SB StartupBasket
All ideas
76 /100 GO Low complexity

FileWise — visa file auditor for study-abroad agents

Audits a student's visa-application file for missing, expired, and mismatched documents before the agent submits it.

EdTech SaaS India+Global Solo-builderWhatsApp-firstAI-agentSMBDocument-intelligence
— views
Evaluation Scores
76/100

GO

Overall Score

16
Problem
13
Demand
12
Build
12
Distrib.
11
Revenue
8
Time
4
Defense

FileWise — pre-submission file auditor for study-abroad agents

1. One-liner

Audits a student’s visa-application file for missing, expired, and mismatched documents before the agent submits it.

2. Trend signal — why now?

Three things collided in the last 12 months.

Student-visa refusal rates exploded. Canada rejected ~52% of study applications in 2024 and peaked near 65% in early 2025. US F-1 rejections hit 41–61% — a 10-year high. Australia hit one-in-three university applicants refused by Feb 2026. The UK’s refusal rate for several key source markets doubled-to-sextupled in the six months to March 2026. The destinations Indian agents send students to are now actively looking for reasons to say no.

The reasons aren’t exotic. Officials and university advisors name the same culprits over and over: incomplete or inconsistent documentation, outdated paperwork, mismatched names/dates across documents, and — the #1 UK refusal reason — failure to prove sufficient, properly-seasoned funds. “Parking” funds (a big deposit that appears two weeks before applying), a balance that dips below threshold on any day in the statement window, a passport that expires before the course ends, a name spelled differently on the transcript than the passport — these are the things that sink files. They are all mechanically checkable.

The money side hurts. Indian students lost ~₹664 crore in 2024 to applications rejected after fees were paid. 13.3 lakh Indian students went abroad in 2024–25. The agent who packaged the rejected file eats the reputational damage and the refund fights.

AI got good enough and cheap enough to read a messy WhatsApp dump of a passport scan, three bank statements, a transcript, an offer letter, and an SOP, and cross-check them against a specific country+university rule set — for a few rupees per file. That wasn’t true 18 months ago.

Provenance:

3. The opportunity

There’s a whole software stack for study-abroad agents — and none of it does this job.

The CRMs (SmartX, Meritto, Inausco, EduCtrl, Dhruvsoft, BoostedCRM) track the pipeline: which student is at which stage, which university replied. They do not read the documents or tell you the file will get refused.

The SOP tools (AbroBot, GyanDhan, Dr Shruti Mehtaa, CampusToolsHub, SOPCreator, Study Metro) check one document — the essay — and most are free. SOP quality is a real refusal factor, but it’s already commoditized and zero-priced. It is not a business.

Nobody owns the pre-submission file gate: take the entire packet and answer one question the agent is terrified to get wrong — “is this file going to get rejected on a technicality, and what exactly do I fix before I submit?” The expensive enterprise traceability/verification players don’t touch consumer visa files; the consumer SOP tools don’t touch the financial/identity/validity checks that actually cause refusals.

The 10× isn’t “AI writes a nicer essay.” It’s collapsing the 45–90 minute manual file-review a careful agent does (and a sloppy agent skips) into a 2-minute structured audit that catches the boring, deterministic killers — expiry dates, fund seasoning windows, balance-dip days, name/DOB mismatches, missing affidavits, gap-year explanation absent — that no SOP tool looks at.

4. Target market

  • Primary customer: Small and solo overseas-education agents in India — the 1–10 person consultancies and individual sub-agents who run on WhatsApp + Google Drive, not the Yocket/Leverage-Edu tier. Each handles ~30–300 student files per intake cycle. Secondary: the same profile in Nigeria, Pakistan, Bangladesh, Nepal, Philippines — all high-volume, high-refusal corridors.
  • Why they buy: “I submitted a file, it got refused on a stupid document mistake, the student blamed me, I refunded, I lost the referral.” A refusal is a direct hit to their reputation and their commission. They will pay to never feel that again.
  • Rough TAM reasoning: India alone sends 13.3 lakh students/yr through a largely unregulated agent network — thousands of small agencies and sub-agents. If 5,000 small agents pay ₹2,000/mo average, that’s ₹12 Cr/yr (~$1.4M) in India before touching other corridors. The corridor count (UK, Canada, Australia, US, Germany, Ireland) multiplies the rule sets, not the build.
  • Why now for them: Refusal rates roughly doubled across destinations in 18 months. The margin for a sloppy file vanished. Agents who used to “wing it” are now getting punished every cycle.

