KAI · Investment Memo #0

Parsefield Ltd · pre-seed · £400k round · decision memo

Date: Day 1  ·  Verified as of: Day 1  ·  Engine: KAI v0.1 (the first real memos run on v0.2)  ·  SLA: applied → memo in 7h 18m  ·  Runtime: frontier model under the FundNow specification

RUN LOGApplication to memo

StageTimestampElapsed
Application complete (all evidence connections received)Day 1 · 09:020 min
KAI-SCREEN verdictDay 1 · 09:4745 min
KAI-VERIFY table completeDay 1 · 13:104h 08m
KAI-MEMO issued for signatureDay 1 · 16:207h 18m

SCREEN verdict (verbatim engine output):

{"verdict":"ADVANCE",
 "reasons":["In scope: UK AI-native B2B software at pre-seed with verifiable revenue",
            "Identity, incorporation and product existence verified at T1/T2 on application",
            "Coherent 18–24-month follow-on story in a funded category"],
 "kill_flags":[],
 "evidence_gaps":["Counterparty confirmation of largest contract","Reference calls pending"]}

1Thesis

Parsefield sells agentic contract-intelligence to mid-market insurers: software agents that read, check and flag policy wordings and binders that today are reviewed manually by outsourced teams. Six paying insurers in nine months, verified £8.7k MRR growing ~2.1x per quarter, sold by a founding pair that unusually combines the technical depth (Imperial NLP PhD, 8% university stake on clean USIT terms) with the market’s trust (eight years broking at Lloyd’s). The category has live budgets and visible follow-on appetite. Evidence quality is high for the stage; one material claim required correction (§3).

Composite7.15 / 10
VerdictRECOMMEND
Team8
Product7
Market7
Traction6
Clock fit7

Scores follow the v0.1 weights. Under v0.2, “Clock fit” is replaced by Capital & round dynamics, integrity becomes a gate rather than a weight, and the decision number is a calibrated probability against a published base rate.

2Verification summary

Twenty-one material claims assessed: 12 VERIFIED (T1/T2) · 4 ATTESTED (T3) · 2 UNVERIFIED (T4) · 2 DISCREPANT · 1 PENDING. Full table in the data room; the two discrepancies below, unabridged.

3Discrepancy register

MATERIALD1. Revenue overstated 8.7%. Application claims £9,400 MRR. Stripe export (T1) reconciles to £8,650: the £750 difference is a one-off pilot-setup fee from Halcyon Underwriting booked in the final week and presented as recurring. Founder response (logged): accepted the correction without dispute; stated the fee was “expected to recur” on pilot conversion. KAI treatment: memo uses £8,650 throughout; claim marked DISCREPANT→restated; no CRITICAL flag, because the underlying trajectory survives the correction. This is what verification is for.

NOTED2. Runway arithmetic. Application claims 7 months’ runway. Bank evidence (T2): £182k cash ÷ £31k monthly burn = 5.9 months. Founder counted a £40k Innovate UK grant instalment not yet received. KAI treatment: runway stated as 5.9 months verified, 7.1 months conditional on grant receipt (grant award letter sighted, T2; payment date unconfirmed → PENDING).

4The company

Agents ingest policy wordings, binders and endorsements; check them against underwriting guidelines and regulation; flag conflicts with citations. Sold per-seat plus volume. Live product (used by KAI during diligence, T1), 140 active seats across 6 insurers, largest signed contract £48k ACV with Bridgewell Mutual (document verified T2; counterparty confirmation PENDING). Repo: 2,140 commits over 9 months, activity profile consistent with a four-person team building in production (T1).

5Team

Dr. Amara Chen (CEO/CTO): PhD, NLP, Imperial College (thesis registry verified, T2); two first-author papers in the relevant sub-field; supervising professor’s attestation (T3, quoted in data room): “the strongest applied student I have supervised in a decade.” James Okafor (CCO): eight years placing specialty risk at Lloyd’s (employment history cross-verified T2; two market references T3). Two engineers, full-time, verified. The rare pre-seed combination: the person who can build it and the person the buyers already trust. The unique knowledge: which specific clauses cost insurers money when missed, and what an underwriter will pay to never miss them.

