An amended standard form is a different contract
An unamended JCT is a known quantity. The same contract with twenty amendments is not, and the usual way a term contract goes wrong is that nobody read them until the first dispute.
A measured term contract arrives as a standard form plus a schedule of amendments. The standard form is a known quantity — priced, insured and argued over for decades. The amendments are where the risk actually sits, and the commonest way a term contract goes wrong is that nobody read them properly until the first dispute.
So Bid-On will read a schedule of amendments and tell you what it changes compared with the unamended form, and what you would ask the buyer about it. Its value is entirely about timing: the questions are on the table before the clarification window shuts rather than after it, which is the difference between a point that can be negotiated and one that has to be priced.
Three things it is not, and each is enforced rather than promised.
It is not legal advice. It says nothing about whether a clause is enforceable, the instruction forbids it, and the output is phrased as questions and observations. It is a first pass to put in front of whoever advises the customer on contracts — which is the decision it is actually for.
It is not a decision. It accepts nothing, rejects nothing and qualifies nothing. Accepting an amended contract binds a company, and no agent mandate in this platform reaches an act that binds one. A test asserts that the bid’s status, its questions, its answers and its returnables are byte-identical after a reading — the only thing that changes is that some questions now exist to ask.
And it is not a substitute for reading them. It marks a point critical only where an amendment moves a risk a subcontractor cannot price or insure: unlimited liability, a cap deleted, fitness for purpose, pay-when-paid, set-off without notice, a payment period materially extended, termination at will. Where the schedule is short or routine it is told to return few findings or none, because a list of twenty points nobody reads is worse than three that get asked.
A finding whose severity the model mislabelled is kept rather than dropped, at major. A point worth raising is worth asking even where the label was wrong, and discarding it would make the reading look cleaner than it was.
One detail worth reporting because the ledger caught it: re-reading the same schedule produces the same findings, and an append-only record refused a write whose only content was that somebody pressed the button twice. Refusing the command would have lost the caller a result they had already paid for. So nothing is written, the findings come back, and the result says plainly that it found nothing new — because “nothing new” and “nothing read” are different answers.
See the record being made rather than described.
Launch the demo environment