You'll embed yourself in different areas of the company and operate part of their processes to understand how decisions are really made: what information is used, what criteria are applied, what exceptions come up and what risks exist.
You'll identify where an agent can improve a decision, where human oversight is needed, and where the decision must remain one hundred percent human. You'll design that human-agent orchestration explicitly.
You'll own the full cycle of one or several agents: problem exploration, goal definition, building, deployment, measurement and continuous optimisation.
You'll design and build agents, defining their instructions, information sources, tools, permissions, workflows and escalation mechanisms.
You'll connect agents with internal and external tools through APIs, MCP servers, databases, automations and SaaS platforms.
You'll create fast prototypes, test them on real cases, and turn errors and results into concrete improvements.
You'll define evaluations and control mechanisms to measure the quality, reliability and impact of the agents on decision quality, productivity and customer experience.
You'll work shoulder to shoulder with the business teams where the agents are deployed and with the technical teams across AI, Data and Engineering.
You'll work directly with tuio's Head of AI and with the co-founder in charge of AI Deployment. This is a mission-critical role for the company, with visibility and responsibility from day one.
During your first months, you'll embed in one of tuio's priority domains. We expect you to:
Learn the domain and operate enough of the process to understand its key decisions and exceptions.
Identify and prioritise opportunities based on their impact on decision quality, frequency, feasibility, effort and risk.
Take ownership of one or two agents, from understanding the problem to a first solution used on real cases.
Define a baseline and an evaluation system to measure whether the agents are improving the operation's decisions.
Iterate on the solutions based on their errors, metrics and the feedback of the people who use them.
The goal won't be to ship a demo, but to prove that an agent improves real business decisions in a way that is measurable, reliable and responsible.
