AI: The Revenue Lifecycle Accelerator
AI: The Revenue Lifecycle Accelerator
AI: The Revenue Lifecycle Accelerator
How artificial intelligence transforms every stage of the revenue journey into the digital economy's biggest engine of value creation
How artificial intelligence transforms every stage of the revenue journey into the digital economy's biggest engine of value creation
Reinaldo Coelho and Haim Mesel
Reinaldo Coelho and Haim Mesel
Triaxis Capital Partners
Triaxis Capital Partners
May 15, 2026
May 15, 2026
Leitura
Leitura
·
8
8
MIN
MIN
01
REVENUE LIFECYCLE
REVENUE LIFECYCLE
Since we published the manifesto on the Revenue Lifecycle, the question we hear most isn't about fintech, nor about the financial infrastructure that the Central Bank built. It is about artificial intelligence, and it always comes, deep down, with the same substance: does the thesis survive the AI era, or is it run over by it?
For the reader who did not follow the previous essay, a short sentence is worth stating: the Revenue Lifecycle is the interpretation, which we defend, that value in the digital economy is distributed along a six-stage journey: acquiring customers, converting intent into revenue, collecting efficiently, financing the operation, protecting against fraud and risk, and moving capital. This journey is also supported by a data infrastructure layer, which is transversal to its stages. It is this journey, and not the isolated product, that organizes where the venture opportunity lies over the next seven years.
The answer to the question about AI is direct. AI does not compete with the Revenue Lifecycle thesis; it accelerates each of its stages and the data infrastructure layer. If the revenue journey describes where value accumulates, artificial intelligence is changing the speed, the power, and, above all, who gets ahead. The path remains the same. What changes is who captures more value along it.
Three overlapping waves
The first mistake in almost every conversation about AI is treating it as a single thing. Predictive AI, in particular, was not born seven years ago: machine learning applied to credit, fraud, recommendations, and demand forecasting was already being used at scale well before the recent explosion of generative AI. What has changed recently is the overlap between this predictive layer, already mature in several sectors, and two new layers: generative AI, which has coexisted with predictive in recent years, and agentic AI, which is only now beginning to emerge from the experimental phase.
Each of the three moves the revenue cycle in a different way. Predictive estimates and recommends: it ranks, classifies, prices, and detects patterns. Generative produces and synthesizes: it writes, codes, reasons over documents, generates images, and interacts with tools. Agentic AI, the wave we are entering now, begins to plan, trigger systems, and execute complete workflows with human supervision where judgment matters.
Predictive estimates and recommends. Generative produces and synthesizes. Agentic plans, triggers, and executes.
The distinction matters because each wave accelerates a different stretch of the journey. A sincere confession: some of our best investments were, deep down, predictive AI companies: anti-fraud, credit origination, operation intelligence; built before the term made headlines. We were already investing in AI applied to the Revenue Lifecycle without calling it that. What changed was not the existence of the opportunity. It was the depth with which each of the three layers penetrates the cycle today.
Since we published the manifesto on the Revenue Lifecycle, the question we hear most isn't about fintech, nor about the financial infrastructure that the Central Bank built. It is about artificial intelligence, and it always comes, deep down, with the same substance: does the thesis survive the AI era, or is it run over by it?
For the reader who did not follow the previous essay, a short sentence is worth stating: the Revenue Lifecycle is the interpretation, which we defend, that value in the digital economy is distributed along a six-stage journey: acquiring customers, converting intent into revenue, collecting efficiently, financing the operation, protecting against fraud and risk, and moving capital. This journey is also supported by a data infrastructure layer, which is transversal to its stages. It is this journey, and not the isolated product, that organizes where the venture opportunity lies over the next seven years.
The answer to the question about AI is direct. AI does not compete with the Revenue Lifecycle thesis; it accelerates each of its stages and the data infrastructure layer. If the revenue journey describes where value accumulates, artificial intelligence is changing the speed, the power, and, above all, who gets ahead. The path remains the same. What changes is who captures more value along it.
