The gap between the pitch and the site
Every software company selling to Indian contractors now has an AI story. Most of them look the same: a dashboard with a glowing neural network graphic, claims about "transforming project delivery," and a demo that shows a perfectly structured query returning a perfectly structured answer.
The contractors we talk to are not impressed by the demo. They are asking a different question: does it work on a 500-page NHAI tender document that was scanned from a physical copy, contains tables OCR'd incorrectly, and was drafted by a PWD office in 2019 using their own clause numbering?
That is the actual document. The question is whether AI can do something useful with it.
Where AI genuinely saves time today
NIT analysis. A typical NHAI or MoRTH tender notice runs 300 to 600 pages. It contains eligibility criteria, performance security requirements, earnest money amounts, completion periods, liquidated damages rates, special conditions, and dozens of other clauses that a bid team needs before deciding whether to bid.
Reading that document properly takes two to three days of a qualified contracts manager's time. AI-assisted extraction can reduce that to 20 to 30 minutes for the structured output: the obligations list, the key commercial terms, the risk flags. The contracts manager then spends their time reviewing the output and making judgements, not reading page 347 to find the LD clause.
This is not theoretical. On 16-criteria go/no-go scoring using a Shipley-model framework, the value comes from having structured data to score, not from the AI making the bid decision. The AI extracts; the BD head decides.
Formal letter drafting. A show-cause reply to an NHAI notice takes a contracts manager two to three hours: find the relevant GCC clause, check the notice period, match the tone to the situation, get the format right. An EOT application is longer. A variation claim is longer still.
AI-assisted drafting, given the contract document and the situation description, can produce a first draft in two to three minutes. The contracts manager reviews and approves. The total time drops from two hours to twenty minutes. On a project generating 40 to 60 formal letters per month, that is 60 to 80 hours saved.
WPI escalation calculation. Under NHAI GCC Clause 44 (and the equivalent MoRTH formula), price variation on RA bills requires pulling DPIIT sub-index data for labour, materials, POL, and plant, applying the correct weightages for the contract type, and calculating component-wise variation. Contractors using spreadsheets for this regularly make errors: wrong base dates, wrong component weightages, negative escalation miscalculated.
AI-assisted calculation, or a properly built WPI escalation calculator, eliminates the formula errors. The numbers are the same as a manual calculation. The time is 5 minutes instead of 90.
Where AI does not help
Judgement calls. Whether to bid on a tender where the employer has a track record of delayed payments is not a data problem. Whether to negotiate a particular GCC clause modification in the pre-bid meeting is not a data problem. AI can provide context (historical payment patterns, clause analysis) but the decision is human.
Relationship-based disputes. A significant portion of Indian construction disputes get resolved through negotiation, not arbitration. The relationship between a contractor's MD and an NHAI project director matters in that process. AI has no role there.
Complex arbitration. NHAI arbitration cases involving claims above ₹10 crore typically require a construction lawyer and a claims consultant. AI can help organise documents and draft submissions. It cannot replace the expert.
The practical question to ask
Before evaluating any AI tool for construction work, ask: what is the actual document, and what is the actual output?
If the answer is "a NHAI NIT" and "a structured obligations list with clause references", that is a real use case. If the answer is "any project data" and "insights", ask for a specific example.
The tools that save time in Indian construction work on Indian construction documents. The NIT from CPWD Lucknow, the GCC from 2018, the site diary in the format the engineer actually uses. Not a generic template.