Focus area
Trigger-event agentic systems
Events that fire actions when conditions align.
Preselected profiles plus trigger events produce opportunities. This was a 2016 formulation of what 2026 calls an agent workflow. The same graph primitives — profile, trigger, action — apply to ad serving, supply chain, fraud, and civic response.
The 2015 direct-response patent proposed the architecture before the vocabulary existed. A preselected profile (identity + score), a trigger event (situation-based rule), and an opportunity (a routed atomic event that some actor can act on). Every field moment-based marketing platform of the last decade is a variant.
2026 generalizes the "advertiser" at the end of the pipeline. The actor who acts on the opportunity can be a care coordinator, a dispatcher, an agent making a payment, or a policy automaton unlocking a benefit. The shape of the pipeline is unchanged. What changes is (a) the telemetry richness feeding the trigger, and (b) the reversibility discipline required to let an LLM-driven agent pull the trigger.
The hardest new engineering problem isn't detection. It's reversibility — an agentic trigger that can send money or lock an account needs a policy envelope, a rate limiter, and a clean undo path. The Family 4 patent described the template. The 2026 build adds the safety surface.
Where it shows up
Three applications.
- 01
Chronic-care patient outreach
Trigger care-team intervention when a patient's logging falls off and their location suggests they are not returning to the clinic.
- 02
Workplace-safety escalation
Lone field worker enters high-risk polygon + biometric stress spike + no scheduled task → dispatcher agent checks in.
- 03
Agentic household replenishment
Pantry IoT reports low + phone location >50 miles from home → agent reroutes delivery to the travel address inside a pre-authorized policy envelope.
Related focus areas