Three overlapping waves
The first mistake in almost every conversation about AI is treating it as a single thing. Predictive AI, in particular, was not born seven years ago: machine learning applied to credit, fraud, recommendations, and demand forecasting was already being used at scale well before the recent explosion of generative AI. What has changed recently is the overlap between this predictive layer, already mature in several sectors, and two new layers: generative AI, which has coexisted with predictive in recent years, and agentic AI, which is only now beginning to emerge from the experimental phase.
Each of the three moves the revenue cycle in a different way. Predictive estimates and recommends: it ranks, classifies, prices, and detects patterns. Generative produces and synthesizes: it writes, codes, reasons over documents, generates images, and interacts with tools. Agentic AI, the wave we are entering now, begins to plan, trigger systems, and execute complete workflows with human supervision where judgment matters.
Predictive estimates and recommends. Generative produces and synthesizes. Agentic plans, triggers, and executes.
The distinction matters because each wave accelerates a different stretch of the journey. A sincere confession: some of our best investments were, deep down, predictive AI companies: anti-fraud, credit origination, operation intelligence; built before the term made headlines. We were already investing in AI applied to the Revenue Lifecycle without calling it that. What changed was not the existence of the opportunity. It was the depth with which each of the three layers penetrates the cycle today.
02
Where each wave accelerates
Where each wave accelerates
In the acquisition and conversion stages, the basic ability to use generative AI quickly became a commodity. Generating content, drafting outbound communications, qualifying leads with a generic prompt—many companies can already do this with similar tools, and differentiation will hardly come from that alone. The basic layer is commoditized. But the deeper layer of the stage, conversely, is where one of the most defensible territories of the new commercial economy is forming: revenue intelligence built on proprietary interaction data.
Platforms that capture, transcribe, and analyze at scale the actual conversations between salesperson and client—in calls, meetings, and messages—accumulate a corpus that the generic model lacks and that a new competitor cannot replicate simply with more GPUs. It is the same logic as transactional anti-fraud, transposed to the commercial funnel: the model learns from every observed interaction, and every new client makes the system better for all others. Whoever combines this accumulation with deep CRM integration, vertical context of the served sector, and established distribution in the channel is building, in Acquire and Convert, a level of defensibility that the model alone never delivers.
In the collection and financing stages, agentic AI tends to create more durable value, especially when combined with four mutually reinforcing elements: proprietary data accumulated at scale, distribution already established in the sector, deep operational integration into the client’s flow, and regulatory capability executed with competence. Agents that conduct conversational collection, originate credit in real-time, and manage receivables find the Brazilian infrastructure in a rare moment of maturity: Pix Automático for recurring payments, Open Finance for credit decisions informed by real data, and the regulatory evolution of FIDCs creating new possibilities for structured credit. It is no coincidence that the heart of the thesis and the heart of the AI opportunity are in the same place. The link is structural.
The more AI shifts from recommendation to execution, the more relevant governance, audit trails, limits on autonomy, and human supervision become. In credit, collection, fraud, and compliance, the agent cannot just be efficient. It needs to be controllable, auditable, and compliant with regulation.
In protection—that is, fraud, risk, identity, and compliance—agentic AI raises the bar. It not only detects the anomaly but investigates, decides, and acts. Here lies one of the most solid forms of defensibility that exist: proprietary transactional data. The model improves with every transaction, and a new competitor cannot replicate that history. It is the rare case where the advantage deepens on its own, over time. In moving capital, we are at the frontier, where the next layer of infrastructure is being built and where the rules are still being drawn.
In the acquisition and conversion stages, the basic ability to use generative AI quickly became a commodity. Generating content, drafting outbound communications, qualifying leads with a generic prompt—many companies can already do this with similar tools, and differentiation will hardly come from that alone. The basic layer is commoditized. But the deeper layer of the stage, conversely, is where one of the most defensible territories of the new commercial economy is forming: revenue intelligence built on proprietary interaction data.
Platforms that capture, transcribe, and analyze at scale the actual conversations between salesperson and client—in calls, meetings, and messages—accumulate a corpus that the generic model lacks and that a new competitor cannot replicate simply with more GPUs. It is the same logic as transactional anti-fraud, transposed to the commercial funnel: the model learns from every observed interaction, and every new client makes the system better for all others. Whoever combines this accumulation with deep CRM integration, vertical context of the served sector, and established distribution in the channel is building, in Acquire and Convert, a level of defensibility that the model alone never delivers.
