Every discipline, read against every other one
Flikt.AI reads your drawings and your specifications together and returns findings in five categories — missing coordination, specification mismatches, documentation gaps, code violations, and spatial clashes — same day.
Most plan review is organized by discipline. The architect checks the architectural set, the engineer checks their own, and each one signs off on work that is internally consistent. The conflicts that reach the field are almost never inside one discipline. They live in the space between two sets, where nobody was assigned.
That space is what Flikt.AI reads. Below is what comes back, in the vocabulary the report itself uses. Every example is a real finding from one of the demo reports published at demo.flikt.ai — you can open the report and read the finding in full.
The five kinds of findings
Every finding is filed under one of five types. The type tells you who owns the answer and what a resolution looks like.
Missing coordination
One discipline calls for something the others never show. This is the largest category in practice, and the one a discipline-by-discipline review is structurally least likely to catch.
From a real reportA mechanical note requires duct smoke detectors and neither the electrical nor the fire alarm plans show them. A fire pump appears in the plumbing scope with no structural support detailed for it.
Specification mismatch
Two documents describe the same assembly differently. Whichever one the field follows, somebody is building to the wrong sheet.
From a real reportWall types call one gypsum thickness and the finish notes call another. A truss bearing elevation does not leave the ceiling height the architectural sheets promise. A spec section requires air chambers in one place and prohibits them in another.
Documentation gap
The set refers to something it never delivers. These are cheap to fix on paper and expensive to discover on site.
From a real reportA corridor reflected ceiling plan defers exit signage to life safety sheets that are not in the bundle. A lighting legend leaves a fixture as TBD. Firestop details filed under the wrong discipline, where the trade that needs them will not look.
Code violation
A conflict with the code the set itself cites. Not our opinion of the code — the standard the documents already point to.
From a real reportPipe penetrations through a rated floor-ceiling assembly with no firestop detail at the penetration. Electrical notes referencing an older code edition than the cover sheet specifies. A panel demand load above the rating of the main breaker feeding it.
Spatial clash
Two physical elements competing for the same space, found in 2D — no model to build, federate, or maintain.
From a real reportWaste and vent risers routed through structural truss zones. A duct crossing a concrete beam. Condenser units sited inside a planting bed, without their clearance.
Sixteen disciplines, one pass
Specifications are treated as a discipline, not an appendix. A requirement in the spec book that contradicts a note on a sheet is a finding like any other. The same holds for anything else you include in the upload — if brand standards come with the set, they are cross-checked against the drawings the same way.
What a finding contains
Findings are written to be forwarded, not summarized. Every one carries the same fields, so a reviewer can go to the sheet and confirm it rather than take our word for it.
What Flikt.AI does not do
Flikt.AI reviews documents. It does not do quantity takeoffs, it does not measure, and it does not build or federate a model. If the work in front of you is model-based clash detection, that is a different tool and we are not a replacement for it.
What Flikt.AI does is read the set you already have, at the stage you already have it — bid documents, permit sets, the revision that landed this morning — and hand back the conflicts before they are built.
Whether that is the right instrument for your project is a fair question, and we answer it in full — including the cases where the answer is no — on is this a good fit.
See the output, not a description of it
Three full reports are published, from real runs with the project identities anonymized. Open one, run the analysis, and read the findings the same way a customer does. Every example on this page is drawn from them — nothing here is illustrative.
RenovationThe AtriumCommon-area renovation. Shows what surfaces when new work meets existing conditions.Open the report →
Single familyMeridian ResidenceCustom residence. A smaller set, where coordination conflicts still concentrate around structure.Open the report →
Frequently asked questions
Do you need a 3D model?
No. Flikt.AI reads the 2D documents — drawings, specifications, and whatever else is in the upload. There is nothing to model, federate, or maintain, which is why the review can run on a bid set or a permit set rather than waiting for coordination to reach a modelled stage.
What do you need from me?
The plan set and the specifications, as PDFs. Include every discipline you have; the value of the review is in the cross-references, so a partial upload finds less. If you have addenda, revisions or brand standards, include those too and they are read as part of the set.
How is this different from asking a general-purpose AI chatbot to review the plans?
A chatbot answers a question about a document you point it at. Flikt.AI reads the whole set at once and checks each discipline against every other one, which is where coordination conflicts live — no single sheet contains the conflict, so no single-document question surfaces it.
Are the findings verifiable?
Every finding cites the sheet numbers and the spec sections it came from, and quotes the language it is reading. The intended workflow is that a reviewer opens the sheet and confirms or rejects the finding, then forwards the ones that stand as RFIs.
What kinds of projects does this work on?
Any documented building project — multifamily, single-family, retail and tenant buildout, hospitality, office, institutional, industrial, and renovation work. The review is document-based, so what matters is the completeness of the set, not the building type.
Catch it on paper, not in the field
Upload a set and see what comes back. The first preview is free.