Operating Partner Model · v2 · 2026

The company that doesn't promise AI.
Delivers it.

An embedded AI operating partnership for Brazilian enterprise — built on the same thesis Anthropic, Blackstone, and Goldman just backed at $10B+ scale. We are building the Brazilian chapter.

The signal — May 4, 2026

The category just got institutionally validated.

Anthropic + Blackstone + Hellman & Friedman + Goldman Sachs formed a new enterprise AI services company — backed by General Atlantic, Apollo, GIC, and Sequoia — to embed Applied AI engineers inside mid-market companies.

"A typical engagement starts with a small team working closely with the customer to understand where Claude can have the biggest impact. Engineers build custom solutions tailored to each organization's operations."

FDE model validated at institutional scale · PE capital backing the category · Brazil: unclaimed. That's PRAGMA.
The problem

Companies have AI. They don't have AI capability.

91%
of companies use AI in at least one capacity
80%
report no measurable productivity impact
$600B
global gap between AI capital deployed and ROI captured
6%
of companies achieving significant AI ROI

What enterprises are missing is not more AI — it's a partner accountable for what happens after the contract is signed.

The manifesto

Four lines. The whole thesis.

01

Change the work

Which tasks exist. Which workflows run. What the company does every day. Before automation, we question whether the work should exist in its current form at all.

02

Change who does the work

Humans → synthetic workers. BPO disruption. The most visible, most measurable line — a client can see this on a headcount report within 90 days.

03

Change how the work gets done

The operating model. The governance. The data architecture. Building the capability infrastructure inside the client so the new workforce operates with new rules, new accountability, new speed.

04

Change the impact of the work

Not efficiency. Not automation. Competitive position. EBIT. Market share. Work that used to cost R$2M now costs R$400K and creates structural advantage — a different category of outcome entirely.

The scope architecture

Five levels. One upsell path the client can see from day one.

PRAGMA operates at five scope levels — from a single task to the entire company. The architecture gives a visible progression the client understands before the first engagement is signed. The higher the level, the less the conversation is about cost — and the more it's about survival.

COMPANY

Enterprise-wide AI capability as infrastructure

Strategic · Existential

Not something you sell — something you earn. PRAGMA functions as the client's internal AI operating function, with cross-client pattern recognition no internal team could have. Reached only after trust is built at every level below.

BU

One business unit fully transformed

Competitive · Market position defended or captured

Individual process wins turn into organizational transformation. Multiple processes, multiple tasks, PEOPLE capability embedded. PRAGMA has enough context to run structured experiments against a production environment it understands deeply.

PEOPLE

Humans operating alongside AI — at every level

Resilience · Cost of dependency avoided

Not sequential — runs in parallel through TASK, PROCESS, BU, and COMPANY. Means the client can sustain what we built — and paradoxically, clients stay longer when they know they could run it themselves.

PROCESS

How tasks connect: workflows, handoffs, decisions

Throughput · Value of full workflow optimized

Once PRAGMA owns a task in production, adjacent tasks become visible — and the argument for expanding into the full process writes itself. The client has already seen the outcome. Now they want it across the workflow.

TASK

A single, bounded, repeatable unit of work

Efficiency · Cost of outsourced task replaced

Lowest friction entry. Clients only have to believe one task should run better. Fastest time to demonstrable outcome. Runs in production from day one — not a POC.

Three commercial models

One capability stack. Three P&Ls.

Every competitor picks one model. PRAGMA runs all three from the same underlying infrastructure — because the synthetic worker that runs inside a WORKFORCE contract is the same agent built into a client's INHOUSE capability.

The BPO killer

PRAGMA WORKFORCE

Replace human BPO workers with synthetic workers. Charge per outcome — not per seat, not per hour. Per invoice processed, per contract reviewed, per ticket resolved. Margin scales with volume, not headcount.

  • Pricing: Per unit outcome
  • Scope: TASK + PROCESS
  • Margin: Highest — scales w/o headcount
  • Analogue: Invisible Technologies ($2B)
The capability builder

PRAGMA INHOUSE

Build the AI capability inside the client. They own it — but we built it, we trained their people, and our OS runs under it. Retainer model. Clients stay because they're invested, not because they're locked in.

  • Pricing: Retainer + success fee
  • Scope: PEOPLE + BU (+ PROCESS)
  • Margin: Strong — IP leverage on delivery
  • Analogue: Distyl AI, Palantir engagements
The strategic operating partner

PRAGMA PARTNER

PRAGMA embedded at company level — functioning as the client's internal AI operating function. Outcome-based fees, equity for transformational engagements, cross-portfolio pattern recognition nobody else has.

  • Pricing: Retainer + equity / carry
  • Scope: COMPANY (all levels)
  • Margin: Highest strategic, equity upside
  • Analogue: Blackstone + Anthropic model

Scope × Model migration matrix

ScopeWORKFORCEINHOUSEPARTNERPricing logic
TASK✓ Per unit✓ RetainerCost of task replaced
PROCESS✓ Per flow✓ RetainerValue of full workflow
PEOPLE✓ RetainerCost of dependency avoided
BU✓ RetainerMarket position value
COMPANY✓ + EquityExistential transformation

The flywheel: WORKFORCE volume trains the agents used in INHOUSE contracts — which generates data making PARTNER experiments more powerful. The more BPO volume we run, the smarter our agents get. The smarter our agents, the stronger our INHOUSE proposition. The stronger that proposition, the more strategic our PARTNER relationships become.

