Quarri - AI-native operations for mid-market industrials

Quarri

AI-native operations for mid-market industrials.

Problem

The optimisation problem.

Supply in

Purchased / extracted

→

Upstream inventory

Pre-transformation

→

Transformation

Manufacturing

→

Downstream inventory

Finished goods

→

Customer POs

Sales pipeline

Regulatory reporting

e.g. stumpage returns

AP / vendor payments

Suppliers, extraction fees

AR / accounts receivable

Customer collections

Extraction and manufacturing — multifaceted and complex optimisation.

Stock-outs

Lost revenue

Cash flow

Working capital tied up in inventory

Aged inventory

Carry cost erodes margin

Throughput vs lead time

Production vs input availability

SKU margin

Invisible across systems

Account margin

Customer-level profitability

The Gap

Mid-market lacks the tools to optimise.

SMB

Mid-Market ($10–200M)

The Gap

Enterprise

Products

NetSuite · QuickBooks

Excel + legacy ERP

SAP · Snowflake

Customisation

None — off-the-shelf is enough

None affordable → Excel workarounds

Palantir, in-house data teams (~$400k+/yr minimum)

Optimisation

Simple — not required

Stuck in Excel

Custom data teams

94%

Spreadsheet error rate

65%

Analyst time on data gathering

A3 - The Gap

Mid-market lacks the tools to optimise — and Excel is filling the gap.

It’s not a bug — it’s a feature. Rigid SaaS forces Excel workarounds; the fix needs a data team mid-market can’t justify.

SaaS is rigid — switching costs are high

Mid-market ERP implementations cost $150–500k with an average 189% cost overrun (Panorama 2025). 30–40% of total IT spend goes to shadow workarounds (Gartner).

So businesses default to Excel

70–90% of firms still rely on spreadsheets across finance and operations (AutoRek 2025, n=500).

But Excel is slow, fragile, and error-prone

94% of business spreadsheets contain critical errors (Frontiers of Computer Science, 2024). Teams spend 65% of their time on data gathering, not insight (FP&A Trends 2024, n=379).

The answer exists — but it’s prohibitively expensive

Consolidate, clean, warehouse your data. Well understood. But it requires $400k+/year before any output, plus constant maintenance. Mid-market simply can’t afford it.

By the numbers

189%

Avg ERP cost overrun (Panorama, 2025)

94%

Spreadsheets w/ critical errors (FCS, 2024)

65%

FP&A time on data, not insight (2024)

$400k+

Annual cost of an in-house data team

The Stack

Deterministic toolkit. Agentic orchestration.

Horizontal AI General-purpose LLMs

Anthropic

OpenAI

Open source

Model-agnostic surface. Skills ship to whichever LLM the customer uses.

Quarri · An MCP plugin

PRODUCT LAYER

Skills

Cash flow Account margin Throughput vs lead time Inventory optimisation

INFRASTRUCTURE LAYER

Deterministic execution

100+ MCP tools Versioned workflows

Context layer

Semantic model Vector store Company glossary Memory of work

Data layer

Data Warehouse Agentic modelling Company schema RBAC Read + write

Connectors & ingestion

Legacy ERP Spreadsheets Databases PDFs External

Deterministic infrastructure.

Persistent · Cheaper · Faster · Contextualised · Vastly larger data.

A4 - The Stack

Deterministic toolkit. Agentic orchestration.

Mid-market doesn’t need more verticalized SaaS. It needs to harness cutting-edge LLMs on the messy reality it already runs on.

< 2 weeks to go-live impact

Weeks 1–2

Connect data — ~0.5 day per legacy source. Absorbed in SaaS fee.

Weeks 3–4

Live dashboards, NL queries, benchmarking. Absorbed in SaaS fee.

Month 2+

Optional paid professional services for workflow automations.

With Quarri

Automate. Optimise. Analyse.

01 Workflow automations

02 Optimisation

03 Analyse

Market

$52B global market. $2.25B serviceable today.

