:leading-relaxed\">\u003Cstrong class=\"font-bold text-slate-900\">A 7-sheet office renovation.\u003C\u002Fstrong> With the written method, the AI counted 555 new devices in total. On the power and data sheet and on the lighting sheet it was within one device of the commercial service. All luminaire types matched. An earlier blind run without the method counted 549 new devices. Be careful with this result: part of the method, the switch rule below, was written from a miss on this very set. It is not an independent test.\u003C\u002Fp>\n\u003Cp class=\"my-5 text-lg leading-relaxed text-slate-800 sm:text-xl sm:leading-relaxed\">\u003Cstrong class=\"font-bold text-slate-900\">A music stage renovation, 6 sheets, no text layer.\u003C\u002Fstrong> The plans were scans, so the AI could only count by looking. It found 55 new devices against 54 from the commercial service. The extra one was a switch for countertop receptacles in a dressing room, which the AI marked as medium confidence. The AI took about two minutes. This was a set the AI had not seen before.\u003C\u002Fp>\n\u003Cp class=\"my-5 text-lg leading-relaxed text-slate-800 sm:text-xl sm:leading-relaxed\">We have no independent reference answer for the full 7-sheet office count, so we do not claim it is right. We compared sheet by sheet where a reference existed.\u003C\u002Fp>\n",{"id":27,"title":28,"kind":20,"html":29},"the-miss-grouped-switches","The miss: grouped switches","\u003Ch2 id=\"the-miss-grouped-switches\" class=\"mt-0 mb-6 text-3xl font-extrabold tracking-tight text-brand-950 scroll-mt-28 before:mb-4 before:block before:h-1.5 before:w-14 before:rounded-full before:bg-accent sm:text-4xl\">The miss: grouped switches\u003C\u002Fh2>\n\u003Cp class=\"my-5 text-lg leading-relaxed text-slate-800 sm:text-xl sm:leading-relaxed\">On the office set, a model counted a switch symbol labeled &quot;S3 (a,b,c)&quot; as one switch. The label means one device for each listed letter, so three switches. The same went for &quot;S4 (a,b,c)&quot;. The sheet has 10 three-way and 3 four-way switches in all. Two S3 groups and one S4 group were each counted as a single switch, so the model missed six switches out of the 248 devices the service counted on that sheet.\u003C\u002Fp>\n\u003Cp class=\"my-5 text-lg leading-relaxed text-slate-800 sm:text-xl sm:leading-relaxed\">We checked the sheet zoomed in and confirmed the commercial service was right. Then we wrote a rule into the method: a group list after a switch symbol means one device per letter, and stacked labels count each switch. With that rule, the AI counted the switches correctly on the same set.\u003C\u002Fp>\n\u003Cp class=\"my-5 text-lg leading-relaxed text-slate-800 sm:text-xl sm:leading-relaxed\">Our routine for a miss is to find the cause, write a rule and test again on a set the AI has not seen. More misses are likely on drawing styles we have not met. We do not know how many.\u003C\u002Fp>\n",{"id":31,"title":32,"kind":20,"html":33},"where-it-fails-or-slows","Where it fails or slows","\u003Ch2 id=\"where-it-fails-or-slows\" class=\"mt-0 mb-6 text-3xl font-extrabold tracking-tight text-brand-950 scroll-mt-28 before:mb-4 before:block before:h-1.5 before:w-14 before:rounded-full before:bg-accent sm:text-4xl\">Where it fails or slows\u003C\u002Fh2>\n\u003Cul class=\"my-5 ml-6 list-disc space-y-2.5 text-lg text-slate-800 marker:text-xl marker:text-accent sm:text-xl\">\n\u003Cli class=\"pl-2\">\u003Cstrong class=\"font-bold text-slate-900\">Scans with no text layer.\u003C\u002Fstrong> The AI has to count by eye, so it is slower and more likely to miss. Use the confidence levels and check these sets harder.\u003C\u002Fli>\n\u003Cli class=\"pl-2\">\u003Cstrong class=\"font-bold text-slate-900\">Unusual notation.\u003C\u002Fstrong> A symbol that the legend does not describe, or a local convention, can be miscounted.\u003C\u002Fli>\n\u003Cli class=\"pl-2\">\u003Cstrong class=\"font-bold text-slate-900\">Hidden scope.\u003C\u002Fstrong> Furniture-fed devices, fire alarm devices from another contractor and &quot;contractor determines quantity&quot; symbols are counted with a flag, since the real number depends on a decision.\u003C\u002Fli>\n\u003Cli class=\"pl-2\">\u003Cstrong class=\"font-bold text-slate-900\">Dense sheets.\u003C\u002Fstrong> Crowded areas need a second look.\u003C\u002Fli>\n\u003Cli class=\"pl-2\">\u003Cstrong class=\"font-bold text-slate-900\">Wire and conduit length.\u003C\u002Fstrong> The length estimate comes from device positions, drawn scale and circuit tags. It is an estimate from the drawings, not a field measurement, and it has been tested on one sample set so far. Devices without a circuit tag on the plan are not measured.