A $204B industry
run on a model
with a hard ceiling.

"Vertical SaaS for consulting" gets passed on because the TAM slide shows a software budget: a few billion dollars, fragmented, with low willingness to pay. That's the wrong number. It was always the wrong number for every category that mattered.

Stripe's TAM slide was never "payment software spend." It was payment volume. Cursor's TAM was never "IDE spend." It was software engineer compensation. The category only gets big when the product stops selling a tool and starts selling the labor itself.

Goldman Sachs estimates generative AI adds $7Tto global GDP and exposes 300 million jobs to automation, weighted toward high-education, high-salary work. That is the consulting labor budget, sized by an investment bank with no reason to talk our book.

The honest objection: consulting isn't standardized like payments. Every firm runs a different methodology, every engagement has a different client. That's true, and it's the point. A generic tool can't win this by copying a Stripe API, because there's no single interchange to standardize. The product has to hold each firm's specific graph of method, client and decision. Heterogeneity isn't a risk to the thesis. It's the reason the data moat exists at all.

The Big Four bill roughly $204Ba year on a headcount of 1.5Mpeople. Revenue doesn't track headcount. Revenue is headcount: a pyramid where junior hours get sold at senior margin. It has worked for a century, and it scales exactly one way. Hire more people.

That's the tell every operator-turned-investor should recognize immediately. A business that can only 2x revenue by nearly 2x-ing payroll is not a margin business. It's labor arbitrage wearing a strategy-firm suit, and it cannot respond to a product that removes the labor without cannibalizing the model that pays the partners.

That's not a guess. In January 2026, McKinsey's global managing partner disclosed publicly that 25,000of the firm's 60,000 'employees' are now AI agents. The largest incumbent in the category is already retrofitting AI onto a cost structure built for people, on defense, inside the exact structural constraint that makes retrofits slow.

This company doesn't work on GPT-3. Consulting is disproportionately intelligence-work: diagnosis, proposal, report, meeting notes, follow-up, dressed up as judgment. It follows rules. It was just too illegible for software to encode until reasoning models could hold a real chain of steps and check their own output.

That threshold moved software from assisting the work to delivering it. Once the machine delivers the output instead of a blank editor, the unit of sale flips: from seat to outcome. Every pricing model built around "software spend" stops being the right comparable.

The comps are already trading on that flip. Cursorwent from $1M to $100M ARR in 12 months, the fastest SaaS company in history, and Harveycrossed $195M doing the same thing for law. Neither is a model company. Both are software that does a specific, previously illegible job and charges for the job done.

consultor.app is the first AI-native infrastructure built specifically for expertise sellers. Not a CRM with a chat widget bolted on. Not Notion with an AI tab. A single agent runs the operation end to end, from first CRM touch to final report, over a proprietary graph that indexes every client, meeting, method and deliverable and consults it before acting.

It doesn't advise. It executes. It researches the client before the call, builds the briefing from the graph, transcribes the meeting and critiques how it was run, writes the report cell by cell, runs code to close the ROI model, and orchestrates specialist sub-agents in parallel. Nothing irreversible ships without a human approving the preview. The machine owns intelligence-work, the consultant keeps judgment.

That distinction is the product decision the category needed, and none of the horizontal AI layers make it. Notion AI writes inside a blank doc with no client history behind it. HubSpot AI and Salesforce Einstein sit on top of a CRM built for transactional selling, not judgment work. Microsoft Copilot drafts inside Office with no method, no engagement graph, nothing to reason over except the file open on screen. Bolting a model onto a horizontal tool doesn't fix the missing layer. It just makes the missing layer type faster.

The obvious objection: models get commoditized, context windows get commoditized, OpenAI ships your feature in a keynote. All true, and none of it is where the defensibility lives.

The moat is the graph. Every method, every client, every decision structured into relationships that only accumulate by actually running consulting operations. No model provider has that data. It can't be scraped, and it can't be reconstructed from a general corpus, because it doesn't exist anywhere except inside the accounts running on this system.

Every engagement makes the next engagement's diagnosis faster and the next report better, for that firm specifically and for the base model the whole platform learns from. That's a compounding curve a fast-follower with a better model still has to start from zero to match.

From client onboarding to proposal generation and benchmarking reports, the agent already executes 270+consulting workflows end to end, not chat replies, finished work product, on live client accounts today. Here's what that turns into for the people running it.

Marcelo turned a 3-day diagnostic into 15 minutes. Ticket size went from $1.6K to $3.6K per project, and his work week went from 60 hours to 32. Juliana scaled from 30 mentees she could barely keep up with to 80across six countries. Churn dropped from 12% to zero in four months. Bruno packaged a decade of methodology into a membership area and went from $0 to $40Kin recurring MRR without adding headcount.

Same pattern each time. Revenue per operator goes up while hours worked go down, and the expansion motion is already organic: solo practitioners pulling in international clients before we've built a single international sales motion.

Brazil has a genuine, under-exported tradition of turning consulting into rigorous method: FGV, Falconi, Cláudio Galeazzi's turnaround discipline, Betania Tanure's change-management practice. None of those names ever crossed the Atlantic as a brand. The thinking did, quietly, without the infrastructure to package and sell it globally.

Pair that domain depth with Brazil's engineering cost structure and you get the same arbitrage that built Havaianas. Simple product, Brazilian execution, global distribution. Except this time the product is software, and the export is judgment infrastructure, not rubber.

We started on the compute side of the pyramid, not the junior-analyst side. No legacy headcount to protect, no partner comp structure to defend. That's the whole unlock.