Three new AI steel takeoff tools showed up in 2026: Ferra (bidferra.com), a demo-gated steel-native tool pairing beam/column detection with bid intelligence; Steel Genie, ALLPLAN's $9,950 takeoff product that generates an estimating-level 3D model with AISC connection checks; and Exayard, a multi-trade AI platform at $99–$249/user/month with a structural steel module. All three read framing plans and detect member marks. None of the three has published an accuracy methodology yet, so the honest answer for most shops is: test on a job you already have final counts for before you trust any of them with a bid.
Who's New in AI Steel Takeoff for 2026?
Two years ago you could count the AI steel takeoff tools on one hand. Now the category has at least six active products, and comparison roundups are multiplying fast enough that half of what you'll read about these tools was written by one of the vendors.
Here's the field as it stands in mid-2026:
| Tool | First seen | Steel-native? | Published pricing | Accuracy methodology published? | |---|---|---|---|---| | Ferra (Bidferra) | Private beta ~April 2026 | Yes | No — demo-gated | No | | Steel Genie (ALLPLAN) | Launched April 8, 2026 | Yes | $9,950 | No | | Exayard | Steel module confirmed 2026 | No — multi-trade | $99–$249/user/mo | No | | SketchDeck (LIFT) | 2025, institutional round 2026 | Yes | Demo-gated | Partial (demo video) | | Beam AI | Established, 1,200+ contractors | No — multi-trade | Quote-based | No |
This article covers the three newest names in detail, because that's where the least independent information exists. If you want the full field including the legacy measurement tools (Bluebeam, PlanSwift, On-Screen Takeoff), start with the complete steel takeoff software comparison.
One pattern before the specifics: all three launched with strong claims and thin public evidence. That's normal for new software — it just means the burden of verification is on you, and vendors that make verification easy deserve more of your evaluation time than vendors that don't.
What Does Ferra (Bidferra) Actually Do?
Ferra is the newest and least public of the three. It surfaced in private beta around April 2026 with minimal public information, raised a reported $3.31M round in late 2025, and as of this writing there is still no self-serve access — you request a demo.
What Ferra says it does:
- Beam and column detection from structural drawings, steel-native rather than multi-trade
- Estimator-in-the-loop review — their positioning language is that the estimator reviews every number
- Revision comparison — compare drawing set revisions without double-counting members, which is a real pain point when addendum 3 lands two days before bid
That last one is the interesting differentiator. Most AI takeoff tools treat each drawing set as a fresh job; when the revised set arrives you re-run everything and manually reconcile. If Ferra's revision diff works as described, that's genuinely useful. But there are no independent user reviews, no published accuracy numbers, no named case studies, and no way to test it without going through sales.
At least one 2026 roundup called Ferra the "strongest fit for structural steel fabricators." Read those roundups skeptically: several of the sites publishing best-of lists this year are affiliated with vendors on the list, and a demo-gated tool with no public users is hard to rank above tools you can try today.
We're preparing a deeper head-to-head as more information becomes public: SteelFlo vs Ferra.
Bottom line on Ferra: promising positioning, real differentiator if the revision comparison holds up, but currently unverifiable. If you book a demo, bring your own drawings and ask them to run a set you have final fabrication counts for.
What Does Steel Genie Do?
Steel Genie launched April 8, 2026, and it's the most significant entrant of the year for one reason: it's backed by ALLPLAN, part of the Nemetschek Group — the same parent as SDS/2. That means enterprise distribution, brand credibility, and a natural channel through detailing software shops.
What Steel Genie does, per its published materials:
- Detects beams, columns, joists, and braces from PDF drawings
- Captures size, length, weight, moment connections, camber, and shear studs
- Runs connection calculations against the AISC Design Guide under LRFD and ASD
- Generates an estimating-level 3D model — not just a quantity list
- Priced at $9,950
Its one published case study is Master Steel, reporting a 67% throughput increase. Independent comparisons through mid-2026 flag two consistent weaknesses: detection is strong on framing plans but weaker on elevations (bracing and lateral systems live there), and revision handling is re-run-per-version with no diff view.
The evaluation question for Steel Genie is really a workflow question. If your shop lives in the SDS/2 or ALLPLAN ecosystem and you want connection-level depth — moment connection counts, camber callouts, shear stud quantities feeding a 3D model — the $9,950 buys you something the lighter tools don't attempt. If you're a 5-15 person shop that needs a fast, verifiable member count to price a bid, you're paying for a lot of model you won't use, and the elevation gap means you're still manually taking off your braced frames either way.
Note also what the 67% number is: one case study, from the vendor, on throughput — not accuracy. Nobody has published what percentage of members Steel Genie finds or misses on a blind drawing set.
What Does Exayard Do?
Exayard is different from the other two: it's a general-purpose, multi-trade AI takeoff platform with a structural steel module, not a steel-native tool. It's also the most transparent of the three on pricing: $99/user/month Starter, $249/user/month Professional, with an AI credit system metering usage.
