Patrick Taylor’s team went to Las Vegas looking for fraud and came back with mistakes. During an early beta test, the engineers kept presenting findings to the person running a shared-services center. None delivered the grand revelation they had hoped for. After a week of adjusting algorithms, they were ready to leave, disappointed. Their prospective customer stopped them. The mistakes were useful. They showed where people needed training and where the business was working well. The software had done something valuable; its makers had almost missed it.
There is a pleasing indignity in having a customer explain your own invention to you. Taylor kept the lesson. Catching wrongdoing could make a persuasive opening argument. Helping a company understand its everyday operations offered a much longer conversation. The expense report, that minor masterpiece of corporate paperwork, could tell a story larger than the price of lunch.
The gap behind the login screen
Taylor co-founded Oversight Systems in 2003. He had noticed a gap between controlling access to a financial system and understanding the transactions moving through it. A business could define permissions, assign roles and accumulate an enormous amount of information. Turning those records into something a manager could understand and act upon was another undertaking.
He spoke with executives before launching the company. The resulting proposition joined financial data with continuous analysis: look for trouble as the transactions accumulate, then put the findings into plain language. Enterprise software already contained plenty of information. Taylor was interested in the distance between possessing it and knowing what to do next.
His earlier jobs had given him several vantage points on that distance. Oracle, Red Brick Systems and Symantec were part of his software education. Later came senior marketing work at Air2Web and Fast-Talk. Across his career, he worked in sales, product management, business development and other operating functions. He had experience explaining products as well as helping shape them.
At Internet Security Systems, he joined as employee number twelve. During his time there, the company grew to more than 1,400 employees and became publicly traded. He led a business unit and oversaw industry relations marketing and product management. That chapter put him inside a company as it moved from a small organization into a much larger one.
Give the accountant fewer needles
By 2011, Taylor was writing about a familiar corporate puzzle: why did computerized organizations, complete with computerized controls, still produce improper transactions? His explanation was practical. There were too many records, too many disconnected systems and too many opportunities for the assumptions behind a control to break down.
Continuous transaction monitoring was his proposed response. Gather the relevant financial activity, analyze it and use the results to investigate suspect transactions and improve the process behind them. The scope included errors as well as fraud. A repeated payment matters to a business even when everyone involved has perfectly respectable intentions.
That distinction gave the work a broader purpose. Correcting one transaction returns a particular sum or repairs a particular record. Understanding why the same mistake keeps occurring can change the next batch. It also gives a manager a more useful assignment than repeatedly rescuing the accounts from the same administrative puddle.
In a 2017 article, Taylor made the workload argument with a small numerical example. Consider 500 expense reports. A random audit of 20 percent means inspecting 100. He described an AI approach that flagged 7 percent, or 35 reports, for human review. The 7 percent figure reflected Oversight’s experience across 20 million expense reports at the time. It was a historical observation, rather than a promise that every organization would produce the same result.
The arithmetic makes the operating question easier to see. A smaller review queue can give people more time to examine each exception. A flag still requires judgment. The attraction lies in choosing where to spend that judgment, especially when a large pile of perfectly ordinary transactions has been demanding its share of the afternoon.
A detective with a feedback loop
Taylor’s first degree was in mechanical engineering. He graduated from Georgia Tech in 1986 and later earned an MBA at Harvard Business School. An engineering idea runs through his explanation of Oversight: a process needs feedback. Observe its output, assess what happened, then use that information to make a correction.
He described the analytical method as evidential reasoning. A suspicious pattern might require several clues, each carrying a different amount of weight. Some evidence could strengthen a concern; other evidence could weaken it. A single rigid rule would struggle with that sort of ambiguity. Context had work to do.
“Instead you have to think like a detective does.”
Patrick Taylor, on evidential reasoning
His example concerned entertaining a politically exposed person. The amount spent has meaning in relation to what is customary in that setting. A statistical baseline can help establish whether a pattern looks unusual. The point of the comparison is to inform an investigation; the software’s suspicion is one part of a longer process.
The feedback loop completes that picture. Detection produces information. Someone must interpret it and decide what response fits. A process can then be adjusted, with later results showing whether the intervention helped. It is an engineer’s way of approaching a very human workplace: observe, respond and keep checking.
transactions→02Weigh
evidence→03Act on
findings↻04Measure
again
Two trillion dollars, with a date attached
Oversight introduced its SaaS platform in 2013. By July 2016, when Luminate Capital Partners announced its investment, the company said its client base had more than doubled in the preceding year. It reported monitoring more than $2 trillion in expenditures annually for corporations and government agencies.
The figures belong to that moment in the business. They describe the scale Taylor and his colleagues had reached thirteen years after founding the company. Insights On Demand, its flagship offering, covered travel and expenses, purchase cards and accounts payable. Partnerships with Concur, Oracle, SAP and TSYS helped connect the analysis to the systems where customers already worked.
Annual expenditures monitored
Taylor identified international expansion as a priority. Luminate’s involvement brought software investment experience and new board members, including Hollie Haynes and Scott Kingsfield. For a founder, that marked another change in the problem set: alongside making the technology useful came the work of extending its reach and building the organization around it.
He was also taking the argument to audiences outside customer meetings. In 2015, his speaking schedule included a session called Rise of the Robot Auditor at an accounting technology workshop, a fraud-detection presentation at a Big Data conference and an Anti-Fraud Collaboration webinar. The subject had a theatrical title and a decidedly practical destination: what audit teams could do with the tools.
Leaving the corner office, keeping the conversation
In April 2018, Taylor discussed his book Robo-Auditing, which brought together his thinking on applying AI in corporate finance. His advice about getting started was restrained: begin with manageable wins, help the team become comfortable and expand from there. The machinery could be ambitious. The rollout could afford some patience.
“The high level is walk and then run.”
Patrick Taylor, discussing Robo-Auditing
That July, Oversight announced Terrence McCrossan as its new CEO. Taylor had led the business for fifteen years and stayed on the board. The change separated the founding chapter from the next stage of management. His board involvement continued until the company’s sale to TCV in 2020.
At TechOperators, his work shifted toward early-stage B2B software investing and advising companies on growth and market positioning. The operator’s experience now had a different audience. His record included the mechanics of selling software, working across functions and staying with one company through years of development.
Atlanta supplied places to share that experience. Taylor appeared among the volunteer mentors for Start It Up Georgia, whose community leaders guided groups meeting weekly. He also joined the host committee for the city’s 2024 Smart Business Dealmakers conference. These are quieter entries in a professional biography than founding a company, but they show where the conversations continued.
Back to the place that taught him to measure
Taylor now serves as Managing Director of Georgia Tech’s Center for Finance & Technology and Professor of the Practice. The center’s mission connects finance education with practice. His route back to campus brings the experiences of a founder, an investor and an author into that work.
His current biography also lists board service with Integrated Biometrics, SafeGuard Privacy and PowerUP, alongside the Georgia Tech College of Computing Advisory Board. The institutional connection has a personal side: the College of Computing describes him as an avid Yellow Jackets fan. The engineer returned with plenty to discuss between games.
The sequence of jobs leaves a set of practical questions for anyone studying a business. What does the customer consider useful? Can the organization turn a finding into a decision? How will it know whether the decision helped? Taylor has approached those questions from several seats: selling software, managing a company, backing founders and connecting finance education with practice. Each seat puts a different part of the same process within reach.
The Las Vegas episode remains a useful way into his career. His team had arrived with a definition of success. The customer gave them a better understanding of what the findings were worth. Across the jobs and titles that followed, the durable question was what someone could do with the information. Sometimes the answer begins with a perfectly ordinary mistake.