Intercom’s Declan Ivory: AI-first support needs people to own how the system works

Intercom’s Declan Ivory: AI-first support needs people to own how the system works

Declan Ivory argues that when AI Agents handle most customer conversations, support teams need to shift from managing queues to improving the system that handles them. In his account, this is not simply a matter of adding automation: teams need clear ownership of content, behavior, workflows, and ongoing improvement. Ivory proposes four foundational roles. An AI operations lead tracks performance and prioritizes fixes; a knowledge manager maintains the content the AI Agent uses; a conversation designer shapes its communication and handoffs; and a support automation specialist builds workflows and backend actions. He says teams can distribute these responsibilities across existing staff before making them formal roles. Other support functions change too. Enablement trains people to work with the AI Agent, QA reviews its behavior and customer experience, and workforce management plans around automation coverage. Ivory emphasizes leaders who can work on the system and coach people, a “player-coach model.” He gives examples from Dotdigital, Clay, Lightspeed, and Intercom, and presents the approach as a pattern he has seen across teams. The organizational stakes in his argument are that responsibility for improving AI performance becomes part of support work when support depends on that performance. The article does not report measured results showing whether these roles improve.

Publication: 23655d88-628e-45fe-8895-8ffb62d970ee

attribution — pass: The byline names Declan Ivory and the publisher is Intercom, so attributing the argument to Intercom's Declan Ivory is supported. The source often speaks as 'we' (Intercom). Describing the recommendations as his argument, and the pattern as one seen across teams, stays within that framing.

That’s the pattern we’ve seen from working with so many teams:

summary — uncertain: The reading is mostly faithful and accessible. However, its final sentence breaks off: 'whether these roles improve.' never says what the roles might improve, which leaves the key limitation unclear. Smallest repair: complete it as 'improve AI Agent performance or customer experience.' It would also help to note that Fin, named in the examples, is Intercom's own AI Agent product. That shows readers that the cited teams use the publisher's product.

At Clay, a dedicated GTM engineer role has been established as part of the ops team with a focus on making support more efficient at scale using Fin.

date — pass: The article's byline date is 25 November 2025, which matches originalPublished 2025-11-25 at day precision.

Declan Ivory25 November 2025 Reading time: 5 min

support — pass: Each claim matches its excerpt. The interpretation that the examples come without measured results is accurate: the source describes role changes at Dotdigital, Clay and Lightspeed but reports no outcome metrics. The question and the limitation do not overstate the evidence.

Lightspeed created a dedicated Digital Engagement team to manage Fin’s optimization

media — pass: The selected media is the verified direct link to the original Intercom article. That is the most direct way for readers to reach the source.