Mentum
Summer 2021AcquiredAI-Powered Assistant for Procurement and Strategic Sourcing
Pressure-test this opportunity
Explore the risks and possibilities with a prompt for ChatGPT, Claude, or your agent.
AI-Powered Assistant for Procurement and Strategic Sourcing
Explore the risks and possibilities with a prompt for ChatGPT, Claude, or your agent.
Mentum was a San Francisco-based startup founded in 2021 by three Colombian-born co-founders — Gustavo Trigos (CEO), Simon Avila (COO), and Daniel Osvath (CTO) — that participated in Y Combinator's Summer 2021 cohort. The company launched as a B2B investment infrastructure play for Latin America, building APIs and embeddable widgets that allowed non-financial apps to offer compliant investment products. After raising a $4.2M seed round led by Google's Gradient Ventures in May 2022, it pivoted entirely to LLM-powered AI agents for U.S. procurement and supply chain teams — a complete geographic, vertical, and technology shift.
Mentum failed as an independent company because its original LatAm fintech product was structurally throttled by country-by-country regulatory friction that made multi-market scale nearly impossible on a $4.2M budget, and its subsequent supply chain AI pivot — while technically credible — never achieved the customer concentration needed to sustain a standalone business.
The outcome was an acqui-hire: Nuvocargo, a cross-border freight and logistics platform, acquired Mentum in October 2025. Only Trigos joined the acquirer as AI Product Engineer. Co-founders Avila and Osvath had departed before the deal closed, and acquisition terms were not disclosed — consistent with a modest technology-and-talent transaction rather than a strategic premium exit.
Image 1 / 1
Mentum was founded in 2021 by three Colombian-born engineers and operators who had each made their way to the United States to study and build careers in technology and finance. [1]
Gustavo Trigos, who became CEO, brought the most distinctive technical pedigree. He earned a Bachelor's in Finance and a Master of Science in Applied Statistics at Hult, and his interest in machine learning began in 2018 with a fake news detection project that was later cited in academic journals. [2] He went on to work as a quantitative analyst at BlackRock's Quantitative Investment Group, applying machine learning and AI to derivatives trading — and was named a 2021 BlackRock Founders Scholar. [3] His time at one of the world's largest asset managers gave him direct exposure to how investment infrastructure worked at institutional scale — and, by contrast, how little of that infrastructure existed for retail and emerging-market investors in Latin America.
Daniel Osvath, who became CTO, came from a distributed systems background, with engineering experience at Confluent and Apple. [4] Simon Avila, the COO, had worked in business operations at Wish, the e-commerce marketplace, giving the team operational and go-to-market experience alongside its technical depth. [5]
The founding insight was rooted in personal experience: Latin America had a large and underserved population of potential retail investors, but the infrastructure to offer compliant digital investment products — mutual funds, ETFs, local stocks — simply did not exist in an embeddable, API-accessible form. Trigos and his co-founders saw an opportunity to build the "Plaid for LatAm investments": a layer of regulated financial infrastructure that any app could plug into, abstracting away the complexity of local broker-dealer relationships and compliance requirements.
Before building the investment API directly, Mentum started with a related but less regulated problem: data orchestration for LatAm wealth managers. The company built SaaS tooling that processed hundreds of thousands of investment data records monthly for some of Latin America's largest wealth management firms. [6] This gave the team early revenue, domain credibility, and a firsthand view of the data plumbing problems that sat beneath any investment product — a logical on-ramp to the more ambitious API infrastructure play.
Y Combinator's S21 cohort provided early validation, capital, and a network that would prove durable: the eventual seed round included co-founders of Plaid and Jeeves as angel investors, suggesting the "Plaid for LatAm" narrative resonated with people who had built exactly that kind of infrastructure in the U.S. market. [7]
The founding team's strengths were real — quantitative finance depth, distributed systems engineering, and operational experience — but the composition also carried a structural gap: no obvious regulatory specialist or LatAm market-entry operator was present at the founding level. That gap would matter.