\u003C\u002Fli>\n\u003C\u002Ful>\n",{"id":35,"title":36,"kind":20,"html":37},"why-you-stay-in-charge","Why you stay in charge","\u003Ch2 id=\"why-you-stay-in-charge\" class=\"mt-0 mb-6 text-3xl font-extrabold tracking-tight text-brand-950 scroll-mt-28 before:mb-4 before:block before:h-1.5 before:w-14 before:rounded-full before:bg-accent sm:text-4xl\">Why you stay in charge\u003C\u002Fh2>\n\u003Cp class=\"my-5 text-lg leading-relaxed text-slate-800 sm:text-xl sm:leading-relaxed\">AI takeoff is a first pass, and you review it the way you review a helper's work. Low and medium confidence lines show where to look. Open questions show what to settle. The totals give the size of the job; your review decides whether to trust them. See \u003Ca href=\"\u002Felectrical-takeoff-software\" class=\"font-semibold text-brand-700 underline decoration-accent decoration-2 underline-offset-2 hover:text-accent-dark\">electrical takeoff software\u003C\u002Fa> for the output, and \u003Ca href=\"\u002Fsample-takeoffs\" class=\"font-semibold text-brand-700 underline decoration-accent decoration-2 underline-offset-2 hover:text-accent-dark\">sample takeoffs\u003C\u002Fa> for the three sets.\u003C\u002Fp>\n\u003Cp class=\"my-5 text-lg leading-relaxed text-slate-800 sm:text-xl sm:leading-relaxed\">To learn how symbols and group labels are used, read \u003Ca href=\"\u002Felectrical-blueprint-symbols\" class=\"font-semibold text-brand-700 underline decoration-accent decoration-2 underline-offset-2 hover:text-accent-dark\">electrical blueprint symbols\u003C\u002Fa>. To see how the count turns into a price, see \u003Ca href=\"\u002Felectrical-estimating-software\" class=\"font-semibold text-brand-700 underline decoration-accent decoration-2 underline-offset-2 hover:text-accent-dark\">electrical estimating software\u003C\u002Fa>, and for the proposal and RFI letters, \u003Ca href=\"\u002Felectrical-bid-software\" class=\"font-semibold text-brand-700 underline decoration-accent decoration-2 underline-offset-2 hover:text-accent-dark\">electrical bid software\u003C\u002Fa>.\u003C\u002Fp>\n",{"id":39,"title":40,"kind":39,"html":41,"faq":42},"faq","FAQ","\u003Ch2 id=\"faq\" class=\"mt-0 mb-6 text-3xl font-extrabold tracking-tight text-brand-950 scroll-mt-28 before:mb-4 before:block before:h-1.5 before:w-14 before:rounded-full before:bg-accent sm:text-4xl\">FAQ\u003C\u002Fh2>\n",[43,46,49],{"question":44,"html":45},"How accurate is it?","\u003Cp class=\"my-5 text-lg leading-relaxed text-slate-800 sm:text-xl sm:leading-relaxed\">We do not give a percentage. In three tests, counts were within one device of a commercial service on the sheets we could compare, with one real miss on grouped switches that we fixed. A new set can still produce errors, so review the count.\u003C\u002Fp>\n",{"question":47,"html":48},"Does it work on scanned drawings?","\u003Cp class=\"my-5 text-lg leading-relaxed text-slate-800 sm:text-xl sm:leading-relaxed\">Yes, but with less certainty. Without a text layer, the AI counts by looking, which is slower and harder. In our one scanned test it was within one device of the commercial service.\u003C\u002Fp>\n",{"question":50,"html":51},"Does my drawing train the AI?","\u003Cp class=\"my-5 text-lg leading-relaxed text-slate-800 sm:text-xl sm:leading-relaxed\">No. We do not use your drawings to train models, and the AI provider's commercial terms do not use API inputs for training. See the \u003Ca href=\"\u002Fprivacy\" class=\"font-semibold text-brand-700 underline decoration-accent decoration-2 underline-offset-2 hover:text-accent-dark\">privacy page\u003C\u002Fa>.\u003C\u002Fp>\n",[53,54,55,56,57,58],{"id":18,"title":19},{"id":23,"title":24},{"id":27,"title":28},{"id":31,"title":32},{"id":35,"title":36},{"id":39,"title":40},[],5,[62,66,71,76],{"path":63,"h1":64,"description":65,"kind":8},"\u002Felectrical-takeoff-software","Electrical takeoff software: what you get from a PDF drawing set","What the takeoff gives you from a PDF electrical drawing set: counted devices by sheet, a mark on the drawing for each one, open items and a CSV with named columns.",{"path":67,"h1":68,"description":69,"kind":70},"\u002Fsample-takeoffs","Sample electrical takeoffs from public bid sets","Three example electrical takeoffs from public bid sets: a park restroom, an office renovation and a scanned music stage set. See how to read each takeoff.","hub",{"path":72,"h1":73,"description":74,"kind":75},"\u002Felectrical-blueprint-symbols","Electrical blueprint symbols on construction plans","Plain-words guide to the symbols on commercial electrical plans: receptacles, switches, S3 and (a,b,c) groups, fixtures, panels, home runs and circuit tags.","guide",{"path":77,"h1":78,"description":79,"kind":8},"\u002Felectrical-estimating-software","Electrical estimating software: from the device count to a bid total","Walk an example from a counted drawing set through assemblies, labor rate, overhead and markup to a bid total, with wire and conduit priced per foot.",[]]