How we work

Three engagement modes — one operating model.

Find where AI matters most

PRAGMA DIAGNOSE

Forward-deployed team embeds for 4–6 weeks. Maps operations, identifies 3–5 highest-leverage opportunities across all five scope levels. Delivers a production roadmap — not a slide deck.

  • R$80–200k · fixed fee
  • Entry point for all clients
Deploy AI into production

PRAGMA BUILD

Embedded engineers design, build, and ship AI systems into actual workflows. WORKFORCE agents, INHOUSE capability, or PARTNER-level transformation — the engagement shape follows the scope level.

  • R$150–400k · /month retainer
  • 3–12 month engagement
Run, monitor, compound

PRAGMA OPERATE

Ongoing governance, agent monitoring, retraining, and scope expansion. Every quarter the OS learns more about your business. AI doesn't degrade — it compounds.

  • R$50–150k · /month + 5–15% success fee
  • % of measured value created
The PRAGMA OS

The proprietary IP that makes the guarantee possible.

The R&D team has one job: compress time-to-capability. Watch everything in open source — LangGraph, CrewAI, AutoGen, LlamaIndex — absorb what works, wrap the best of it into proprietary operating systems deployed for every client. What takes a competitor 6 months to build, PRAGMA deploys in 3 weeks — because the OS already exists, with Brazilian compliance baked in, and connectors to SAP / TOTVS / Oracle already built.

Synthetic worker runtime

TASK OS

Pre-built agent templates for high-frequency enterprise tasks. NF-e, SPED, REINF, TOTVS, SAP, Oracle connectors. LGPD compliance built in. Quality assurance loops — the agent grades its own output.

Workflow orchestration layer

PROCESS OS

Multi-agent orchestration across dependent tasks. Handles handoffs, decision points, exceptions. Pre-built for fiscal reconciliation, procurement, compliance review, document intelligence.

Human-AI collaboration layer

PEOPLE OS

The interface between synthetic workers and humans working alongside them. Escalation logic, judgment-call routing, feedback loops. Ensures the org sustains what PRAGMA built — without permanent dependency.

Business unit transformation layer

BU OS

Cross-process orchestration, shared data architecture, BU-level dashboards and measurement. When multiple PROCESS OS deployments run inside the same BU, BU OS ties them into a coherent operating model.

Enterprise AI infrastructure

COMPANY OS

Company-wide governance, data mesh, agent registry, performance architecture. Makes AI a core competency — not a project. Built on top of all four lower OS layers.

The compounding advantage: The 50th deployment of a fiscal reconciliation agent is materially smarter than the first — and no competitor has run 49 previous ones. Services firms have people. PRAGMA has IP that compounds independent of headcount.

Why Brazil · Why now · Why PRAGMA

A market with the deepest moat and zero coverage.

The structural moat

Brazil's fiscal/tax infrastructure — NF-e, SPED, REINF, eSocial, LGPD — is the most complex in the world. Most outsourced, most intelligence-heavy, most auditable. A US-born AI company cannot replicate this knowledge without years of local investment. That's not a constraint. That's a moat.

The white space

~8,000 Brazilian companies in the R$500M–R$20B range. Zero dedicated FDE firms. Zero Claude Partner Network members covering LatAm mid-market. Zero equivalent of the Anthropic + Blackstone company operating natively in Portuguese with local regulatory depth.

≈8,000
Brazilian mid-market companies in scope
0
Dedicated FDE firms in Brazil today
0
Claude Partner Network members covering LatAm mid-market
18-month sequencing

From wedge to platform leverage.

0–6 MO

Land the wedge

2 anchor clients via fiscal/tax AI. TASK + PROCESS via WORKFORCE. DIAGNOSE → BUILD. Apply to Claude Partner Network. Hire first 2 FDEs.

6–12 MO

Expand scope + partners

First clients move TASK → PROCESS → PEOPLE. First PE fund as portfolio Operating Partner. 2 commercial partner agreements. PRAGMA Insights launches.

12–18 MO

Build the OS

PRAGMA OS formalised. TASK OS + PROCESS OS in production. R&D team wrapping open source. Agribusiness AI discovery. Second PE fund.

18+ MO

Platform leverage

OS compresses delivery cost. PARTNER-level clients. Series A: 'the Distyl AI of LatAm.' BU OS + COMPANY OS deployments. LatAm expansion via partners.

Others theorize AI.
PRAGMA practices it.

If your company is already investing in AI and has nothing to show for it in production — that's the problem PRAGMA exists to solve. Not with a new model. With a team embedded in your operations, accountable to your results, and an OS that gets smarter with every engagement.

Start with PRAGMA Diagnose →