Specifications get reissued. A designer sends revision C, and somewhere in two hundred pages a material has changed, a tolerance has tightened, or a clause has been added that puts a cost on you.
Finding those changes is somebody senior reading both versions side by side for an afternoon, and it is exactly the kind of work people put off. When it gets skipped, the change is discovered on site, which is the most expensive place to discover anything.
Contracts have the same problem with higher stakes. The version you are asked to sign is rarely the version you sent, and the difference between them is where the risk sits. Reading a contract properly takes hours, so it gets skimmed, and the clause that matters is the one nobody read.
We build the AI inside APSIS Business Components, a construction software platform. Two of the capabilities in it deal with documents that change.
Specification comparison takes the old version and the new one and reports what actually changed. Contract review takes the contract you sent and the contract you got back, and finds where the risk has moved. You can then ask it questions in plain English about what a clause means for you.
Comparing documents sounds simple and is not. A page can look identical and read differently, and it can look completely different because a paragraph moved and read exactly the same.
So the comparison runs two ways at once, independently. One looks at the pages as pages, and catches anything that moved, was added or was removed. The other reads both versions as language, and catches the changes that matter but do not look like anything: a material substituted, a figure altered, a requirement that used to be optional.
Each of those runs and reports on its own. If one finishes and the other is still working, you see that, rather than staring at a spinner wondering which part failed. If something goes wrong, the comparison can be run again and the new attempt keeps a link to the one it replaced, so there is a history rather than a mystery.
And as everywhere else in the platform, when someone corrects what the AI found, the system keeps both: what the AI said, what it was changed to, and who changed it. On a specification that matters, because the record of what changed is the thing you will be arguing from later.
Contract review works the same way, comparing what you sent against what came back, and it adds something the specification side does not need.
You can ask it questions. What is our liability if the programme slips. Does this cap our exposure or theirs. What changed about payment terms. The answer comes back with the part of the document it came from, so you are reading the actual clause rather than trusting a summary.
That last point is the whole design. An answer without its source is an opinion, and nobody should make a contractual decision on an opinion from software. An answer with the clause attached is a shortcut to the paragraph you needed to read, which is genuinely useful and safe to act on.
Corrections here are recorded against the specific clause, so a note that "our reading of clause 14.3 differs" survives, with the name of whoever made it.
Every change in a reissued specification has a price, and the price goes up the later it is found. Caught at review, it is a conversation and possibly a variation. Caught on site, it is rework, delay and an argument about who should have spotted it.
The same is true of a contract. A clause read before signing is a negotiation. The same clause read after signing is a loss.
None of this replaces the person doing the reviewing. It puts the differences in front of them in minutes, so the hours they spend go on judging what the changes mean rather than on hunting for them.
Comparing documents is one of the things we have built into the APSIS platform over more than a year. The others: reading drawings and schedules, processing procurement paperwork, pricing quotes, drafting bids and answering questions from project documents.
They run on the same foundations. Each arrives faster than the last, and they work together. A specification compared in one place can be questioned in another. A change found in a reissue can be priced against the bill it affects. That is the advantage of building capability into a platform rather than bolting separate tools onto the side of a business.
It is also the reason this work has continued for over a year rather than ending at a pilot. Read how we build the AI inside the platform →
Tell us where your team loses the most time. We will tell you honestly whether AI pays there, what it takes to build, and what we have already delivered for businesses like yours.
We use cookies to measure how the site is used and, if you allow it, for advertising. Nothing is set until you choose. Privacy policy.