Mentum's original product was a B2B2C investment infrastructure layer for Latin America. The core idea: any business — a neobank, a payroll app, a consumer fintech — could embed compliant investment products into its own interface without building the underlying regulatory and brokerage infrastructure from scratch.
The product consisted of two components. First, customizable APIs that connected a client's application to local broker-dealers and investment vehicles — mutual funds, ETFs, and local equities — in a given LatAm market. Second, embeddable widgets: pre-built UI components that a non-financial app could drop into its product to give users an investment experience without designing one from scratch. [15]
The regulatory abstraction was the core value proposition. In Latin America, offering investment products requires working through licensed broker-dealers who are registered with each country's financial regulator. Mentum's API sat on top of those broker-dealer relationships, so its clients didn't have to negotiate them independently. The analogy to Plaid — which abstracted U.S. bank account connectivity — was deliberate and resonated with investors who had watched that model succeed.
Before the investment API, Mentum had built data orchestration tooling for LatAm wealth managers, processing investment data records at scale. [6] This earlier product gave the team domain credibility and likely served as a distribution channel into the broker-dealer relationships the API would eventually depend on.
At the time of the seed round in May 2022, the product was still in beta, with plans to launch in four markets. [9] Whether it ever exited beta is not publicly documented — a critical gap in the record.
After the pivot, Mentum built an LLM-powered AI assistant for procurement and supply chain teams at mid-market U.S. manufacturers. The product automated what the company called "source-to-pay" workflows: the full cycle from identifying and evaluating suppliers, through bidding and quoting, to generating purchase orders and reconciling spend. [11]
The design philosophy was deliberately non-disruptive. Rather than asking procurement teams to adopt a new platform, Mentum's agents operated inside existing workflows — primarily email inboxes — and integrated with ERP systems already in use. A procurement manager could receive a supplier quote by email; the agent would parse it, standardize the data, compare it against existing bids, and surface a recommendation, all without the user leaving their inbox. [16]
The system was built with a human-in-the-loop architecture: agents proposed actions, but users verified and approved before anything was executed. [16] This was a deliberate trust-building mechanism in a context — procurement decisions involving real vendor contracts and purchase orders — where errors carry direct financial consequences.
The underlying technology used LLMs to aggregate and standardize unstructured data from emails, spreadsheets, and documents: the messy, heterogeneous inputs that characterize real procurement workflows. The product targeted mid-market manufacturers in renewable energy, automotive parts, and consumer packaged goods — sectors with high procurement volume and limited internal automation resources. [12]
What distinguished the supply chain product from generic AI tools was its domain specificity and workflow integration depth. Rather than a general-purpose chat interface, Mentum built agents trained on procurement-specific tasks, connected to the actual data flows (email, ERP) where procurement work happened.
Phase 1 (LatAm Fintech): Mentum's customers were B2B — specifically, consumer-facing fintech apps, neobanks, and digital platforms in Latin America that wanted to add investment features without building regulated brokerage infrastructure. The end users were retail investors in LatAm markets who lacked access to digital investment products. At the time of the seed round, Mentum had two broker-dealer partnerships and was in integration with two YC-backed Colombian fintechs, with 25+ additional companies expressing interest. [17]
Phase 2 (Supply Chain AI): After the pivot, Mentum targeted procurement and supply chain teams at mid-market U.S. manufacturers — companies large enough to have meaningful procurement volume but without the internal automation resources of enterprise players. Sectors included renewable energy, automotive parts, and CPG. [12] The company engaged 300+ VPs and Directors of Supply Chain and visited dozens of factories during its customer discovery phase. [12]
For the LatAm fintech product, the opportunity was real but diffuse: Latin America had hundreds of millions of people with limited access to formal investment products, and the embedded finance wave was creating demand for exactly the kind of infrastructure Mentum was building. The challenge was that "large addressable market" and "accessible market" are different things in a region where regulatory fragmentation means each country is effectively a separate go-to-market motion.
