Voice AI inside IICRC S520 protocols.

Mold intake under S520 has a sharper protocol boundary than water under S500, and an AI agent's scope has to match it. Restoration operators running mold remediation work should look at AI voice agents through a different lens than they would for water intake. The protocol drives the conversation and the conversation drives the scope.

How S520 differs from S500

S500 covers water damage with three categories ranging from clean to grossly contaminated. The protocol decisions on a water loss are primarily about source water, drying approach and material disposition. S520 covers professional mold remediation with three conditions ranging from normal indoor environment to actual mold growth. The protocol decisions on a mold job are about containment, work practices, post-remediation verification and worker protection.

The intake conversation for mold work has different inputs. The presence of visible mold growth. The estimated affected area in square feet. The HVAC system status. The occupant health considerations. Whether the building has been water-damaged previously. Whether anyone has done sampling. These data points map into S520 condition determination, which maps into the containment approach and the work scope.

What an AI agent captures on mold intake

An AI voice agent built for restoration mold intake captures the S520-specific data points in the same way it captures the S500 data on water intake. Visible mold presence and location. Affected area estimate. HVAC status. Occupant sensitivity or health concerns. Building history of water damage. Previous sampling, if any. Photographs of visible growth. The agent walks the homeowner through these items in a conversational form.

The protocol mapping is similar to S500. Inputs map to a preliminary condition determination and a recommended containment approach. The dispatcher receives a structured intake with the condition assessment and the equipment recommendation. The estimator gets the inputs needed for a defensible scope.

Where the AI scope is narrower on mold

The AI scope is narrower on mold than on water for two reasons. First, S520 condition determination has a sharper boundary than S500 category determination. A misclassification on mold has greater downstream consequences, including potential health exposure issues. The threshold for AI confidence in the condition assessment has to be higher.

Second, mold jobs are more likely to involve occupant health concerns that the AI cannot adequately triage. A homeowner reporting respiratory symptoms or asthma is a case that needs human attention, both for the scope question and for the customer relationship. The AI should flag and escalate.

What stays with humans

On mold intake, the AI hands off to humans whenever the call involves occupant health concerns, visible mold growth beyond a small contained area, building situations with multiple ownership parties, or any case where the homeowner is asking about sampling or testing. These cases need an estimator or PM on the phone, not an intake agent.

The AI's job on these calls is to recognize the pattern early and route to human within the first 60 seconds. A slow handoff on a mold call is worse than no AI at all, because the homeowner has already started a conversation with a system that cannot complete it.

The pilot considerations

A company running AI voice for mold intake should pilot it on a narrow case set first. Smaller affected areas. No reported occupant health concerns. Single-ownership buildings. The company builds confidence in the AI's handling of the bounded cases before expanding scope. This is more conservative than the water-intake pilot pattern and the right approach for the domain.

The estimator's role

On mold work, the estimator's role in the workflow is larger than on water work. The estimator does the protocol determination, the scope and the pricing. The AI's contribution is the front-door intake that feeds the estimator clean inputs. The companies that have integrated AI voice into mold operations have done so with the estimator's process at the center of the design. The AI serves the estimator, not the other way around.

Positioning AI voice for mold work

The right way to position AI voice for mold restoration is as a bounded-scope intake tool that handles the structured data capture and routes the judgment calls to the estimator. The company that expects the AI to handle every mold call end to end will hit edge cases that should not have been handled by AI. The company that uses the AI to clean up the intake data and protect estimator time gets the value. The framing matters.