The structural steel module handles:
- Beam and column mark detection from framing plans
- Grid-to-grid length extraction and schedule cross-referencing
- W-shapes, base plates, gusset plates, shear tabs, moment connections
- Tonnage calculation and export to Excel, ProEst, and Sage
- "Smart Estimates" — turning quantities into branded proposals
The pricing is genuinely accessible and the export story is decent. Two cautions:
First, multi-trade platforms historically go shallow on steel. Reading structural drawings is its own problem — schedule marks that resolve on a different sheet, notation that varies by detailer, the same member appearing on plan, elevation, and section views. Tools built for counting doors and measuring drywall tend to underestimate this.
Second, Exayard runs aggressive content marketing — including "90% time savings" claims and self-authored best-of-2026 roundups where Exayard ranks favorably. No accuracy methodology has been published. When a vendor's loudest public output is SEO comparison content rather than verification data, weight your evaluation accordingly.
Bottom line on Exayard: the cheapest way to experiment with AI takeoff if steel is one of several trades you estimate. If structural steel is your business, test the steel module hard before relying on it — specifically on schedule-driven drawings and anything with heavy elevation content.
What Questions Should You Ask a New AI Takeoff Vendor?
The full checklist is in questions to ask AI takeoff vendors, but for 2026's crop of new tools, five questions do most of the work:
| Question | Why it matters | |---|---| | "Can I run my own drawings before buying?" | Demo-gated tools control what you see. A vendor confident in detection lets you upload your own set. | | "What's your accuracy methodology — and is a human in the loop?" | "95%+" means nothing without knowing what's counted, on what drawings, and whether a person verified it. | | "How do you handle elevations and sections, not just plans?" | Framing-plan-only detection misses bracing, lateral systems, and misc steel. This is the known gap in at least one 2026 entrant. | | "What happens when a revision set arrives?" | Re-run-and-reconcile vs. a real diff is hours per addendum. Only one new tool even claims to solve this. | | "Which section standards do you cover?" | See the next section. If you ever see BS, EN, or AS/NZS callouts, most new tools can't read them. |
And one meta-question: ask how the count is derived. If the tool infers quantities rather than showing every detected member instance, you can't audit it — and a count you can't audit is a count you'll re-do manually the first time it burns you.
How Do New Tools Compare on Standards Coverage?
Here's where the 2026 entrants show their age. All three launched US-first, which in practice means AISC W-shapes, HSS, and the common plate/angle callouts. If your drawings ever reference anything else, coverage matters:
| Tool | AISC (US) | BS/IS (UK/India) | AS/NZS (AU/NZ) | EN (Europe) | GB (China) | CFS/SSMA (cold-formed) | |---|---|---|---|---|---|---| | Ferra | Yes (claimed) | Unknown | Unknown | Unknown | Unknown | Unknown | | Steel Genie | Yes — AISC connection engine | No published support | No published support | No published support | No published support | No published support | | Exayard | Yes (W-shapes, plates) | No published support | No published support | No published support | No published support | No published support | | SteelFlo | Yes | Yes | Yes | Yes | Yes | Yes |
SteelFlo covers 4,500+ steel sections across all six standards, which as of 2026 no other tool in the category attempts. That's not a niche concern: on offshore-designed projects, PEB packages with cold-formed secondary members, or any set where a UB, UC, HEA, or SSMA stud callout appears, a US-only section database silently misreads or skips those members. Ask any vendor what happens when their detector hits a 310UB40 or a 600S162-54 — cold-formed is the sleeper issue, since plenty of "structural steel" jobs carry SSMA studs and tracks in the same drawing set.
Which Tool Should a Small Fabricator Try First?
For a small shop evaluating in 2026, sequence by verifiability, not by feature list:
- Start with anything you can test free on your own drawings. One run on a job you've already fabricated teaches you more than three vendor demos. Compare by shape family, not total tonnage — tonnage hides offsetting errors.
- Try Exayard second if budget is the binding constraint and steel is one of several trades you bid. At $99/user/month the downside is small; verify the steel module on schedule-heavy drawings first.
- Evaluate Steel Genie if you're an SDS/2 or ALLPLAN shop with connection-level estimating needs and $9,950 doesn't move your software budget. Test elevations specifically.
- Watch Ferra, but don't wait for it. Until there's self-serve access or independent reviews, it's a bookmark, not a candidate.
Whatever you pick, run the same acceptance test: a drawing set with known final counts, compared by shape family, including at least one revision cycle and at least one sheet that isn't a framing plan. A tool that survives that earns a spot in your bid workflow.
If you'd rather start with the established option: SteelFlo detects every steel member label on your drawings, shows each one as a box you can verify or reject — 95-99% accuracy with that human verification step — and covers 4,500+ sections across 6 standards. Your first 3 takeoffs are free, so you can run the acceptance test above on your own drawings before spending anything.