XMarts room

Summer 2022

Rima AI

Document and spreadsheet automation for accountants

Visit website

What they do

Document and spreadsheet automation for accountants

How close to us

5, very close

One reusable automation layer sold to accountancy firms, keeps know-how, audit tracking; active small firm direct W5 match.

Wedges

  • W5 One layer for one type of small firm direct
  • W8 The other 90% of an AI skill adjacent
L1 Company brain (knows the organisation) 1, light
L2 Agent team (runs agents, tools, approvals) 1, light
L3 Does and learns (finishes, checks, measures, updates the brain) 2, clear

Who buys

Firm size
sme
Buyer type
business
Role
tax resolution firm owner or case manager; accounting firm owner
UK
no
Their own description (from YC)

Rima helps accountants automate document and spreadsheet work with accuracy and cell-level audit tracking.

For accountants that deal with messy docs and spend their time in Excel, Rima enables them to build workflows in natural language to extract and process financial data from PDFs or ERPs into their Excel templates and reconcile their spreadsheets in minutes. Accountants build these workflows once and our agents do the work for them going forward, preserving their institutional memory and know-how.

Full teardown

Rima AI (YC Summer 2022): teardown vs XMarts

In one line: Rima sells AI that reads client documents and fills in forms and spreadsheets for small accounting firms, and today sells mostly to US tax resolution firms (IRS debt cases) [site] [yc]. Threat: 2, because it is the same pattern as W5 (one layer for one type of small firm) with the same kind of buyer, but it is tiny, US-only and stops at document work [site] [yc].

1. Fact file

Item Finding Source
What they say Home page: "AI for Tax Resolution Firms". YC line: "Document and spreadsheet automation for accountants". [site] [yc]
Founded, team, funding Founded 2021, YC S22, 5 people, Austin TX. Funding: no primary source. One tracker lists about $600K pre-seed, another a $125K seed, from third-party data. UNVERIFIED. [yc] [docs]
Founders Gwanygha'a "Gana" Gana (EE and data science degrees from Georgia Tech, Harvard MBA, ex-founder) and Cedric Foudjet (ex-founder and VC in Africa and Southeast Asia, self-taught cloud engineer, Wharton MBA). What YC likely saw: two repeat founders who had run a wholesale business with over 4,000 small customers, so they know document-heavy back offices (inference from the launch post). [yc] [launch]
Customers Home page and case studies name no customer. One firm with 300+ active IRS cases (they claim). A second "mid-market" tax resolution firm. About 500 cases a month for one onboarded firm (they claim). Press named one accountancy customer: Sunbridge Advisors. They claim over 100 accountants use it (May 2026). [site] [press] [launch]
Price Not public. Demo-led. A "single-user license" (one user, up to five clients) and a one-time use option (one case, one run) exist for solo agents. No amounts. Jobs page shows 0 open roles. [site] [yc]
Path Garage, wholesale auto-parts marketplace for African retailers (YC S22) → Rima, accountant document agents, launched May 2026 → home page now tax resolution only; the old accountant pages return HTTP 410 (gone). [yc] [press] [site]

2. Their proposition, reverse engineered

Part What they do
Who buys Tax resolution firms, CPA and tax teams, solo enrolled agents (US) [site]. Earlier: accounting firms and in-house finance teams [launch].
The pain Staff spend hours keying numbers from bank statements, transcripts and bills into IRS forms. "They aren't practicing accounting. They're acting as data janitors." [press]
The promise "Move your team from data entry to case review." [site] Drafts for staff to check, correct and approve [site].
How it works Upload messy documents. An agent fills the form using the firm's rules and edge cases. A checklist flags gaps. Staff review against the source and approve. Every task is logged [site]. Earlier version: Rima interviews an accountant, captures edge cases, builds a reusable "Blueprint" workflow [launch] [press].
The output A form draft (433-F, 433-B, 433-A OIC, 1040, P&L) with flags and an audit trail [site].
Proof style Time per document. 90 min to 30 min per 433 form; a P&L from a week to 2 hours; 300+ cases. They claim all of these. No money result [site].
Who sells it Founder-led demos (book-a-demo link). Blog and guides for search. Five "Rima vs" pages posted 25 Sep 2026 [site] [blog].
Price model Not public. Licence per user and per case hints at per-seat plus per-run (inference) [site].
The story Generic AI gets "55% of the way there" and lacks each client's edge cases; the fix is written rules, not a better prompt [launch] [site]. People keep the final judgment [site].

3. Positioning

  • Category they claim: AI for tax resolution firms; before that, "the first AI accounting agent" [site] [press].
  • Category a YC partner would put them in: vertical AI for back-office document work, now in one niche (inference).
  • What they want to be known for: the safe, audited way to turn client documents into a reviewable form draft [site].
  • Where they sit on the map: L1/L2/L3 = 1/1/2, buyer = sme, business, US.

4. Three layers vs ours

Layer Them (score + evidence) Us (Marton's logs show)
L1 Company brain 1. Firm rules and edge cases are written into each workflow and reused across clients [launch] [site]. The guide says rules are written down, and no memory feature is described [docs]. A vault of about 1,400 pages, 112 call transcripts and about 8,700 customer quotes, searchable by meaning.
L2 Agent orchestration 1. One data-entry agent per workflow plus a review queue with named approvers [site]. No team of agents is claimed. About 5,600 jobs run by agents from three makers, each checked by a different AI before merge.
L3 Execution and learning 2. Does the work, then checks it: gap flags, plausibility checks, source tie-out, human approval, audit trail [site] [docs]. Better "after each workflow" is a customer feeling, not a measured loop [launch]. About 160 failure records, each with an automatic check. Offer feedback lines fell from 44 to 18 as 22 rules were added. Only about 15% of jobs reached main, and merges were unreliable in October.

