The era in which services become software.

The first era of enterprise software, from the 2000s onwards, was about tools. SaaS sold control: a CRM to manage leads, an ERP to manage processes, a PMS to manage projects. Mental work was still performed by humans — software merely provided structure.

The next era is different. AI agents don't replace the tool — they replace the person who operated the tool. The CRM of the future is not a place where a salesperson logs activity — it is an agent running the prospecting. The ERP of the future is not a place where accountants classify transactions — it is an agent that classifies and reconciles.

Foundation Capital calls this Services-as-Software. Sequoia calls it the largest opportunity of the next decade. Satya Nadella, in his 2024 conversation with Dylan Patel, was blunter still: "The future of SaaS is to be replaced by agents."

MVA operates at that frontier.

The global movement is real. Brazilian specificity is what makes the opportunity singular.

Over the past five years, Brazil built public digital infrastructure few countries have. Pix delivered real-time payment rails. eSocial and Siape opened payroll-linked data in real time. Open Finance exposed clients' financial positions to authorized third parties. Electronic Judicial Process turned opaque courts into structured data sources.

Each of those achievements is, in practice, a national API. Building vertical agents on top of them unlocks operational context that would take years to emulate elsewhere.

At the same time, Brazilian traditional markets still suffer massive labor inefficiency in repetitive tasks. Law firms manage 50 thousand cases with manual triage. Mid-size banks originate payroll loans via correspondents and paper. Investors miss judicial auctions because nobody can read the official gazettes in time.

There is, therefore, a productive disparity: cutting-edge digital infrastructure, archaic market operations. This is exactly the terrain where vertical AI agents deliver 3x to 10x productivity multipliers. This is exactly the terrain MVA chose to operate in.

Not every vertical deserves a product. Not every pain point deserves a five-year investment. We apply three filters, in this order:

1. High volume of repetitive intellectual work. Without volume, the ROI of substitution doesn't pay off. A process that happens a thousand times a month in a firm is already a candidate. A thousand times a day, certain business.
2. Licensable or public data base. An AI agent without context is a guess. Electronic judicial process, eSocial, Bacen and courts are our data floor. Without that structured base, the vertical becomes LLM brute-forcing.
3. Measurable pain with a buyer willing to pay. Substituting 40% of a lawyer's time has clear monetary value. Cutting 10 days from the payroll-lending cycle, likewise. If the pain isn't convertible into savings or incremental revenue, the product has no traction.

When all three align, MVA builds. Leiloa.ai and LawPilot came out of that filter.

Pure product holdings lose touch with the market. Pure consultancies lose operational credibility. MVA does both deliberately — and that creates value compounding in three directions.

First: advisory funds the products' patience. A vertical product in a regulated market doesn't generate cash in month six. Advisory does. That gives the holding an investment horizon pure-product competitors rarely achieve.

Second: products validate advisory. Technical advice carries more weight coming from someone who operates a real product in production with real numbers. When MVA recommends architecture or stack in a Tech DD, it's because we use it daily — not because we read it in a paper.

Third: advisory is market radar. Sitting inside PE/VC M&A deals gives unique visibility into what the market is building, which verticals are maturing, where there are vulnerable entrants and incumbents. That radar feeds the product pipeline itself.