Why can't SMEs negotiate favorable payment terms with large buyers?

Small suppliers sign 120-day payment clauses because they fear losing the buyer. Here's a contract copilot, "45 Days", that one person could build with Claude in a weekend.

The problem

Who feels it: Small manufacturers, job-workers and service vendors selling to corporates and PSUs. They pay staff and vendors in 30 to 45 days but get paid far later, and they rarely push back for fear of losing the account.

The law is already on their side on paper. MSMED Act Section 15 caps agreed credit at 45 days, and Section 16 adds compound interest at three times the RBI bank rate. One law-firm case describes a 120-day clause cut to 45 days within three weeks (Legal 500).
Buyers have a tax reason to comply. Section 43B(h) ties their deduction on micro and small supplier purchases to paying on time (ICAI journal).
Recovery after the fact is slow. As of June 2026, MSME Samadhaan had 256,892 applications worth Rs 55,244 crore, with 58,148 disposed of. An amendment bill passed by the Rajya Sabha aims to tighten timelines but was not yet law at the time of that report (BCR).
Market size signal. A 2022 study put delayed dues at over Rs 10.7 lakh crore, about 6% of GVA (ICAI journal). Treat it as dated.

Who's already working on it

Existing players cover the money, not the negotiation. TReDS platforms (RXIL, Invoicemart, M1xchange) finance invoices, but under 1% of registered MSMEs are onboarded (Whalesbook). Recordent handles buyer credit checks, reminders and legal notices (BizzBuzz). We didn't find a tool focused on the moment before signing.

The solution idea

"45 Days" is a copilot that reads a buyer's PO or contract before you sign, shows what the payment terms cost you, and drafts a polite counter-offer backed by the law.

Clause scanUpload a PO or MSA; it extracts credit period, start trigger, retention, set-off and acceptance clauses, and flags conflicts with the 45-day cap, with page references.

Cost of termsConverts "120 days" into rupees at your borrowing rate, compared with 45 days and with TReDS discounting.

Counter-offer drafterA relationship-safe email and replacement clause in English or Hindi, in firm, neutral or soft tone, framing 43B(h) as the buyer's own tax benefit.

Buyer leverage cardChecks whether the buyer likely falls under the TReDS mandate (turnover above Rs 250 crore, or a CPSE) and suggests TReDS settlement as an option.

Overdue ladderIf the invoice slips, calculates statutory interest and prepares a reminder, a notice draft and an ODR claim pack.

How Claude builds it

Which Claude capabilities do the work

Claude CodeBuilds the web app, database, tests and deployment.

Claude API in the productReads scanned and digital contracts, returns clauses as structured JSON, and drafts counter-offers.

SkillsAn "msme-payment-terms" skill holds the legal rules, clause library and tone guide, so every output follows the same checked source.

ArtifactsA landing page and clickable cost calculator to show SME owners before the app exists.

Connectors (MCP)Gmail and Google Drive during pilots to pull POs and draft replies for review; Vercel to deploy.

Scheduled tasksA weekly overdue-invoice sweep, and a watch on the amendment bill so the rules stay current.

Weekend build plan

Saturday morning — rules and test set: Collect 10 real or anonymised POs and MSAs. Write the skill with Sections 15, 16 and 43B(h) and have a CA or lawyer check it. Hand-label expected clauses as a test set.
Saturday midday — prototype: Ask Claude for an artifact with the landing page and calculator. Show it to three SME owners.
Saturday afternoon — scan: Use the starter prompt below in Claude Code to build upload, extraction and the results page. Run the test set and fix misses.
Saturday evening — calculator: Add cost of terms and statutory interest as plain code, with no model involved, and unit-test the arithmetic.
Sunday morning — drafter: Add the counter-offer drafter with three tones and two languages, plus the leverage card.
Sunday afternoon — ship: Deploy to Vercel, add login and auto-delete of uploads, and put an "information only, not legal advice" notice on each output.
Sunday evening — pilot: Onboard five suppliers through a CA or industry association. Set a scheduled task to review which counters were sent and answered.

Starter prompt for Claude Code

Build a Next.js + TypeScript + Postgres web app called "45 Days" for Indian small suppliers.

Flow: user uploads a PO or contract PDF -> Claude API extracts payment clauses as JSON
(credit_days, start_trigger, retention, set_off, acceptance_period, page_reference, quote)
-> app shows flagged issues -> user enters invoice value and annual borrowing rate
-> app shows the rupee cost at the contract term, at 45 days, and at a user-entered TReDS discount rate
-> user picks tone (firm/neutral/soft) and language (English/Hindi)
-> Claude drafts a counter-offer email and replacement clause.

Rules:
- Keep legal rules in /rules/msme.md (MSMED Act s.15 and s.16, Income Tax s.43B(h)), loaded into the prompt. Do not let the model cite anything outside that file.
- Every flag must show the exact quote and page from the contract. If a clause is not found, say "not found"; never guess.
- All money and interest calculations are deterministic code with unit tests, never model output.
- Ask for Udyam category. If the supplier is not micro or small, warn that these protections may not apply.
- Show "Information only, not legal advice" on every output. Delete uploads after 24 hours.

Start by proposing the schema and file structure, then build the upload and extraction step first.
Add a test script that runs extraction over /fixtures and compares it with expected JSON.

Monetisation

Suggested pricing, not yet validated.

Free: one contract scan a month with the cost calculation.
Rs 999 a month: unlimited scans, counter-offers and the overdue ladder.
Rs 2,999 per ODR claim pack.
Channel: white-label for CAs, plus referral income from invoice-discounting platforms.

Riskiest assumption, and how to test it

Assumption: The barrier is information, and an SME owner will send a counter-offer once they know their rights. Bargaining power and fear of losing the buyer may matter more than any tool.

Test first: Before building the full app, take 20 live or upcoming contracts through a manual version using Claude and the prompt above. Count how many owners send the counter and how many get better terms. A reasonable bar: at least 5 of 20 send and at least 2 improve. If owners won't send, pivot to the overdue ladder, where the relationship is already strained.

Whether the 45-day protections cover a given supplier depends on registration and category. Medium enterprises and traders may fall outside them, and the amendment bill could change the details. Check with a lawyer.