Financial institutions have invested heavily in artificial intelligence, deploying coding assistants, automated workflows and increasingly sophisticated AI agents. But how can banks translate improvements in individual productivity into meaningful efficiency gains across the organisation? And as agents take on more complex workflows spanning multiple business functions, how can institutions maintain control over their activities, access to enterprise systems and accountability?
Promenaut, the enterprise AI orchestration platform provider, is addressing these challenges with a control layer designed to coordinate human employees and AI agents. Developed in collaboration with HSBC, its platform operates across existing cloud, AI model and knowledge infrastructure, providing a common framework for managing agentic workflows.“AI has made individuals faster,” says Monika Fernando, Global Strategic Account Manager at Promenaut, speaking to TradingTech Insight. “Someone working within a particular process can be much more efficient, but those improvements aren’t necessarily translating into wider organisational productivity. When you have a workflow involving multiple people and handovers between different teams, you still encounter bottlenecks. For example, a lot of banks have rolled out AI coding tools to their developers, and they’re now producing code much faster. But the quality of the requirements they’re being given hasn’t necessarily improved, so the bottleneck has effectively moved upstream to the requirements stage,”
Promenaut’s approach is to connect specialised agents across business functions, allowing them to share information and coordinate activities as work passes between teams. This extends beyond automating individual tasks to managing processes involving multiple agents, employees and enterprise applications.
A Common Control Layer
Promenaut supports models from providers including Anthropic, OpenAI and Google, and is designed to operate across different cloud environments without requiring institutions to standardise on a single provider.
Its AI Resources (AR) framework, modelled on human resources functions, provides detailed execution logs, agent performance scoring and alerts when performance deteriorates. Agents can monitor other agents, with human intervention incorporated where required. Access controls mirror those assigned to employees, restricting the data and workflows available to individual users.
“Promenaut is the operating layer for agentic workflows in regulated enterprises,” explains Fernando. “We sit above the existing technology stack, bringing people and AI agents together as one governed workforce — with oversight of what every agent does across it. We include a governed knowledge graph: clients can connect an existing one or build on ours, with full governance over how agents access and build on it. We also ship a factory of pre-built agents and workflows — teams of agents doing defined pieces of work that clients can use from day one, extend, or build on top of.”
The availability of pre-built agents and workflows raises the question of how much of this functionality banks need to source from specialist providers, rather than develop themselves. Many financial institutions already have established cloud environments and internal development teams, and are building their own agentic capabilities using existing infrastructure.
Fernando argues that these implementations frequently remain focused on individual agents, while coordinating multiple agents across different systems and providers introduces additional requirements. Whether institutions need a dedicated orchestration platform will depend on the complexity of their deployments and the extent to which they want governance capabilities independent of their underlying technology providers.
Capital Markets Applications
Software delivery is among the areas attracting interest from financial institutions. Promenaut’s requirements intelligence capabilities can generate and evaluate requirements documents, drawing on organisational knowledge to identify relevant precedents, including previously rejected proposals. The platform can then create Jira tickets and reprioritise development backlogs.
In capital markets, front-office applications are currently more focused on supporting sales and client servicing than trading execution, where regulatory concerns remain a constraint, as Fernando explains. “From a front-office capital markets perspective, where we see a big opportunity is on the sales side, particularly around client intelligence. For example, salespeople often start their day by reading the news and distributing information through client chats. Much of that content is quite generic. The opportunity is to personalise it for individual clients and automate some of the manual work involved.”
Other applications include updating CRM systems from meeting notes, generating follow-up communications and evaluating business initiatives against organisational objectives, such as proposed expansion into new markets.
Regulatory change management is another potential area, with agents monitoring regulatory announcements, interpreting new requirements and identifying affected systems. In post-trade operations, Promenaut is exploring workflows that anticipate settlement failures and initiate communication with counterparties to resolve discrepancies before settlement. These applications are at different stages of development and shouldn’t all be regarded as established production deployments.
Deployment and the Economics of Scale
Promenaut offers both self-hosted and vendor-hosted deployment options. Larger financial institutions are expected to favour containerised deployments within their own environments, allowing them to retain control over their data and infrastructure. The company is also exploring more modular offerings that would enable clients to adopt individual workflows without deploying the entire platform.
AI expenditure is becoming another operational consideration as institutions increase their use of coding assistants and agentic workflows. Token consumption can rise rapidly, making it harder to establish the costs of individual applications and assess their contribution to business performance.
“One of the things our platform allows you to do is allocate AI spending by team, or even by individual user if you want that level of granularity,” says Fernando. “It can also identify situations where you don’t need to use the latest or most expensive model to achieve the required result. If a less expensive model can deliver the same quality of output, the platform can recommend that alternative. You can then attribute the costs to individual teams and start to measure the return on investment. You can see how much a particular team is spending on AI and assess what value it’s getting from that expenditure.”
As financial institutions expand their use of AI agents across multiple business functions, the ability to coordinate their activities, maintain effective governance and manage costs will become increasingly important. Dedicated orchestration platforms such as Promenaut offer one approach to bringing these capabilities together, supporting the transition from isolated AI applications to more integrated, enterprise-wide workflows.
Subscribe to our newsletter
