Synextra launches AI agents for high-volume regulated work

Automating regulated workflows

Synextra, a specialist in Microsoft Azure, is launching a new service called Agents of Output. This offering is specifically designed for governance-led sectors that handle high volumes of regulated work. The company aims to help industries like accountancy, legal services, and insurance automate repetitive but complex tasks.

The service focuses on tasks that require strict adherence to rules, such as conducting anti-money laundering checks. It can also handle the drafting of engagement letters, which are standard documents in professional services. By automating these processes, firms can manage large workloads more efficiently while maintaining regulatory standards.

The goal of the service is to allow organizations to scale their operations without needing to increase their total headcount. By offloading high-volume tasks to AI agents, human teams can spend more time on high-value work that requires personal expertise. This approach addresses the growing demand for efficiency in sectors where transparency and compliance are non-negotiable.

Managed AI and risk ownership

Unlike many traditional AI solutions, Synextra takes on the risk of the build and the ongoing operation of the agents. The company manages every agent on behalf of the client for as long as the service is in use. This means the client does not have to worry about the technical maintenance or the underlying infrastructure of the AI.

The pricing model is designed to provide measurable value by focusing on outputs rather than hours or user seats. There are no setup or consulting fees, and there are no charges based on the number of users or the hours spent on the project. Instead, each compliant output starts at a price of £0.05, plus a monthly platform fee.

Synextra also provides a guarantee regarding the viability of the automation. CEO Chris Piggott stated that if an agent is not worth building, the company will say so and will not bill the firm for finding that out. This ensures that clients only invest in automation that provides a genuine return on investment.

Integration and testing process

The process for deploying an agent begins with a free design phase offered by the company. During this stage, engineers map out the current workflow and agree on the scope, timeline, and acceptance tests before any commitment is made. This ensures that the final product aligns with the specific needs of the organization.

Once the design is finalized, engineers connect the AI agent directly into the firm's existing systems. This integration allows for immediate value because it does not require the client to undergo additional training or make significant system changes. The agents are tested using the firm's own work to ensure they pass the agreed-upon acceptance tests before going live.

During the development phase, the engineers work closely with the firm to ensure the AI behaves as expected. They monitor the performance and adapt the agent as processes or requirements change over time. This hands-on approach ensures that the quality of the output remains high and that any issues are addressed proactively by the specialists.

Accuracy and accountability

A key feature of Agents of Output is its ability to provide a clear audit trail for every action taken. The service is designed for organizations that must be able to demonstrate exactly how each output was produced. This level of transparency is critical for staying compliant with various industry regulations.

The agents follow a strictly defined specification and are programmed to refer any exceptions to human staff. This ensures that while the AI handles the bulk of the work, humans remain in control of the process. All changes made by the agents are logged and reversible, which maintains accountability for every piece of data produced.

In a recent deployment for a UK conveyancing group, the agent was used to extract special conditions from mortgage offers. During the trial phase, the agent identified 99.9 per cent of the conditions with no false matches. This high level of accuracy demonstrates the technology's ability to handle nuanced data in a regulated environment.