Startup Spotlight: Tugce Bulut builds Eloquent AI to automate bank workflows without APIs

Why it matters

Eloquent AI is betting that financial firms will adopt AI faster when it works inside their existing systems. Bulut's enterprise-building experience is relevant; the company's revenue and automation claims still need clearer operating evidence.

Tugce Bulut (@TugceBulut) has built a second company around a familiar enterprise problem: important work still gets done by people moving between software systems that were never designed to work together. Her San Francisco startup, Eloquent AI, wants AI operators to handle those screen-based tasks for banks and fintechs, including customer onboarding, account issues and compliance workflows.

The pitch grows out of a founder who has already built for large-company customers. Bulut founded Streetbees, a consumer-insights company that used mobile contributions and machine learning to analyze behavior for brands. Her background also includes strategy consulting, training at Cambridge and relevant awards. Y Combinator says she scaled it to 200 employees, raised $80 million in venture capital and secured multi-million-dollar enterprise deals with Fortune 500 clients. Streetbees described using deep neural networks to analyze consumer-submitted stories, photos and video for large consumer-goods companies. That history gives Bulut experience selling AI-enabled software into enterprises; it does not, by itself, show that financial institutions will trust an autonomous system with regulated decisions.

Eloquent AI is listed in Y Combinator's Spring 2025 batch. Its co-founder, Aldo Lipani, is an associate professor of machine learning at University College London. Lipani holds a PhD, and his prior work and research include large language models, conversational systems and evaluation. That academic focus meets Bulut's enterprise-building experience in a product whose hardest test is less fluent conversation than consistent execution across unfamiliar software and tightly controlled processes.

The interface is the integration

Eloquent AI's central product choice is to interact with the software financial firms already use. Its product description says the AI learns from standard operating procedures and observed employee workflows, then controls browser and desktop applications to complete multi-step tasks. Its Y Combinator launch materials describe examples such as account unfreezing, identity checks, repayment adjustments and claim-status updates. The goal is to avoid waiting for custom API integrations or engineering teams to connect every system.

That approach targets a real deployment bottleneck. A bank's operation may cross a support platform, internal dashboard, customer record and compliance checklist, with older tools mixed in. An API-first automation project needs systems access and integration work before an agent can act. A screen-operating agent can potentially start with the same interfaces an employee uses, reducing one kind of implementation work. It also inherits a different problem: software screens change, permissions can be easy to misconfigure, and an incorrect click can have consequences beyond a bad answer.

Eloquent AI says its proprietary multimodal model, Oratio, uses both language and visual information. Its Y Combinator launch materials describe secure, auditable interactions, continuous simulation and rigorous evaluation. Those are important design claims in financial operations, where a system needs to show what it did and why. Public product descriptions do not provide enough detail to assess how Eloquent AI tests exceptions, handles a changed interface or decides when a human must take over.

The product has also broadened from the original financial-services operator pitch. Eloquent AI's website now presents Fixer for customer service, Closer for inbound sales, Navigator for onboarding and support, and custom agents. It lists integrations with familiar business platforms and says customers can use APIs, SDKs and low-code customization. The combination suggests Eloquent AI is selling both an out-of-the-box route for teams that lack engineering capacity and a configurable platform for larger deployments. The no-API argument is the entry point, while the wider product page shows Eloquent AI is not limiting itself to screen-only automation.

Fast traction, with a short measurement window

The headline traction figure comes from Eloquent AI's Y Combinator profile: Eloquent AI says it reached $500,000 in annual recurring revenue in four weeks. The same profile says the operator can automate up to 96% of tasks. Both numbers are Eloquent AI claims. The profile does not define the four-week starting point, say how much of the claimed ARR was live at that point, or explain the tasks and workflows counted in the 96% figure.

An investor post from Duke Capital Partners added an early customer count: 10 paying clients, naming Cleo, OakNorth Bank and Vouch, plus more than 150 companies on a waitlist. That offers a glimpse of initial commercial interest, but the investor did not break down which named customer used which product, how much work was in production or what the waitlist represented. A waitlist is a measure of interest, not deployed revenue. Eloquent AI's website continues to advertise the 96% automation figure, while its careers page claims revenue grew 11 times in a year and says it onboarded major banks and fintechs. The page does not give a baseline or period for the revenue multiple.

For a repeat founder, these are useful early indicators of whether the customer problem is urgent enough to command budget. They are not yet a substitute for results that show reliability across real cases. The crucial missing operating detail is the denominator: what counts as a task, what conditions trigger escalation, and how much human review remains around an automated resolution. An agent resolving 96% of a narrow, well-defined set of routine cases would mean something different from one completing 96% of every incoming regulated workflow.

Eloquent AI is also selling a deployment claim, not just an AI model. If observing an employee's work can produce a safe, reliable operator quickly, Eloquent AI can reduce the time and internal effort a financial-services buyer must spend before seeing value. The four-week ARR claim is consistent with Eloquent AI's argument that customers can deploy quickly, though the profile does not establish that rapid setup caused the reported revenue or explain how many customers contributed to it.

Investors are backing execution in a crowded category

In September 2025, Eloquent AI announced a $7.4 million seed round led by Foundation Capital, with EJF Ventures, Duke Capital Partners, Massive Tech Ventures, Logo Ventures and Y Combinator among the named participants. Bulut said in a LinkedIn post that the round closed in three days and was 12 times oversubscribed. That oversubscription figure is her account of the process, rather than a public filing.

The later funding record is less tidy. In April 2026, investor Revo Capital said Eloquent AI had closed an $8.4 million round led by Foundation Capital with Revo participating. Eloquent AI's careers page now describes a $10 million seed round. The public figures do not reconcile into one confirmed total, so the most precise account is that the September 2025 announcement put the round at $7.4 million, with later investor and company materials citing higher amounts. No valuation is established by those announcements.

Eloquent AI is entering a category with other teams selling specialized AI for financial operations. Gradient Labs, for example, announced in June 2026 that it had increased its Series A to $26 million to build specialist agents for finance, including lending, disputes and identity checks. The competitive field also includes Primitive, Unique and Heron, as well as horizontal AI and automation platforms. Eloquent AI's stated distinction is its emphasis on learning workflows from observed actions and working through existing browser and desktop interfaces. That is a product design choice, not a proven moat. The category's commercial test will be whether financial institutions can deploy such systems with enough control to make automation worthwhile without creating a new layer of operational risk.

Bulut's second enterprise bet

Bulut's Streetbees record gives Eloquent AI a founder with experience building an AI-enabled product for demanding enterprise customers and scaling a company around it. Her earlier business gathered consumer behavior from people in their daily lives; Eloquent AI aims to take action inside the systems financial employees use at work. The products differ, but both depend on translating messy human processes into software that can operate at scale.

Lipani's research background complements that bet. Models that navigate interfaces must interpret visual context and language while being evaluated on outcomes, not just on whether an answer sounds convincing. Eloquent AI describes Oratio and its evaluation tools as built for that environment. The publicly described product architecture remains Eloquent AI's account; its effectiveness will depend on performance across actual customer workflows, especially when a screen changes or an unusual case falls outside an employee's standard procedure.

Eloquent AI's most promising thesis is practical: financial firms already have systems and procedures, and replacing them is slower than teaching software to work through them. Bulut has previously persuaded large customers to buy an AI-based product. The next proof is whether Eloquent AI can turn its early revenue claims into repeatable deployments where customers can measure completed work, exception rates and the human oversight still required. Until those details are visible, $500,000 in claimed ARR and a 96% automation ceiling describe a strong sales pitch, not yet a public operating record.