For the supply chain AI product, Mentum cited a $300B/year market opportunity in procurement and supply chain, driven by shifting regulations and geopolitical trends including nearshoring and supply chain diversification post-COVID. [18] This figure is a broad category estimate rather than a serviceable addressable market for Mentum's specific product, but it reflects the genuine scale of procurement spend flowing through mid-market manufacturers.
Phase 1 — LatAm Fintech Infrastructure:
Mentum's competitive position in LatAm fintech was structurally difficult. The "Plaid for X" model works when the underlying connectivity problem is solved once and then scales horizontally — Plaid's value came from connecting to thousands of U.S. banks through a single integration. Mentum's equivalent required not just technical integration but regulatory licensing and broker-dealer relationships in each country, meaning the marginal cost of adding a new market was high and the network effects were weak.
Competitors included regional players building similar infrastructure (Flink in Mexico, Hapi in the U.S. targeting LatAm investors) and the broker-dealers themselves, who had incentives to build their own digital distribution rather than cede margin to an API intermediary. The structural disadvantage was that Mentum's go-to-market required convincing broker-dealers to partner — the same entities who were potential competitors — while simultaneously convincing fintech apps to integrate before the product had launched in their market.
Phase 2 — Supply Chain AI:
By 2023–2024, the supply chain AI space was crowded with well-funded competitors. Established procurement software vendors — Coupa, Jaggaer, SAP Ariba — were adding AI features to existing platforms with large installed bases. Newer AI-native entrants including Zip, Tonkean, and Pactum were raising significant capital to automate procurement workflows. The competitive axes that mattered most were distribution reach (how many procurement teams could you reach through existing ERP integrations or partner channels) and workflow depth (how much of the source-to-pay cycle could you automate reliably).
Mentum's position was strong on workflow depth — the human-in-the-loop design and email-native approach were genuinely differentiated — but weak on distribution. With two employees and no disclosed enterprise sales infrastructure, reaching mid-market manufacturers at scale required either a channel partner or a viral adoption mechanism. Neither was evident. The incumbents had the distribution; Mentum had the product philosophy. That asymmetry is difficult to overcome without a Series A to fund a sales motion.
Mentum's revenue model differed across its two phases, and the company never disclosed revenue figures at any stage — itself a signal about the scale achieved.
Phase 1 (LatAm Investment API): The model was API-based SaaS, likely structured as a combination of platform fees and transaction-based revenue on investment flows processed through the API. This is the standard infrastructure-layer model: charge per API call, per transaction, or as a percentage of assets under management flowing through the platform. The data orchestration SaaS that preceded the investment API was likely subscription-based. No revenue figures were disclosed at any point.
Phase 2 (Supply Chain AI): The model was almost certainly SaaS subscription, likely priced per seat or per workflow volume, targeting mid-market manufacturers. No pricing or revenue data was ever made public.
Inferred unit economics: Mentum raised a total of $4.2M across two rounds. [19] At 7 employees at the time of the seed round and a San Francisco cost base, annual burn was likely in the range of $1.5M–$2M, implying roughly 2–2.5 years of runway from the seed close in May 2022. The YC job listing at some point claimed "7+ years of runway" — a figure that is difficult to reconcile with a $4.2M raise at SF burn rates unless the team had contracted significantly. [20] By the time of the YC company listing, Mentum had 2 employees, [21] which would dramatically reduce burn and extend runway — but also signals a company that had wound down most of its operations before the acquisition. All burn estimates are inferences from headcount and location data, not disclosed financials.
The absence of any revenue disclosure across four years of operation is consistent with a company that never achieved meaningful commercial scale in either product phase.
Mentum's disclosed traction metrics are sparse and concentrated in the early fintech phase.