5. Product sketch (MVP)

  • Drag-and-drop (or WhatsApp-forward) the student’s document set — passport, transcripts, bank statements, offer/CAS letter, SOP, sponsor docs.
  • Pick destination + university + visa type. App loads the matching rule set.
  • Red/Amber/Green readiness report in ~2 minutes: each rule passed/failed with the exact reason.
  • Financial checks: required balance vs tuition+living for the right months; fund-seasoning window (no unexplained recent “parking”); no balance-dip below threshold on any day in the statement period; statement format/recency valid.
  • Identity/validity checks: passport valid through course end + buffer; name, DOB, parentage spelled identically across passport/transcript/offer/bank docs.
  • Completeness checks: every document the chosen corridor requires is present; gap-year/employment explanation present if there’s a gap; sponsor affidavit + relationship proof present if third-party funded.
  • Per-rule fix note in plain language the agent can forward to the student (“bank balance dips to ₹X on 14 May — keep it above ₹Y every day for the full 28-day window”).
  • Shareable PDF “file-ready” report the agent can keep as proof they checked.

6. AI angle — what’s load-bearing

Remove the AI and there is no product. The core is multimodal extraction (read a phone-photo of a bank passbook, a scanned transcript, a CAS letter in varied layouts) plus cross-document reconciliation (is “Priya S. Kumar” on the passport the same person as “Priyaa Kumar” on the transcript?) plus reasoning over a structured but fuzzy rule set (“does this funding story hold together for a UK Tier-4?”). The deterministic date/number checks are scripted, but getting clean structured data out of a chaotic real-world document dump is exactly the job that only became cheap and reliable with current multimodal models. A pre-LLM version would need an army of OCR-template-builders per document type per country — which is why this didn’t exist before.

7. Localization angle

India-first is the wedge, not a nice-to-have. WhatsApp-forward intake (agents live in WhatsApp), ₹-pricing tiers (₹999–₹4,999/mo where a $49 global tool can’t land), and Indian-document literacy (PAN, Aadhaar-linked statements, passbook-style bank docs, India-specific affidavit formats, NABARD/cooperative income proofs for rural sponsors). The same engine re-skins per corridor (Nigeria→UK, Pakistan→Australia) by swapping rule sets and local document formats — geography is the expansion path, not a rebuild.

8. Business model — path to $1M–$5M ARR

  • Pricing: ₹999/mo (solo, ~25 files), ₹2,499/mo (small agency, ~100 files), ₹4,999/mo (multi-counsellor, unlimited) — plus a ₹49/file pay-as-you-go for the long tail.
  • ACV: ~₹24,000–30,000/yr ($290–360) for the agency tiers; blended ~₹20,000 ($240).
  • To $1M ARR: ~3,500 paying agents at ₹24k/yr ≈ ₹8.4 Cr ≈ $1M. India has well over that many small agents.
  • To $5M ARR: ~15,000–18,000 agents across India + 4–5 other corridors, or land mid-tier agency chains (50–200 counsellors) on per-seat deals. Needs the corridor rule sets built out and a couple of chain logos.
  • Expansion path: more corridors → more seats per agency → adjacent files (dependent visas, post-study work, scholarship-doc audits) → a white-label “file-ready badge” agencies show students.

9. Go-to-market wedge — first 100 customers

  • Scrape the agent directories. India’s agent ecosystem is listed all over — IDP/British Council certified-agent lists, ICEF Agency directory, state-wise “study abroad consultants in ” SEO pages, and the 50+ India study-abroad Facebook/Telegram groups. Build a list of 2,000 small agencies with names + WhatsApp numbers.
  • Send a free audit of a real (anonymized) refused file. Cold-DM: “Forward me one file you’re about to submit — I’ll send back the red flags free in 5 minutes.” A refusal-traumatized agent will take that bet. Convert the wow into a trial.
  • Refusal-season timing. Hit them in the Aug/Sept and Jan intake crunch when files are stacking up and one refusal stings most.
  • Counsellor-influencer partners. A handful of India study-abroad YouTubers/Instagram coaches (the SOP-tips crowd) have exactly this audience; affiliate them on a per-signup cut.
  • “Refused-file” content. Tear down anonymized real refusal letters (“here’s the one date that killed this UK file”) — the agents share these themselves.

10. Build complexity — justification

Low–Medium. Off-the-shelf multimodal LLM for extraction, a standard web stack, and a per-corridor rule set that’s mostly config (dates, thresholds, required-doc lists) once the first country is done. No custom models, no proprietary data needed to start — the rules are public. The honest hard parts are (a) extraction reliability on garbage scans and (b) building/maintaining accurate rule sets per corridor as policies shift. A technical solo founder with a domain advisor ships India→UK + India→Canada in 8–12 weeks; each additional corridor is days-to-weeks of rule work, not a rebuild.

11. Gating checklist

GatePass?Note
Legal in target market✅A document-checking tool is legal; it must not fabricate/alter docs (that’s the dark-pattern line — see §13).
Ethical — no harm / dark patterns✅Helps submit honest, complete files. Hard rule: never generate or “fix” fake financials.
Market exists (evidence above)✅13.3 lakh students/yr, surging refusals, ₹664 Cr losses, dense adjacent software market.
1–5 person team can build this✅Off-the-shelf AI + config-driven rules.
Launchable with <$50K / ₹40L✅Solo/pair build; inference + hosting is the only real cost.

