AI Agents in Finance: Why Artian Exists
At Artian, we believe a seismic shift is underway. Enterprises everywhere are starting to rethink how they manage their most critical workflows — and AI is at the heart of this transformation.
The early signs are unmistakable; businesses are moving beyond rigid RPA and basic chatbots. They're beginning to embrace intelligent agents that can reason, adapt, and collaborate — just like we saw during a previous revolution in another high-stakes arena: Wall Street.
Learning from the Reinvention of Trading
To further appreciate the parallels, let’s step back twenty years. The adoption of decimal pricing and Reg NMS in the early 2000s was enabling high-speed electronic trading in the equity markets. The resulting disruption created giants like Citadel and Two Sigma on the buy-side, and created sell-side opportunities to capture significant market share for early movers like Credit Suisse, Goldman Sachs and Morgan Stanley with their AES, Sigma and BXS platforms.
The Story of Clack
Around that time, the FX trading landscape offered no equivalent electronic access. The interbank FX spot market operated either through voice brokers or two electronic platforms: Reuters Matching and EBS. However, both Reuters and EBS were only accessible via screen and keyboard — there were no APIs available for order entry. This limitation posed a significant challenge for firms looking for algorithmic trade execution. In response, Lehman Brothers devised a remarkable and little-known workaround. Their engineers built Clack, a physical device that programmatically pressed keys on the EBS terminal, effectively simulating human interaction.
Clack, 2003-2004, earned its name because it was so loud that it needed to be operated with egg crate foam on the walls.
It was adapted, as in hacked together, starting with a machine that was used to test keyboards.
What does this story have to do with Artian? Well, there’s no mysterious link between Clack and Lehman’s bankruptcy, or anything as dramatic. The more insightful aspect of this episode is actually the relative insignificance of Clack in the end. There is no doubt that this was a neat piece of technology for sure — we love it — but it was a hack nonetheless.
In many ways, Clack was analogous to screen-scraping and button-clicking RPA tools, adapted from technologies designed for testing, not building scalable systems. Within a couple of years, Clack was shelved as the FX trading landscape went through a radical shift with the adoption of FIX APIs, and algorithmic market-making systems became the norm. These autonomous systems were composed of rules-driven C++ engines, with non-AI models, but they were legitimate “agents” with significant real world impact.
The Story of Excel Macros
While Clack operated in the realm of exchange connectivity, another hurdle in the automation of trading was in pricing. For any non-trivial securities asset classes, computational estimation of risk is a key factor in pricing and hedging strategies. Early on, traders did this using Excel models. Then they started writing code using VB macros that plugged into their spreadsheets — this was a huge unlock because traders could now develop sophisticated financial engineering models well beyond the arithmetic functions built into Excel. While this was a significant step up from using pen-and-paper or calculators, it was riddled with problems.
The models and macros lived locally on a trader’s desktop PC. If a trader was on vacation, that pricing model was inaccessible and the risk wasn’t properly managed. So, traders started sharing the models by emailing the spreadsheets, with embedded macros, to each other. But then the versions of the models diverged in irreconcilable ways. Then some enterprising firms started running clusters of headless PCs with desktop versions of Excel being invoked by DOS scripts to run the models.
You may now be chuckling, but it wasn’t obvious then that they were headed down the wrong path. Meanwhile, Goldman Sachs built SecDB to revolutionize institutionalized asset pricing through a complete rethinking of how derivatives trading risk was managed. Goldman’s approach is now broadly adopted all across Wall Street with incremental differences.
Again, what does this second story have to do with Artian? There is a huge difference between increasing the productivity of an individual employee in a firm versus transforming how the firm operates. We see LLM-enabled chatbots as the Excel VB macros of today. The chatbot paradigm inherently limits how effectively the superpowers of generative AI can be leveraged.
Caretakers, Innovators and Changemakers
Around that time, financial firms faced a crossroads. Early "automation" efforts simply recreated manual trading processes with spreadsheets and pursued other clever but eventually misguided attempts. They were a step forward, but they did not fundamentally change the game.
The true transformation came when firms abandoned old workflows and built autonomous, model-driven agents that could make decisions, adapt in real-time, and outmaneuver human competitors. Quantitative trading reshaped the financial landscape, leaving behind those who clung to legacy thinking.
The organizational dynamics of how this played out are perhaps even more important than the technological change at play. The caretakers with the power and authority to drive change were often satisfied with incremental progress, which showed that they weren’t simply resting on their laurels, but they often also had too much at stake to pursue truly disruptive change. A second group of innovators often had thought-provoking ideas and impressive prototypes, but could not translate those gains into meaningful impact.
The people that took the risk of trying something new, but also weren’t afraid to put in the immense effort required to make those new things work in the challenging environments of each global bank — those were the real changemakers. Some are recognized broadly for their contribution, but many are unsung heroes satisfied by their transformational and lasting impact on their firms and industry.
Today, the same organizational dynamics are playing out again across the broader enterprise world with AI.
Tackling Today’s Automation Trap
Right now, most enterprise "automation" is stuck in the past:
RPA bots that follow rigid scripts across off-the-shelf interfaces
AI chatbots that slightly speed up customer service tickets
Basic programmed workflows that collapse under complexity
These are the spreadsheets and Clack scripts of today's enterprises: useful but inherently limited. In a world that’s increasingly unstructured, dynamic, and complex, this approach is reaching its breaking point.
Unstructured Data, Real Reasoning and True Intelligence
The next leap forward isn’t about pushing buttons faster or patching broken workflows with prettier interfaces. It’s about mastering unstructured data — the emails, PDFs, contracts, instructions, and real-world artifacts that drive modern business. It’s about human-like reasoning: making judgment calls when the data is messy, adapting to novel situations, collaborating across systems, and learning continuously. It’s about AI agents, not bots.
AI agents that can:
Extract, understand, and structure insights from messy documents
Reason through ambiguity with incomplete information
Integrate with decades-worth of highly bespoke and specialized internal applications
Collaborate across business units, regulatory environments, and complex systems
Operate responsibly, dynamically, and autonomously at enterprise scale
This is not the world that legacy RPA and chatbots were built for. This is the world Artian was built for.
Artian’s Multi-Agent AI Systems: The Way Forward
At Artian, we’re not duct-taping yesterday’s workflows. We’re reimagining how enterprises work — and how automation works. Existing technology platforms struggle to adapt to a future accelerated by AI agents. We know because we were there.
We, the founders of Artian, each have over 25 years of experience operating at the intersection of AI, financial markets, and enterprise technology. We have led the design and deployment of mission-critical AI systems in some of the world’s most complex and high-stakes environments — JPMorgan Chase, Google, Morgan Stanley, Knight Capital — bridging cutting-edge research with real-world enterprise impact.
Before founding Artian, we were in prime transformational roles within the highest perches of Wall Street, but when we looked out onto the emerging landscape, we could not find the offerings that would help us achieve our desired outcomes. So, we committed ourselves to solving that problem for all the changemakers out there, especially within financial services.
We know how to do it because we’ve done it before:
We’ve built platforms reaching millions of users and handling trillions of dollars in critical flows.
We’ve deployed autonomous, multi-agent systems that navigate complex, competitive environments.
We’ve launched machine learning systems trusted with the world’s most sensitive customer data at massive scale.
This isn’t theory for us. It’s hard-earned experience and conviction. What we have set out to accomplish at Artian isn’t just disruption — it’s our next disruption. Welcome to the future of enterprise automation.