Phase 1 (LatAm Fintech, as of May 2022):
The "25+ companies interested" figure is a pipeline metric, not a revenue metric. Two integrations in progress after 12+ months of operation in a regulated market is a slow pace for a company with $4.2M in the bank.
Phase 2 (Supply Chain AI, 2024):
The 300+ discovery conversations demonstrate serious customer development discipline. The absence of any named customers or revenue metrics in the acquisition announcement — despite Nuvocargo having every incentive to highlight commercial validation — suggests the supply chain product had not yet achieved meaningful commercial scale.
The most important failure in Mentum's story happened before the pivot — and it was structural, not executional.
Trigos described the LatAm regulatory process as "burdensome" at the time of the seed round: to enter each country, Mentum had to reach out to the 10 top broker-dealers in that market, understand local regulations, and build relationships before doing any business. [10] That is not a sales problem that better outreach can solve — it is a structural go-to-market tax that applies to every market entry, regardless of how good the product is.
The math is punishing. With plans to launch in four markets and a $4.2M seed round, Mentum was attempting to build regulated financial infrastructure across multiple Latin American countries — each with its own regulator, its own broker-dealer ecosystem, and its own compliance requirements — on a budget that would be modest for a single-market U.S. fintech. The product was still in beta 12+ months after founding and at the time of the seed close. [9] Whether it ever exited beta is not documented in any public source.
The attempted remedy was the seed round itself: $4.2M to fund market entry and broker-dealer relationship-building across four countries. The outcome was that the capital was insufficient for the regulatory surface area the company was trying to cover. The "Plaid for LatAm" analogy that attracted investors was structurally misleading: Plaid's connectivity problem was solved once and scaled horizontally across thousands of U.S. banks. Mentum's equivalent required a separate regulatory and partnership motion in each country, making the marginal cost of expansion high and the path to network effects long.
The founding team's composition compounded this. Trigos brought quantitative finance depth; Osvath brought distributed systems engineering; Avila brought operational experience at a consumer marketplace. None of the three had an obvious background in LatAm financial regulation or broker-dealer relationship management — the exact capabilities the go-to-market required most.
The pivot to supply chain AI was a rational response to a genuine market opportunity. The emergence of capable LLMs in 2022–2023 (GPT-4, Claude) made it technically feasible to build the kind of unstructured-data-parsing, workflow-automation product Mentum built in its second phase — a product that would have been impossible to build reliably at founding. The customer discovery discipline was real: 300+ conversations with supply chain VPs and factory visits represent serious market validation work. [12]
But the pivot also came with a structural problem that the team could not solve with the resources available. By the time Mentum was building supply chain AI, the competitive landscape included well-funded AI-native entrants (Zip raised $43M Series B in 2022; Pactum raised $25M Series A in 2022) and incumbent procurement platforms (Coupa, SAP Ariba) with large installed bases adding AI features. Mentum's differentiation — email-native agents, human-in-the-loop design, workflow depth — was real, but differentiation without distribution is a product, not a business.
The team had contracted to 2 employees by the time of the YC listing. [21] A 2-person team building enterprise supply chain AI and selling to mid-market manufacturers — a segment that requires relationship-based sales, implementation support, and ERP integration work — is structurally undersized for the go-to-market motion the product required. The attempted remedy was deep customer discovery and a focused product build. The outcome was technology that was compelling enough to attract an acquirer but not a customer base large enough to sustain a standalone company.
Mentum's supply chain AI product faced a category-level structural challenge that affected many AI-native startups in 2023–2025: the risk that a well-resourced incumbent would absorb the core capability.
Nuvocargo CEO Deepak Chhugani said: "The Mentum team really understood these agent-based workflows, especially in email and messaging environments, and that experience is going to accelerate our roadmap." [22] That framing — "accelerate our roadmap" — is the language of a feature acquisition, not a product acquisition. Mentum's AI agent capabilities were valuable as a component of a larger logistics platform; they were not, on their own, sufficient to anchor a standalone company with a defensible market position.