6Traction & economics

Verified MRR £8,650 (from £4,100 three months prior, ~2.1x/quarter, T1). Six logos, zero churn to date (9-month history, too short to lean on). ACV range £6k–£48k. Gross margin ~81% on current serving costs (T2, recomputed). Burn £31k/mo; verified runway 5.9 months → the raise is needed and honestly sized (£400k ≈ 12 months at planned burn).

7Market & the wedge

Mid-market insurers spend on manual wording review today; the displaced cost is real outsourcing invoices, not a hypothetical budget (two customer references confirm current spend, T3). Who dies first: the manual review vendors, not the incumbents’ software. Wedge risk: horizontal document-AI platforms moving down-market; defence is workflow depth and the Lloyd’s-trust channel, which generic platforms lack. Path to £10m revenue requires ~200 mid-market logos or Lloyd’s syndicate expansion: plausible, not assumed.

8The Clock plan

The 18–24-month question: can Parsefield reach a strong follow-on round? Verified trajectory (2.1x/quarter from a real base) reaching ~£40–60k MRR inside 15 months would place it firmly in range for UK/EU insurtech and applied-AI Series A appetite, which remains active in this category. Likeliest month-36 door: secondary into an A/B round (partial sale, capital returned, tail retained). Exit door plausible but not underwritten (consolidators acquire in this space). Roll door available if trajectory is strong but round timing slips. Clock fit: 7/10.

9Risks

  1. Concentration: top logo = 24% of verified MRR; a Bridgewell churn resets the trajectory. Resolved by: counterparty confirmation + renewal evidence (requested).
  2. Category compression: horizontal AI vendors bundling “good-enough” review. Resolved by: watching win/loss vs named platforms quarterly via telemetry.
  3. Founder-dependence in sales: all six logos closed by Okafor personally. Resolved by: first non-founder sale within 12 months (tracked as a calibration prediction).

10Calibration block: public, permanent

#PredictionPResolves
C1Verified MRR ≥ £25k0.45Month 12
C2Priced round at ≥ £12m pre-money0.50Month 18
C3Company operating (not wound down/acqui-hired under cap)0.80Month 24
C4≥1 of current top-3 customers churned0.30Month 12
C5First non-founder-closed sale0.55Month 12

In production every prediction is stated next to its base rate, Brier-scored as it resolves, and public-tier predictions resolve only on public events (v0.2 rule).

11Recommendation & terms

INVEST £120,000 via SPV on the offered SAFE (£4.2m cap), joining alongside the named lead, conditional on: (a) counterparty confirmation of the Bridgewell contract; (b) restated revenue figure acknowledged in the round docs; (c) live Stripe telemetry connection maintained post-investment. Entry at £4.2m cap against verified £8.65k MRR is full but defensible given team quality and category heat; the discrepancy episode raised confidence in process integrity: the founder corrected without friction.

12Human decision

DecisionSignerDateEngine agreement
(demonstration: no decision executed)Muhamad Kaissi, FundNown/an/a
Instrument note (v0.2 per-rail rule). This fictional deal is papered on a SAFE for demonstration. A real UK deal seeking SEIS/EIS would instead use an HMRC-compliant advance subscription agreement (≤6-month longstop, no investor protections) or priced equity, because the YC SAFE’s terms breach HMRC’s advance-assurance conditions and would forfeit backers’ reliefs. KAI v0.2 screens for this at intake.

This memo took KAI 7 hours 18 minutes from complete application to decision. Your turn: register for early access for a screen result in 48 hours and a decision within ten working days, on a published SLA, once underwriting opens. Backers: every memo like this, in full, on every deal.