The latest news from Warrington based Synextra who know AI agents are only as good as the built in compliance. Here’s the word;
Synextra, the Warrington-based Microsoft Azure specialist, has today launched Agents of Output, a new fully managed service that designs, builds and runs AI agents for governance-led sectors such as financial services, legal and insurance. The agents handle high-volume, regulated work, from drafting engagement letters to running anti-money laundering checks. They are designed to help organisations grow without a matching rise in headcount, giving teams more time for work that needs their expertise. Because the agents work inside existing systems, in most cases there is nothing new to learn and the benefit starts from day one.
Firms that buy AI agents usually carry the risk that they will not work, and are then left to run them themselves. Agents of Output takes on the risk of each build, and its engineers run every agent on the organisation’s behalf for as long as it is in use. There are no set-up or consulting fees and no charge per seat or hour. Each output that meets the agreed standard costs from £0.05, plus a monthly platform fee.
Chris Piggott, CEO of Synextra and Agents of Output, said: “AI rarely fails in the demo. In governance-led firms, it fails at sign-off, so that’s what we build for. If an agent isn’t worth building, we’ll say so, and we won’t bill a firm for finding out. When it is, we agree the standard every output must meet before we build, and we don’t hand it over and walk away.”
That standard is set during a design phase with no consulting fees. Agents of Output maps the current process and agrees the scope, timeline and acceptance test, so the business case is clear before any commitment.
During the build, Agents of Output’s engineers connect the agent to existing systems, so nothing is replaced. They then run it on the firm’s own work to make sure it passes the acceptance test before it goes live.
Once the agent is live, Agents of Output manages it in full. Its engineers stay on hand throughout, working alongside the people who use it and adapting the agent as processes, volumes and needs change. They monitor its performance and apply model updates, so any drop in quality is caught before the organisation notices. No in-house AI expertise is needed.
One of these agents was built for a UK conveyancing group, where every mortgage offer had to be read in full to find the special conditions. These are the requirements a lender sets before it releases the money. Because each lender words and places them differently, every offer took considerable time. The agent now pulls out each condition and shows where it appears, so conveyancers check the findings rather than the whole document. In trials on offers the group had already checked, it found 99.9% of conditions with no false matches. The group now has access ahead of a full rollout.
As in that example, the firm’s own people stay in control. Each agent makes only the decisions agreed in its specification and refers anything else to staff. Every change is logged so it can be reversed. This matters because whoever signs off the work remains accountable for it, even when an agent produced it.
Agents of Output is available from today. Free discovery sessions at agentsofoutput.com help identify the processes worth automating.
Be the first to comment