Demo · multi-industry edition

Thirteen workflows. Agents reason, people decide.

Four building workflows, three workforce workflows, five insurance workflows, and one that runs inside the PMO itself, each a chain of AI agents. You pick the input, set the operating rules in plain English, and watch the agents hand work to each other. A person signs off before anything leaves the page. It runs live on real AI models; switch the badge to rehearsal for scripted outputs, so the story still works offline.

Why all six look the same. Every workflow here is the same four moves. The sensors change; the pattern does not. What we bring is delivery of the pattern: the definitions agreed before build, the person who owns the decision, and the measurement that proves the benefit.

1. Data in one place

BuildingsBills, meters, weather, asset attributes, retrofit options. Estimates flagged, never silent.

WorkforcePunches, schedules, pay registers, benchmarks, policies by province, collective agreements.

2. Agents reason

BuildingsRead, normalise, detect, diagnose, plan, price, draft. Each agent hands structured output to the next and says where a rule changed its call.

WorkforceForecast, schedule, check rules, audit pay, explain, score, recommend. Same hand-off discipline.

3. A person decides

BuildingsProperty manager, asset manager, ESG lead. Approve, dismiss, or send back. Dismissals tune the rules.

WorkforceSite manager, payroll practitioner, HR business partner. Nothing publishes or pays on its own.

4. Measure

BuildingsGJ, kWh, dollars, tonnes against a weather-normalised baseline; misalignment year moved.

WorkforceOvertime hours, exceptions per pay run, minutes per payroll cycle, regretted departures.

Guardrails every agent carries

Rules win. The operating rules you set are injected into every agent's instructions and outrank its defaults. Agents must say where a rule changed a call.
No invented numbers. Agents label every figure as given, calculated, or estimated, and give a range where they estimate.
Privacy. Workforce agents work on the minimum data, never profile individuals beyond the task, and flag anything that would need consent under BC privacy law.
A person owns the outcome. Drafts are drafts. The last step in every chain is a named human role, and the page records their decision.
Vendor claims are vendor claims. Any savings figure is compared with the independent range before it is booked.

Inputs are synthetic and written in the page. In live mode each agent is a large language model called through ResonAi's server, so no key is needed and nothing you type is stored. Rehearsal mode plays scripted outputs for the first preset of each workflow. Emission factors quoted: natural gas 49.87 kgCO2e/GJ (BC methodology), BC grid 9.9 gCO2e/kWh (2024), Ontario grid 30 gCO2e/kWh as an illustrative placeholder. All costs in Canadian dollars.

Thirteen workflows. Agents reason, people decide.

1. Data in one place

2. Agents reason

3. A person decides

4. Measure

Want one of these running on your data?