$52.2B

Global TAM

Mid-market industrials · GDP-scaled

$28.1B

4-geo TAM

US + EU + UK + Canada · bottom-up

$10.7B

SAM

Target industries · SaaS >74% · ~81% GM

$2.25B

SOM · Addressable now

SaaS >65% · ~79% GM

Key SOM verticals

Forestry & Lumber

Existing customer base · 60% SOM

Manufacturing

Target vertical · 60% SOM

Wholesale & distribution

Target vertical · 60% SOM

Bottom-up: US Census + Eurostat + ONS + StatCan firm counts × Quarri ACV in 4 geos. Global TAM extrapolated by IMF 2025 GDP. SOM modifiers: US/UK/CA 1.0× (signed deals), EU 0.1× (regulatory/language friction). Russia excluded.

Traction

3 months post-MCP.

Research & Development · months 1–6

No target verticals

100+ product interviews across industries
Hand-cranked pilots
Stress testing pricing
Developing infra
No live deployments

Commercial proof · months 7–9

$47k ARR

Working pipeline

$130k

Contract expansion / pilot conversion

$300k+

Qualified pipeline

3

Anchor customers

Pipeline concentration: Forestry, Mills, Manufacturing

Concentration: 1 anchor account currently 77% of ARR. Q3 2026 target: 5+ accounts, no single account >25%.

Revenue mix & margin

SaaS 75% 90%+ gross margin

Services 25% 40% gross margin

Already SaaS-dominant. Services is the wedge into mid-market; agentification pulls services margin up over time → see The Flip.

A6 - Traction & Pricing

Three live customers, four pilots. $24k average land contract.

3 live paying customers and 4 active pilots / POCs at an average land contract of $24k. Clear path to expansion across 5 of 7 via additional services, seats and connections.

3 live + 4 pilots

Paying customers + active POCs · 7 contracts total

$24k

Average LAND contract value

5 of 7

Clear expansion path via services, seats, connections

Pricing tiers — the land

Annual contract, billed monthly · dev days are an annual entitlement that resets each contract year.

Essentials

$500/mo

3 sources · 3 users · Guided setup

Scale

$1,200/mo

5 sources · 5 users · 4 dev days / year

Business

$2,000/mo

7 sources · 10 users · 10 dev days / year

Corporate

$3,000/mo

10 sources · 25 users · 15 dev days / year

Share of wallet — the expand

Onboarding starts at Scale tier ($1,200/mo). Within 6 months customers add sources, users, and services — driving expansion. Blended margin steps down as services mix rises; absolute margin grows.

Unit economics + per-tier margin in data room

A6b - GTM & Distribution

Forestry & Lumber — a wedge into mid-market industrials.

Beachhead profile

50–500

Employees

$10–100M

Revenue

CEO / CFO

Buyer + Champion

Legacy ERP

Operating Systems

Forestry & Lumber → replicates across mid-market industrials

Forestry & Lumber beachhead		Mid-market industrials
Forest management	→	Resource extraction · allocation · contract fulfilment
Tree length & lumber transportation	→	Supply chain & logistics · bills of lading · delivery reports
Lumber mills	→	Processing & manufacturing · maintenance · inventory

GTM — PE distribution

Live

Active JV conversation with a leading North American PE fund for portfolio-wide Quarri deployment. The fastest mid-market distribution channel: one decision-maker, dozens of portfolio companies.

Tier	Fund size	Portfolio company	% PE raise	Quarri serves
Small Cap	< $500M	$10–300M	13%	Yes — beachhead
Mid-Market	$500M–$5B	$100M–$1B	52%	Yes — primary
Large / Mega	$5B+	$500M–$10B+	35%	Anthropic / OpenAI / Palantir

Why mid-market · ~1M+ boomer-owned businesses (~$5T) entering PE need professionalising. 75% of PE deals are bolt-ons, each creating multi-entity reporting needs. Sources · McKinsey GPM 2026, PitchBook US PE Breakdown 2025.

Market Context

Why foundational models won't build this.

100× cheaper per execution

LLM only

$1

per invoice · probabilistic, 300+ pages of context

Quarri tool

$0

per invoice · deterministic, cached

Every $1 of compute Quarri saves
is $1 of lost revenue for the horizontal AI provider.

Model companies will never ship a cheaper tool — even from their own deployment teams.

Plus · no model lock-in Quarri is the deterministic layer. Skills and tools lift-and-shift to whichever LLM or harness wins next.