In the collection and financing stages, agentic AI tends to create more durable value, especially when combined with four mutually reinforcing elements: proprietary data accumulated at scale, distribution already established in the sector, deep operational integration into the client’s flow, and regulatory capability executed with competence. Agents that conduct conversational collection, originate credit in real-time, and manage receivables find the Brazilian infrastructure in a rare moment of maturity: Pix Automático for recurring payments, Open Finance for credit decisions informed by real data, and the regulatory evolution of FIDCs creating new possibilities for structured credit. It is no coincidence that the heart of the thesis and the heart of the AI opportunity are in the same place. The link is structural.
The more AI shifts from recommendation to execution, the more relevant governance, audit trails, limits on autonomy, and human supervision become. In credit, collection, fraud, and compliance, the agent cannot just be efficient. It needs to be controllable, auditable, and compliant with regulation.
In protection—that is, fraud, risk, identity, and compliance—agentic AI raises the bar. It not only detects the anomaly but investigates, decides, and acts. Here lies one of the most solid forms of defensibility that exist: proprietary transactional data. The model improves with every transaction, and a new competitor cannot replicate that history. It is the rare case where the advantage deepens on its own, over time. In moving capital, we are at the frontier, where the next layer of infrastructure is being built and where the rules are still being drawn.
"AI does not replace the thesis. It accelerates everything it has always meant to say."
"AI does not replace the thesis. It accelerates everything it has always meant to say."
03
Context, and why vertical wins
Context, and why vertical wins
There is a layer that is not a stage, but crosses all of them: data intelligence. An AI agent is only as good as the context it receives. A generic model, however advanced, does not know the specific customer, the transactional history, or the real-world operation. Whoever unifies a company's dispersed data and makes it actionable by an agent controls the layer that all other stages consume.
Data has ceased to be fuel for reports and has become fuel for decisions.
This is why vertical software tends to capture more value than horizontal in this era. Where context, sector-specific data, local regulation, and operational workflow matter more than the model's generic capability, the vertical company operates with an advantage. An agent that understands healthcare regulation, clinical language, and hospital workflows captures value that a generic agent cannot, and builds a barrier that the model alone cannot tear down.
When you cross the vertical axis (the sector) with the Revenue Lifecycle axis (the stage), a two-dimensional map emerges. The most defensible companies of the next decade will occupy a precise cell on this map: a stage of the cycle, within a vertical, driven by AI with proprietary data. "Intelligent billing for healthcare." "Credit origination for power generation." "Vertical anti-fraud for financial services." It is at this intersection that value concentrates.
The model, in the application layer, is usually not the advantage
If a single idea must survive this essay, it is this one, with an important qualification: in the application layer, the model is rarely the sustainable competitive advantage. For foundation model companies, AI infrastructure, highly specialized models, or models trained with exclusive data, the model can indeed be a central part of the advantage. But for the vast majority of application startups—and this is the category Triaxis operates in—the model alone does not defend a position.
For a large portion of corporate use cases, today's leading AI providers offer capabilities that are sufficiently similar that mere access to the model is not a competitive advantage. There are, indeed, relevant differences between them in terms of cost, latency, context window, multimodality, adherence to tools, privacy, reasoning quality, regional availability, and integration ecosystems. These differences matter for architectural decisions, but rarely determine who wins a category.
"Using AI," therefore, is not a competitive advantage: it is a condition of entry. The question that separates a thesis from an illusion is not "does this company use good AI?", but "does this company have something that AI, alone, does not provide for free?". Proprietary data, distribution, vertical context, integrations built over years, and executed regulatory compliance. These are the walls that protect competitiveness.
AI is the how. The control point in the revenue cycle is the what. We bet on the cycle, with AI as an execution tool, not on AI hoping that it happens to land somewhere in the cycle.