This is a structural pattern in enterprise AI: when the core technology is built on top of foundation models that any well-resourced competitor can also access, the defensible moat must come from distribution, data, or workflow integration depth. Mentum had workflow integration depth. It did not have distribution or proprietary data at scale. Nuvocargo had the distribution — enterprise relationships with major shippers and manufacturers — and could absorb Mentum's workflow expertise into a platform that already had the customer relationships to deploy it.
By the time of the acquisition, only Trigos remained. Co-founders Avila and Osvath are absent from all acquisition coverage, and the YC listing showed 2 employees at some point before the deal. [21] [23] The departures of two of three co-founders before an exit is a signal worth noting, though the reasons are not publicly documented. Whether they left because of strategic disagreement over the pivot, personal circumstances, or simply because the company's trajectory made departure rational is unknown. The outcome — a single-founder company at acquisition — is consistent with a company that had wound down most of its operations before the deal closed.
Nuvocargo's acquisition of Mentum closed in approximately two months, from first conversation in August 2025 to close in early October. [13] A two-month deal process with no disclosed competitive bidding is consistent with a transaction where the seller had limited leverage. The deal was Nuvocargo's third acquisition since inception and second in 2025, [24] suggesting it was a programmatic acqui-hire strategy rather than a one-time strategic bet.
Whether investors recovered any capital from the $4.2M raised is not disclosed. Given the acqui-hire framing and the absence of any revenue metrics in the acquisition announcement, a return at or below invested capital is the most plausible inference — though alternative explanations exist and the deal terms remain private.
Regulatory complexity in emerging markets is a go-to-market cost, not a moat. Mentum raised $4.2M to build investment API infrastructure across four LatAm markets, each requiring separate broker-dealer relationships and regulatory compliance. The same regulatory friction that made the market underserved also made it expensive to enter — and $4.2M was insufficient to cover the country-by-country partnership and compliance work required. The "Plaid for LatAm" framing attracted investors who had seen the Plaid model succeed in the U.S., but Plaid's connectivity problem was solved once and scaled horizontally; Mentum's equivalent required a separate regulatory motion per country, making the capital requirements fundamentally different.
LLM capabilities genuinely unlocked Mentum's second product — but timing advantage is not the same as competitive advantage. Mentum's supply chain AI product would have been technically impossible to build reliably in 2021 when the company was founded; GPT-4 and Claude made it feasible by 2023. The pivot was correctly timed to the technology curve. But every well-funded competitor had access to the same foundation models, meaning Mentum's workflow depth and human-in-the-loop design were differentiators that could be replicated by incumbents with larger distribution networks. Being early to a capability is valuable; being the only one with that capability is what creates a defensible business.
Enterprise AI products that live in existing workflows need distribution infrastructure to scale, not just product depth. Mentum's most praised design decision — building agents that operated inside email inboxes rather than requiring platform adoption — was validated by the acquisition. Nuvocargo specifically cited Mentum's expertise in "agent-based workflows, especially in email and messaging environments." [22] But a 2-person team selling to mid-market manufacturers through relationship-based enterprise sales is structurally undersized for that go-to-market motion. The product philosophy was right; the commercial infrastructure to deploy it at scale was absent.
A pivot that changes geography, vertical, and technology simultaneously is effectively a new company — and should be evaluated as one. Mentum's shift from LatAm fintech APIs to U.S. supply chain AI changed every dimension of the business: customer profile, regulatory environment, technology stack, competitive landscape, and sales motion. The $4.2M raised for the fintech product had to fund both the wind-down of that product and the build-out of an entirely new one. No additional capital was raised. The result was a technically credible second product built by a team that had contracted from 7 to 2 employees — a resource constraint that made commercial scale difficult to achieve before the runway ran out.