A7 - Why Foundational Models Won’t

LLM vendors are paid for tokens. Quarri is paid for impact.

Economic misalignment on AI deployment: LLMs want token usage and lock-in. Operators want lower cost and no lock-in.

LLM economic incentive

LLM vendors charge by the token. Product strategy follows revenue: deep-research modes, agent loops, extended thinking. These maximise tokens per question. An agent loop solving a complex workflow can burn 100k+ tokens per run.

Quarri economic incentive

Quarri is paid based on the impact it delivers. Every workflow we encode shortens the loop, caches the answer, and routes the next call to a deterministic tool. Customers get a faster, cheaper answer. We capture a share of the economic value the workflow delivers. Aligned incentives.

No vendor lock-in — this helps Quarri

MCP is an open standard. Quarri’s 100+ tools work for any agent that speaks MCP — Claude today, GPT or Gemini tomorrow, an open-source harness next.
Customers ride the LLM frontier. When a more powerful model or harness lands, customers benefit without re-platforming. They never marry one vendor.
Pan-agent data plugins. The deterministic + permanent layer becomes the substrate every AI tool in the org calls into — not just one chat product.

Head-to-head

	Quarri	Palantir	Fivetran + dbt + Hex	LLM (raw)
Time to value	< 2 weeks	6–18 months	3–6 months	Per session
Cost	$6–70k/yr	$1M+ implementation	$400k+/yr (data team)	Token cost only
Needs a data team	No	Forward-deployed engineers	Yes	Yes (to be useful)
Vertical depth	Forestry & industrials	Enterprise only	Generic	Generic
Permanence (memory)	Yes — encoded in tools	Yes (custom builds)	Stored in code	None

A7b - Competitive Landscape

Two markets exist. Neither serves mid-market.

Data tools need a technical team. Automation tools were built for API-native SaaS — not the legacy systems mid-market actually runs on.

Data Stack · fragmented by role, not outcome

Layer

Tool

Tech team required

ELT / Ingestion

Airbyte · Stitch · Whalesync

Data Engineer

Warehousing

Snowflake · BigQuery · DuckDB

Data Engineer

Modelling

dbt · Cube

Analytics Engineer

Visualisation

Tableau · Looker · Omni

Data Analyst

Analytics

Hex · Python / SQL

Data Scientist

Quarri replaces the entire stack — no technical team required.

Automation Stack · built around modern SaaS, not legacy

Quarri

Seed/A

B-C/Acq

Series D+

Listed

Bubble ~ valuation / total raised

Newly funded entrants

Ciridae

$20M

Seed · May 2026

Bespoke AI agent deployments for mid-market. Forward-deployed engineers.

ciridae.com

Deployferry

~$1M

Seed

AI agents for industrial operations across manufacturing & physical-asset businesses.

deployferry.io

Netter

YC seed

Y Combinator

Centralises scattered mid-market data into one ontology; deploys analytics, workflows and ML via chat.

netter.ai

Example Workflows

Three things Quarri enables that AI can't do alone.

4 hrs → 2 min

Forestry organisation · Automated reporting

Weekly reconciliation replaced. Same workflow, encoded once.

$300k

Forestry organisation · Mis-accrual caught

Mis-accrual flagged in pilot week 1, missed across 8 months of invoices.

$60k

Marina operation · Pricing exception

Pricing exceptions surfaced across 1,000s of line items — margin unlocked.

This has saved me days of work and is actually accurate versus what we were doing.

Operations Lead

Forestry customer, North America

A8 - Example Workflow

Stumpage accrual & release.

Faster Mistakes eliminated Logic preserved

$200M+ forestry business. Live on Quarri in pilot week 1 — caught $300k of miscounted accrual and now runs end-to-end in 5 minutes per month.

Data sources

Operations database — volumes harvested by site, species, period
Government tax rates — stumpage rates by region, by counterparty
Government invoices — PDFs, parsed and matched
15 Excel tabs — previously the source of truth, retained for traceability

Mapping system — 7 logic steps + 3-tier waterfall

Quarri built the full mapping framework end-to-end: location, department, and government counterparty mapping, encoded as a 7-step logic chain resolved by a 3-tier waterfall decision tree. Versioned, named, audit-ready.

