Lessons from a contractor's closed jobs, shown to the bid team on new tenders
TL;DR Bidrecall turns a construction contractor's finished project records into a searchable memory. When the bid team prices a new tender, it pulls up what went wrong on similar jobs, like late approvals, design changes and failed subcontractors. Mid-size EPC and building contractors pay, because underpriced bids lose money for years.
Indian projects run late for reasons that repeat. A study of four completed hospital blocks at AIIMS New Delhi found every one overran its timeline, by roughly one to four years, with delays in statutory clearances, weak planning and poor communication among the causes[1]. A 2025 survey of Chennai residential projects ranked twelve recurring causes of cost overrun[2].
Contractors know this. But the knowledge sits in site engineers' heads, chat groups and closing reports nobody opens, and when a project manager leaves the lessons walk out the door with them.
So the estimator pricing the next hospital or flyover tender starts from rate schedules and not from what actually happened last time.
Bidrecall reads the documents a contractor already produces - daily progress reports, minutes of meetings, change orders, delay notices and final account statements. It pulls out events such as a three-month wait for a fire NOC or a piling subcontractor walking off site, and tags each by project type, state, client and cause.
When a new tender arrives, the estimator uploads the scope. Bidrecall shows the ten most relevant past problems, what they cost in days and rupees where records allow, and who on staff dealt with them.
It also prompts site teams with a short close-out interview at handover, recorded by voice in English or Hindi and transcribed, so new lessons keep coming in. AI handles extraction and matching, and people review every lesson before it enters the library. What the bid team gets is a warning list, not an automated price.
The managing director or head of estimation at a contractor with ₹200–2,000 Cr turnover pays (my estimate of the target band). Estimators, planning engineers and project managers use it. I found no Indian product sold for this job at a public price. My guess is that such buyers might accept ₹8–20 L a year per firm, plus a one-time charge to load past projects.
Language models can now pull events out of scanned, mixed-language site reports at low cost, which wasn't practical a few years ago. The PMI's project management guide treats lessons learned as a standard project output[3], so public and private clients increasingly expect contractors to show them. Large Indian contractors, the biggest with revenue near ₹2.9 lakh Cr[4], set the bar that mid-size firms get pushed to match.
I found no public figure for what Indian contractors spend on knowledge-management tools. The narrowest signal is the pain itself, since four of four AIIMS projects in one study ran over time[1].
Adjacent - a UK project-controls company sells AI schedule analysis to large engineering firms and rail operators[5]. That shows contractors will pay for software that learns from project data.
That UK vendor analyses schedules and critical paths but doesn't describe a lessons library[5]. One professional-services software vendor markets an engine that learns from past projects, but for consulting firms and not builders[6]. Indian site apps cover daily progress and materials. Large contractors keep internal lessons databases that hardly anyone opens at bid time.
The gap is a product that puts past lessons right in front of the estimator.
| Difficulty | Moderate (36/100) |
| Starting capital | ₹70 L–1.4 Cr |
| Time to launch | 9–12 months |
Extraction works fine on clean documents, but Indian site records are scanned, handwritten in places and all over the place. Loading a firm's past projects is manual work you will underprice at first. And selling needs a champion senior enough to change how bids get prepared.
No licence needed. Software for contractors has no sector permit, so the baseline applies with strict data separation.
Building this outside India? The categories above usually have a local equivalent, so check with your own regulator.
Don't let AI set a bid price or add lessons unchecked. People review every lesson, and the estimator decides the number.
Want to work out the AI layer for Bidrecall? Talk to Yash →
Here's what can go south -
I think it's worth a shot for a founder who has worked in estimation or project controls at an Indian contractor and can get past projects loaded. If you're a pure software founder without industry access then skip it. First prove that an estimator changes a bid price or a contract clause because of a Bidrecall warning.
Plan for roughly ₹70 L–1.4 Cr, much of it for loading past project records and a long sales cycle with senior estimators.
No sector licence. You need company registration, GST, DPDP compliance and contracts that keep each firm's records separate.
Yes, contractors everywhere repeat old mistakes. Check local data protection law and contract confidentiality terms first.
The free Bidrecall validation kit has the line-by-line capital breakdown, the effort scoring, the contract and data checklist, a 14-day validation sprint with pass/fail targets, a 90-day launch plan, 10 interview questions for estimators and project managers, and 5 assumptions to test. Download the kit →
Thinking about building Bidrecall or something close to it? The make-or-break work is getting a contractor to hand over years of messy site records and then actually use the result at bid time, and that's the kind of thing I help founders plan for. If you need a hand, you know where to find me. Talk to Yash →
For illustration only. Names are fictional and ideas hypothetical; sourced facts are real but may be out of date. Nothing here is legal, financial or professional advice, and Yash Khatri is not personally liable for any use of it. Read the full disclaimer.