The economics of inference reinforces discipline. The unit cost of using AI tends to fall in many cases, but consumption can grow faster than efficiency gains, especially in agentic architectures that execute multiple calls, verifications, and interactions with external systems to complete a single task. Even if current prices reflect intense competition among major providers, it is not prudent to assume that this economic structure will remain unchanged. A company whose margins only close at today's prices carries a risk that the enthusiasm of the moment usually hides. The question to ask is simple: does this business model survive an inference price two or three times higher? The answer separates the business from the spectacle.
There is a layer that is not a stage, but crosses all of them: data intelligence. An AI agent is only as good as the context it receives. A generic model, however advanced, does not know the specific customer, the transactional history, or the real-world operation. Whoever unifies a company's dispersed data and makes it actionable by an agent controls the layer that all other stages consume.
Data has ceased to be fuel for reports and has become fuel for decisions.
This is why vertical software tends to capture more value than horizontal in this era. Where context, sector-specific data, local regulation, and operational workflow matter more than the model's generic capability, the vertical company operates with an advantage. An agent that understands healthcare regulation, clinical language, and hospital workflows captures value that a generic agent cannot, and builds a barrier that the model alone cannot tear down.
When you cross the vertical axis (the sector) with the Revenue Lifecycle axis (the stage), a two-dimensional map emerges. The most defensible companies of the next decade will occupy a precise cell on this map: a stage of the cycle, within a vertical, driven by AI with proprietary data. "Intelligent billing for healthcare." "Credit origination for power generation." "Vertical anti-fraud for financial services." It is at this intersection that value concentrates.
The model, in the application layer, is usually not the advantage
If a single idea must survive this essay, it is this one, with an important qualification: in the application layer, the model is rarely the sustainable competitive advantage. For foundation model companies, AI infrastructure, highly specialized models, or models trained with exclusive data, the model can indeed be a central part of the advantage. But for the vast majority of application startups—and this is the category Triaxis operates in—the model alone does not defend a position.
For a large portion of corporate use cases, today's leading AI providers offer capabilities that are sufficiently similar that mere access to the model is not a competitive advantage. There are, indeed, relevant differences between them in terms of cost, latency, context window, multimodality, adherence to tools, privacy, reasoning quality, regional availability, and integration ecosystems. These differences matter for architectural decisions, but rarely determine who wins a category.
"Using AI," therefore, is not a competitive advantage: it is a condition of entry. The question that separates a thesis from an illusion is not "does this company use good AI?", but "does this company have something that AI, alone, does not provide for free?". Proprietary data, distribution, vertical context, integrations built over years, and executed regulatory compliance. These are the walls that protect competitiveness.
AI is the how. The control point in the revenue cycle is the what. We bet on the cycle, with AI as an execution tool, not on AI hoping that it happens to land somewhere in the cycle.
The economics of inference reinforces discipline. The unit cost of using AI tends to fall in many cases, but consumption can grow faster than efficiency gains, especially in agentic architectures that execute multiple calls, verifications, and interactions with external systems to complete a single task. Even if current prices reflect intense competition among major providers, it is not prudent to assume that this economic structure will remain unchanged. A company whose margins only close at today's prices carries a risk that the enthusiasm of the moment usually hides. The question to ask is simple: does this business model survive an inference price two or three times higher? The answer separates the business from the spectacle.
04
Where this is heading
Where this is heading
Looking at the next three to seven years, three movements seem clear. First, agentic AI will migrate from pilot to operation. What is impressive in a demo today will become the standard way of executing collection, credit, fraud prevention, and financial operations, and the advantage will remain with those who have the data, context, and operational maturity to trust a decision to an agent. Second, the competitive frontier will move down from horizontal to vertical. Generic AI capabilities become built-in features of large platforms, and value migrates to specialists with deep proprietary context in each sector. Third, the Brazilian financial infrastructure stops being a backdrop and becomes an active platform. As Open Finance, Pix Automático, and the new generation of FIDCs mature, space opens up for an entire layer of companies that originate credit, manage risk, and move capital with AI at the center of the decision. Brazil has, at this intersection, an advantage that few emerging markets possess.
The Revenue Lifecycle will not change. What will change is the depth to which intelligence embeds itself in each stage of it.