12. Feasibility score

AxisWeightScoreNotes
Problem intensity2016/20Hair-on-fire at refusal time; a refusal directly costs the agent reputation + commission. Not felt evenly every day, which keeps it off 17+.
Demand evidence1513/15Hard numbers: refusal rates, ₹664 Cr losses, documented top refusal reasons, a funded adjacent software market. Skeptic nods.
Build feasibility1512/15Off-the-shelf AI + config rules. Extraction on bad scans and rule-set upkeep are the real work, not blockers.
Distribution clarity1512/15Named lists (agent directories, FB/Telegram groups), a strong free-audit hook, seasonal timing. Conversion still unproven.
Revenue mechanics1511/15₹ pricing fits wallets; path to $1M is a believable agent count. Churn (seasonal usage) and willingness-to-pay vs free SOP tools are the risks.
Time to first revenue108/10Free-audit-to-trial-to-paid can close in weeks during an intake cycle.
Defensibility104/10Execution moat only. Rule sets are public; SOP-tool players or a CRM could bolt this on. Win is the corridor-rule library + agent workflow lock-in built over 12 months.
Total10076/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy · domain-expertise-required — needs solid multimodal/extraction engineering and someone who actually knows visa-file rules per corridor (ex-agent or visa officer as advisor).

Key assumptions to validate (3–5)

  1. Assumption: Small agents will pay ₹1,000–2,500/mo for a pre-submission check rather than rely on free SOP tools + manual eyeballing. How to test: 30 paid pilots at the real price during one intake cycle; measure renewal, not just signup.
  2. Assumption: Extraction is reliable enough on real WhatsApp-quality scans that the report is trusted, not second-guessed. How to test: Run 200 real (anonymized) files; measure false-pass / false-fail rate against an expert review. Need <5% misses on financial/validity rules.
  3. Assumption: The free-audit cold-DM converts at a workable rate. How to test: 200 cold DMs, measure file-forwarded rate and trial-start rate; need ≥10% forward.
  4. Assumption: Corridor rule sets can be kept current cheaply as policies shift. How to test: Track UK/Canada/Australia rule changes over one quarter; measure hours to update.

Risk flags

  1. Commoditization / incumbent bolt-on: SOP tools or CRMs add a “file check” feature. Mitigate by going deeper on financial/identity rules and owning the agent workflow before they notice.
  2. Liability / trust: If the tool greenlights a file that then gets refused, the agent blames the tool. Frame as “flags risks, doesn’t guarantee approval,” keep a clear confidence/uncertainty signal per rule, and never imply approval certainty.
  3. Ethics line: Agents may want it to manufacture compliant-looking financials. Hard no — refuse to generate/alter financial documents; that’s fraud and a brand-killer. The product checks honesty, it doesn’t fake it.
  4. Seasonality: Usage spikes around intake cycles → churn between cycles. Mitigate with annual plans and off-season value (re-checks, dependent-visa files).

14. Structured verdict

Score:                  76/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Technical founder (multimodal/doc-extraction) + an ex-agent or visa-rules advisor
Time to revenue:        6–10 weeks (close paid pilots inside one intake cycle)
Capital to launch:      ₹4–8 lakh ($5–10K) — mostly inference + hosting + a few months runway
Top 3 assumptions to validate first:
  1. Agents pay ₹1–2.5k/mo and renew — 30 paid pilots across one intake cycle
  2. Extraction <5% miss rate on financial/validity rules — 200 real anonymized files vs expert
  3. Free-audit cold-DM ≥10% file-forward rate — 200 DMs
Kill criteria:
  - Abandon if <20% of 30 paid pilots renew after their first intake cycle
  - Abandon if extraction false-pass rate on financial/validity rules exceeds 10% after tuning
  - Abandon if a free SOP/CRM incumbent ships an equivalent full-file audit before your India→UK+Canada v1 lands

15. Next step — 1-week validation sprint

  • Day 1–2: Build the agent target list (200 small India agents with WhatsApp numbers) from IDP/British Council/ICEF directories + 10 FB/Telegram groups. Hand-write the India→UK financial+validity rule set (it’s public).
  • Day 3–4: Manually (or with an LLM + me in the loop) audit 5–10 real anonymized files agents forward; send each a red/amber/green report. No product yet — concierge it.
  • Day 5: Cold-DM 50 agents the free-audit offer. Decide go/no-go on a falsifiable bar: ≥8 of 50 forward a real file, and ≥3 say “I’d pay ₹X/mo for this” with a number attached. If fewer, the pain isn’t worth paying for — and I’d rather know in a week.

Interested in a detailed proposal?

Get a deep-dive with market research, competitive analysis, and implementation roadmap.

Contact us

info@startupbasket.ai