Accrual + release with smart matching

Quarri assigns the stumpage at the time of harvest, then unwinds the accrual when the matching government invoice arrives — sporadically, sometimes many months later. A smart matching system reconciles accrual to release across periods, flags variances, and preserves the original judgement.

Outcome

$300k

Over-accrual caught in pilot week 1

100+ hrs/yr

Manual, judgement-heavy time replaced

5 min

Per month end-to-end, including release

“I’ve done more with AI in the last 6 weeks than I have in the last 6 months.” Head of Operations · Forestry customer, North America

Why Now

Every AI tailwind.

MCP shipped

Feb 2026 · the mid-market stack became buildable. Quarri shipped 100+ tools in week 1.

App store inevitable

Distribution flows through LLMs. Build for them, not against them.

Vertical AI on foundational models

Inject verticalisation into general-model power. Don't lose horizontal AI — extend it.

Build for LLMs. Don't build against them.

A9 - Why Now

Mid-market AI only became viable in early 2026.

The infrastructure to deliver agentic operations to mid-market only became powerful enough ~3 months ago. Quarri shipped on day 1 of the inflection.

Mid-market is falling behind

60%

AI projects fail pilot → production (Gartner 2025)

87%

Say AI critical to operations (Deloitte ’25)

54%

Prioritising AI agents (Deloitte ’25)

LLM milestones — the inflection

Nov 2024

Anthropic launches Model Context Protocol (MCP)

Open standard for agents to call structured tools. The plumbing arrives.

May 2025

Claude opens user-facing custom MCP integrations

End users can plug their own MCP servers into Claude. Quarri now has a distribution path.

Sep 2025

Claude ships file creation — Excel, PowerPoint, Word, PDF

In-chat workplace artefacts. Outputs land in the tools customers already use.

Oct 2025

OpenAI Apps SDK launches at DevDay (preview)

ChatGPT opens to third-party apps with inline iframe rendering — closed preview.

Oct 2025

Claude Skills framework formalised

Reusable, named, versioned skill packages. Our 100+ tools become first-class citizens.

Feb 2026

Claude Opus 4.6 launches — the inflection

1M-token context, dramatic step-up in agentic reasoning. This is when the workplace-skill execution layer became powerful enough. Quarri shipped in week 1.

Sources · anthropic.com/news (MCP, Integrations, Skills, Opus 4.6), openai.com/index/introducing-apps-in-chatgpt.

Market opportunity

$52.2B

Global TAM · GDP-scaled

$28.1B

4-geo TAM · bottom-up

$10.7B

SAM · Target industries

$2.25B

SOM · Addressable now

Geography	TAM	SAM	SOM	SOM modifier
US	$11.4B	$3.9B	$1.4B	1.0× · signed deals
EU-27	$13.5B	$5.5B	$0.3B	0.1× · language/regulatory friction
UK	$1.6B	$0.7B	$0.3B	1.0× · home market
Canada	$1.6B	$0.6B	$0.3B	1.0× · signed deals

Bottom-up · US Census Bureau (2022 SUSB), Eurostat SBS, ONS Business Demographics, Statistics Canada (33-10-0717). Filtered by industry inclusion matrix and 50–500 emp band × tiered Quarri ACV. Global TAM extrapolated to 32 additional economies via IMF 2025 nominal GDP at $496 / $M GDP (bottom-up rate). Russia excluded.

Market sizing model (v2.1) in data room

The Moat

Moats compound outward.

Vertical data set

Opt-in benchmarking + optimisation. External data sets.

Skills corpus

Productised customisation; lifts and shifts to new customers and becomes industry-specific insight.

Context layer

Semantic model learns from previous deployments.

Data layer

Agentic modelling skillset deepens.

Connectors

Built once — less lift every new client.

A10 - Tools & Skills

100+ tools. Skills mapped to the optimisation problem.

Every Quarri skill is composed from a deterministic data + context layer and 100+ MCP tools the LLM invokes autonomously.

