Gideon Ebose has spent much of his career arranging things other people barely notice. A design system that made more than 100 lottery brands behave like relatives. A pension product meant to gather scattered accounts. Workplace software that connected one corporate system to another. Now there is Rulebase, where the objects being arranged are customer calls, compliance rules, dispute queues and the small operational promises that have a habit of going missing between tabs.
It is an unusual route to artificial intelligence, though a perfectly sensible one for a product designer. Ebose did not arrive through model research or a grand theory of machines. He arrived through interfaces. Interfaces teach a merciless lesson: a process can look elegant on a slide and still become a maze when a person has to use it at 4:47 on a Friday afternoon.
Financial services provides mazes in abundance. A customer writes in. A support agent interprets the problem. A policy lives elsewhere. A quality analyst may review the conversation days later, if it lands in the small sample the team has time to inspect. A dispute moves into another system. The customer waits while the institution performs a complicated dance behind a ticket marked “open.” Rulebase is Ebose’s attempt to give that dance a score.
A designer enters the ledger
Ebose studied history and economics at the University of Essex from 2014 to 2017. The combination is almost suspiciously apt for his present work. Economics asks how systems distribute choices and incentives. History keeps a record of what actually happened after the theory met people.
His early résumé moved between finance and design. There was a trade-finance analyst role at Ebury, contract UX work, and a pension-aggregation product at Payseed. At Muncho, he was selected to represent his university in the Santander Entrepreneurial Awards and built the digital material and prototypes for the entry. The jobs supplied a practical education in institutions: money is abstract until somebody needs to find an account, complete a form or explain an exception.
At Gatherwell, a British platform for local lotteries, he became lead product designer. He built a design system spanning more than 100 lottery brands and worked on a new offering that, according to his published résumé, sold 80,000 tickets in four weeks. He also built data pipelines and ran tests to measure conversion. The interesting detail is not the ticket count. It is the blend of typography and telemetry, of making a product coherent and then asking whether the coherence worked.
Gatherwell was acquired by Jumbo Interactive while Ebose was there. He later joined Microsoft, working on product for virtual communication and Microsoft Viva. In a 2023 Microsoft community post, he introduced himself simply as a designer on the Viva team and asked IT administrators to discuss the awkward business of connecting human-resources systems. Even inside a global software company, he was looking at the seams.
The products before the product
Ebose met Chidi Williams in London. Both are Nigerian, and their skills overlapped without matching. Williams had worked on wealth-management systems at Goldman Sachs and built Buzz, an open-source speech-to-text application. Ebose had spent years turning complicated business systems into usable products. Before Rulebase, the pair tried several ideas together, including an AI customer-feedback tool.
Rulebase itself began in 2024 with a different front door. Its early public pitch focused on voice fraud: detecting account takeovers, phishing and AI-generated voice attacks in financial services. The founders spoke with compliance, risk and fraud teams. Their underlying insight widened. The problem was not confined to a suspicious voice. It lived across the operating environment where customer conversations became decisions, obligations and follow-up work.
They shifted toward quality assurance and compliance. Sentinel, an early product, scanned calls, email and support conversations against a company’s own policies and regulatory requirements. Then came Beam, a real-time copilot for support agents. The company’s broader product became an AI “coworker” for the back office, joining review with action.
“Building intelligent compliant machines.”Gideon Ebose’s public shorthand for the mission
The evolution reads less like a sequence of costume changes than a camera pulling back. Voice fraud was one risky conversation. QA was the review of many conversations. The coworker was the machinery required to do something useful with what the reviews revealed. By the time Rulebase joined Y Combinator’s Fall 2024 batch, the founders had found a sturdy premise: customer interactions are the front door to a surprising amount of financial operations work.
What the machine is allowed to do
Automation in financial services comes with a delightful catch: the work most worth automating is often the work an institution cannot afford to automate carelessly. A missed greeting is one thing. A mishandled dispute, an ignored complaint or a regulatory deadline is another. Rulebase’s product therefore sits between appetite and accountability.