The journey remains
Companies come and go. Technology waves follow one another ever faster. Predictive gave way to generative in the public spotlight, which gave way to agentic in institutional pilots, and there will be a fourth layer we do not yet know how to name. But the revenue journey is structural. It is not a technological fad; it is the anatomy of how value is created and captured in a digital economy, and it stands firm while the tools transform.
That is why we invest in the cycle, and not in the technology of the moment. The revenue journey is the stable path; artificial intelligence is the most powerful acceleration we have ever seen coupled with it.
AI does not replace the thesis. It accelerates everything it has always meant to say.
Looking at the next three to seven years, three movements seem clear. First, agentic AI will migrate from pilot to operation. What is impressive in a demo today will become the standard way of executing collection, credit, fraud prevention, and financial operations, and the advantage will remain with those who have the data, context, and operational maturity to trust a decision to an agent. Second, the competitive frontier will move down from horizontal to vertical. Generic AI capabilities become built-in features of large platforms, and value migrates to specialists with deep proprietary context in each sector. Third, the Brazilian financial infrastructure stops being a backdrop and becomes an active platform. As Open Finance, Pix Automático, and the new generation of FIDCs mature, space opens up for an entire layer of companies that originate credit, manage risk, and move capital with AI at the center of the decision. Brazil has, at this intersection, an advantage that few emerging markets possess.
The Revenue Lifecycle will not change. What will change is the depth to which intelligence embeds itself in each stage of it.
The journey remains
Companies come and go. Technology waves follow one another ever faster. Predictive gave way to generative in the public spotlight, which gave way to agentic in institutional pilots, and there will be a fourth layer we do not yet know how to name. But the revenue journey is structural. It is not a technological fad; it is the anatomy of how value is created and captured in a digital economy, and it stands firm while the tools transform.
That is why we invest in the cycle, and not in the technology of the moment. The revenue journey is the stable path; artificial intelligence is the most powerful acceleration we have ever seen coupled with it.
AI does not replace the thesis. It accelerates everything it has always meant to say.
Author
Reinaldo Coelho and Haim Mesel
Reinaldo Coelho and Haim Mesel
Triaxis Capital Partners
Triaxis Capital Partners
·
Triaxis Capital
Triaxis Capital
Get in touch
São Paulo - SP
Rua Funchal, 411
Cj 64 - Vila Olímpia - CEP 04551-060
Telefone: +55 (11) 3044-4060
Recife - PE
R Sant'Anna, 267
Salas 201 e 202 - Santana - CEP 52060-460
Telefone: +55 (81) 3221-6501
Triaxis Capital LTDA – CNPJ 15.333.310/0001-03 This website is for informational purposes only and does not constitute a public offering of securities. Investment funds are not guaranteed by the administrator, the manager, or any insurance mechanism. Past performance is not indicative of future results. Before investing, read the fund’s prospectus and consult a qualified professional.
Compliance & Regulatory
Get in touch
São Paulo - SP
Rua Funchal, 411
Cj 64 - Vila Olímpia - CEP 04551-060
Telefone: +55 (11) 3044-4060
Recife - PE
R Sant'Anna, 267
Salas 201 e 202 - Santana - CEP 52060-460
Telefone: +55 (81) 3221-6501
Triaxis Capital LTDA – CNPJ 15.333.310/0001-03 This website is for informational purposes only and does not constitute a public offering of securities. Investment funds are not guaranteed by the administrator, the manager, or any insurance mechanism. Past performance is not indicative of future results. Before investing, read the fund’s prospectus and consult a qualified professional.
Compliance & Regulatory
Get in touch
São Paulo - SP
Rua Funchal, 411
Cj 64 - Vila Olímpia - CEP 04551-060
Telefone: +55 (11) 3044-4060
Recife - PE
R Sant'Anna, 267
Salas 201 e 202 - Santana - CEP 52060-460
Telefone: +55 (81) 3221-6501
Triaxis Capital LTDA – CNPJ 15.333.310/0001-03 This website is for informational purposes only and does not constitute a public offering of securities. Investment funds are not guaranteed by the administrator, the manager, or any insurance mechanism. Past performance is not indicative of future results. Before investing, read the fund’s prospectus and consult a qualified professional.
Compliance & Regulatory