5. Same and different

Topic Same or different Them Us
Core belief Same Generic AI misses a firm's edge cases; written rules fix it [launch] [docs]. Same idea: rules added from failures and feedback (Marton's logs show).
Buyer type Same Small accounting and tax firms [site]. Target in brief: SMEs and scale-ups; W5 example is accountancy firms.
Owner of the result Different Staff approve every draft; the firm owns the outcome [site]. Aim: own one number for the client (brief). Not proven.
Kind of work Different Fills forms from documents. One step in a case [site]. The wedge covers a whole layer of a firm's work (brief).
Country Different US only: IRS forms and the Internal Revenue Manual [site]. UK mid-sized firms (brief).
How it captures know-how Different Interview with the accountant, then a built workflow [launch]. Learns from calls and from checked results (Marton's logs show).
Size Different 5 people, no named customers on the site [yc] [site]. Two founders, no outside customer yet.
Delivery Same Founder-built, per-firm setup [launch]. Founders plus the system.
Proof Different Minutes saved per form, claimed [site]. No customer number yet.
Price Unknown Not public [site]. Not set.

6. Wedge overlap

Wedge Fit Why
W5 One layer for one type of small firm direct Same model: one reusable layer for one kind of firm, built once and sold to many. Same buyer type (small accounting and tax firms). They chose US tax resolution; our example is UK accountancy (payroll, VAT, forecasts) [site] [brief].
W8 The other 90% of an AI skill adjacent They say rules need an owner and review when they change, and write rules as reviewable items [docs]. That is skill upkeep, but only for their own workflows.

7. Whitespace (where they stop)

  • Layer gap: they stop at document-to-form drafts. No month-end close, VAT, payroll or forecasts, and no agent team [site]. To own it we need one firm's recurring workflow run end to end by our agents for 4+ weeks.
  • Buyer gap: US tax resolution only now. UK and European accountancy, and any firm not doing IRS collection cases, is unserved by them [site]. To own it we need UK accountants who say they have the pain.
  • Number gap: their numbers are minutes per form, claimed, with no named customer [site]. To own it we need a before/after number a firm partner would sign, in money or hours per month for the firm.
  • Price and speed: not enough evidence either way (no public price [site]; no onboarding time found). Not listed.

8. For the YC application

  • Proven today: Marton's logs show his system runs agents that do work and check it: about 5,600 jobs, each checked by a different AI, with about 160 failure records that each added an automatic check. Rima's public results are minutes saved per form, claimed, with no named customer [site].
  • Draft answer to "how are you different from Rima?": DRAFT. "Rima fills in forms for US tax firms from client documents. We are going for the whole recurring workflow of a UK accountancy, with agents that check their own work and learn from each fix." Founders say it their own way.
  • Future claim (needs proof): "We run the monthly work for a UK accountancy and own one number." Needs one firm's workflow run for 4+ weeks with a before/after number, plus the firm's consent to use it.
  • What we understand that they don't: a firm pays for the finished month, not for a faster form. needs proof (no UK accountant has used the system; nobody but Marton has run it).
  • Copyability (M3): medium. The tech is easy for them: they already have rules, review queues and audit trails [site]. But a UK move means new rules (VAT, payroll, UK filing), a new market and a 5-person team that just narrowed to one US niche [site] [yc] (inference). They would not likely chase a wider layer soon, but a partner could say they could. Not hard, so M3 cannot score above 3 against Rima alone.
  • Why now (M4): None found in this teardown. Do not use "agents can now do the work": that is general AI progress.
  • Do not claim: that Rima is "only a document tool" without saying it keeps firm rules and audit trails. "No one does accountancy workflows" (Rima did, until it narrowed). That its pivot proves the accountant market is bad: the reason for narrowing is UNVERIFIED.

9. Ideas worth taking

  1. Narrow hard to one form and one buyer, with a "Rima vs X" page for each rival. Cheap way to meet a buyer who is already searching [blog].
  2. Write each firm rule as Situation, Rule, Prohibition so a reviewer can read it. Easy for a buyer to trust [docs].
  3. A cell-level or field-level link back to the source page. Buyers in regulated work ask for it [docs] [site].

10. Unknowns

  • Why they dropped the accountant pages (410 gone) and went to tax resolution. Check the Wayback Machine for the old pages and their dates.
  • Real price, real customer count, and whether the 300+ cases come from one firm or several.
  • Funding: no primary source. Look for a SAFE or seed announcement.
  • Whether the "interview an accountant" onboarding still exists in the tax product. The new product pages do not say how workflows are built [site].
  • The SOC 2 wording conflicts between pages ("in observation" on one, "Type II certified" badge on another) [site].
  • Dates of the February 2026 blog posts (the same date on five posts suggests a bulk upload) [blog] (inference).

Sources

Technical detail
Slug
rima-ai
Lens hash
179096c479
Pass
deep
Stale
no