Its QA layer reviews customer interactions across calls, email and chat. Specialist agents can then support workflows such as onboarding, remediation, reactivation and retention. The software connects with the tools already holding the work and preserves human approval for consequential actions. The point is not to hand an institution to a chatbot. It is to turn a fragmented chain of evidence and action into a system someone can inspect.
This is where Ebose’s design background becomes more than an agreeable origin story. AI products are frequently described in terms of what the model knows. Operations products succeed or fail on what happens next. Who sees the alert? Which rule produced it? What may the software do without approval? Where is the evidence when an auditor asks? Every answer is a design decision wearing sensible shoes.
Ebose has been direct about the value of domain knowledge. “You need to understand Mastercard’s rules, CFPB timelines,” he has said. The sentence contains the business thesis in miniature. Generic intelligence is cheapening. A reliable map of rules, context and handoffs is harder to assemble, and harder still to keep current.
The charm of the unglamorous
In September 2025, Rulebase announced a $2.1 million pre-seed round led by Bowery Capital, with Y Combinator, Commerce Ventures, Transpose Platform VC and angel investors participating. The company said its software was already deployed with Rho and a Fortune 50 financial institution. Its stated customers later included Qonto and Lesaka. At FinovateSpring in May 2026, Ebose presented Rulebase’s compliance agents to an audience accustomed to hearing financial software promise a tidier tomorrow.
Money and logos are conventional startup milestones. The more revealing measure may be the company’s expanding vocabulary. It now talks about complaints, disputes, audit preparation, onboarding and accounts that stall before a first transaction. These are not separate inventions so much as species of unfinished business. Each begins with a signal, travels through several owners and risks becoming invisible along the way.
Ebose has written about onboarding customers by stating clearly what the product can do with confidence, setting weekly expectations and delivering value quickly. It is modest advice, and therefore useful. In a market thick with synthetic colleagues and heroic verbs, clarity about limits has commercial force. A bank does not need software to sound enchanted by its own possibilities. It needs the software to remember the deadline.
The founder’s recurring material is not artificial intelligence. It is the handoff: the quiet moment when one system, person or promise must meet another.YesPress
There is a humane edge to that focus. Rulebase has described personalized improvement reports that show analysts and service agents what they got right and where they can improve. The reaction, Ebose observed, was not chiefly about speed or accuracy. People valued specific feedback because so much operational labor becomes visible only when something fails. Automation can be used to erase labor. It can also make good labor legible.
That distinction may determine whether the AI coworker becomes a tiresome metaphor or a useful product. Ebose’s version is at its most convincing when it behaves less like a personality and more like excellent institutional memory. It notices every conversation, carries context into the next system, asks for approval when judgment matters and leaves the filing cabinet in better order than it found it.
A larger back office
For now, Rulebase is focused on financial services across the United States, Europe and Africa. The company has pointed toward adjacent regulated fields such as insurance, where the same combination of high volume, procedural detail and human consequence appears. Ebose’s stated ambition is broader than QA: pull as many manual back-office tasks as possible out of fragmented tabs and into coordinated workflows.
The ambition is large, but the path remains pleasingly specific. Start with the conversation. Find the obligation inside it. Attach the relevant rule. Give the work an owner. Keep the evidence. This is not a science-fiction future. It is an operations meeting that ends on time.
Ebose’s career suggests that systems become interesting to him at the moment they threaten to become unmanageable. A hundred brands. Corporate integrations. Millions of customer interactions. He responds by looking for the common structure and then giving it an interface. Rulebase is the largest version of that habit so far.
There will always be more glamorous work than reading the fine print. There may be less consequential work. Somewhere, a complaint is waiting in a queue, a customer has been asked for the same document twice, and an analyst is opening a fourth tab to learn what the first three failed to say. Ebose is building for that exact, ordinary moment. It is where the future most often forgets to look.