Optimisation challenge

Data locked in ERP / Excels / PDFs

Quarri product / skill

Named, versioned, composable

ERP · Supplier feeds · POS

→

Inventory optimisation

AR ledger · AP ledger · Bank feed

→

Cash flow

ERP inventory tables · Warehouse logs

→

Inventory ageing

Production schedules · Ops DB · Maintenance logs

→

Throughput

ERP cost tables · Freight invoices · Sales channels

→

SKU profitability

Customer master · Pricing/rebate tables · Service tickets

→

Account margin

Ops DB · Govt tax tables · Govt invoice PDFs · 15 Excel tabs

→

Accrual & release

Vendor statements · GL · Supplier portals

→

Systems reconciliation

Data flow — sources, tools, skills

Indicative weighting — not actual tool count

All skills and tools are designed to fit and orchestrate together — more than the sum of their parts.

The Flip

Quarri will run deployments solo.

Three forces drive services-to-software economics.

Margin trajectory

Services GM Blended GM

As services agentify, margin tail rises toward SaaS economics.

Growing data corpus

Every deployment is captured — call scripts, emails, workbooks, final automations.

Trains the deployment agent.

Horizontal AI gets more powerful

Foundational models improve. Our skills ride the tailwind.

Cheaper compute, smarter orchestration.

Skills become reusable products

First deployment of a skill is custom work. Every next deployment is lift-and-shift.

Skill creation cost falls. Margin recovers.

Three flywheels: data, model, product. All compound to margin.

The Team

Operator + architect.

Theo Leslie

CEO · Operator

Director of Strategy at Worldpay — $100m+ ARR product launches
VP Growth at fintech (Red Sky) — built lending product 0 → $500k ARR
Founding team for delivery division at major mid-market hospitality group
Chartered accountant (PwC)
Lived the Excel-heavy data pain for 10+ years

David Jayatillake

CTO · Data architect

3x founder (1 exit — Delphi Labs acquired by Cube)
Former VP of AI at Cube (semantic layer)
Founded 2 semantic layer startups
Leading voice in the semantic layer space (high-profile thought leader)
Data leadership at Lyst and Worldpay
20 years in data infrastructure and engineering

The operator who lived the pain + the architect who owns the technical stack. $13.5M raised between us — both founders have shipped venture-backed products at scale before. Expanding to 6 FTEs (including founders) in 2026.

The Ask

$2.4M

First raise. Founder-funded to ARR.

Plan for invested funds

Commercial

$5.1M 18-month exit ARR
Signed PE partnership

Team

6 immediate FTEs

Product Milestones

Proactive intelligence · Automated agentic deployments

$5.1M exit ARR ≈ 100 accounts at blended $50k ACV by month 18.

A13 - The Ask, Financials & Roadmap

$2.4M raise. $47k to $5.1M ARR in 18 months.

$2.4M to compound traction into a category-defining position in mid-market industrials. Two founders full-time, expanding to 6 FTEs in 2026. Closing Q2/Q3 2026.

Use of funds

Current ownership

David Jayatillake 50% · Theo Leslie 50%.

Financial projections

	CY26	CY27	CY28	CY29
Revenue	$61k	$2.4M	$11.7M	$35.2M
Exit ARR	$231k	$5.1M	$20.2M	$49.5M
GM	76%	61%	63%	64%
Contracts	5	165	542	1,096
Headcount	6	16	27	39
Rev / FTE	$10k	$152k	$431k	$904k

GM compresses to 61% in CY27 as onboarding peaks. Steady-state per-customer GM 75–95%.

Raise timeline

Apr–May · first meetings
May–Jun · lead term sheet
Q2/Q3 2026 · close

Product (18 mo)

Optimise Quarri toolkit for token-efficient LLM use
Deeper analytics & context memory
Bank of reusable workflows
AI-powered implementation agent

Hires

BD1 · Aug 2026 (M5)
Midlevel Dev · Aug 2026 (M5)
SDR1 / GTM · Sep 2026 (M6)
Senior Dev · Nov 2026 (M8)

Roles, not literal heads — some filled through AI agents as the platform matures. Full quarterly forecast, cohort GM build-up, and 3-statement model in the data room.

Financial forecast & ARR milestones in data room

Get in Touch

Theo Leslie

Co-Founder & CEO

theo@quarri.ai · quarri.ai

linkedin.com/in/theo-leslie-quarri