For organisations considering AI agents, the most useful questions are practical: what work can the system carry out, what information does it need and when should a person step in? Above all, how will the organisation know whether it is making a difference?
The Agentic AI Innovation of the Year category sponsored by Microsoft Ireland at the 2026 AI Awards offers four examples. Future Proof Knowledge, Oac Snacks, SPRYT and yieldHUB apply the technology to care administration, business planning, patient appointments and semiconductor testing.
Their shortlisted projects show how organisations are exploring AI agents across very different settings, each with its own requirements for accuracy, oversight and performance.
What is agentic AI?
An AI agent uses tools and information to work through a task, choosing its next steps within the permissions it has been given.
The degree of independence matters. Anthropic distinguishes between workflows and agents: workflows follow predefined paths, while agents can dynamically direct their processes and tool use. Its advice is to start with the simplest solution that meets the need, as greater autonomy can introduce additional cost and complexity.
For a business evaluating a system, the important question is what it can do, how reliably it can do it, and where human judgment remains essential.
Meet the Finalists
Future Proof Knowledge uses AI agents to support regulated care workflows, including clinical deadlines, reauthorisation, caregiver training and billing. Its architecture connects structured patient context with verified information, helping teams automate tasks while retaining traceability and oversight.
Oac Snacks developed a multi-agent planning system supporting commercial, supply and finance decisions. Defined agent roles, evidence-based debate and human approval help challenge assumptions, assess forecasts and keep business decisions traceable within clear limits on autonomy.
SPRYT’s Asa is an AI agent managing appointment booking, rescheduling, reminders and patient queries through familiar messaging channels. Conversational AI and predictive analytics support appointment attendance, helping healthcare teams coordinate patient journeys and reduce administrative workload.
yieldHUB Live brings agentic AI to semiconductor testing, with agents monitoring data, analysing performance and supporting operational decisions. The platform helps manufacturers improve yield, reduce unnecessary retesting and use existing production capacity more efficiently through real-time insights.
Measuring what matters
These applications support different kinds of work, so their success should be measured against the processes they are intended to improve.
In care administration, that could mean checking whether tasks are completed accurately and on time, and whether staff can trace actions back to their sources. In business planning, it could mean assessing whether recommendations make assumptions clear and help decision-makers consider alternatives.
For appointment coordination, useful measures could include successful bookings, unresolved requests and the workload passed back to staff. In semiconductor testing, an evaluation could compare the system with existing practices, examining errors, unnecessary repeat work and operating costs.
These are suggested evaluation measures, rather than reported results from the finalists. The starting point is a clearly defined process and a baseline against which progress can be measured.
Start with the task and the boundaries
“The first question for a leadership team is which task an AI agent should be trusted to carry out. Start with a clear business problem, define where human approval is needed, and measure the result before expanding.” – Mark Kelly, Founder of AI Ireland
Before launching a pilot, teams should agree on three things:
The task: Define the process, the information required and the problem being addressed.
The permissions: Establish what the agent can read, recommend and carry out, including where approval or escalation is required.
The evidence: Record a baseline and measure the full process, including the time people spend checking and correcting outputs.
An agent’s responsibilities should expand as evidence supports it. Recording exceptions and failures helps teams understand where the system needs improvement and where human involvement remains necessary.
Join us at the Irish AI Awards 2026
Join us on Tuesday, 17 November 2026, at the Radisson Blu Royal Hotel, Golden Lane, Dublin, as we celebrate Ireland’s AI community’s outstanding achievements and reveal this year’s winners.
Tickets are available now. Secure your place and join us for an event celebrating the people, organisations and innovations shaping the future of artificial intelligence in Ireland.
Discover more from AI Ireland
Subscribe to get the latest posts sent to your email.
