Guest · Customer experience and personalisation
Gareth Cummings
CEO, eDesk
Gareth Cummings is CEO of eDesk, the AI-led customer support platform for online sellers, processing more than 14 million conversations a month and supporting over $25 billion in annual eCommerce transactions. A Trinity College Dublin graduate and winner of the Kenneth Mullican Memorial Medal, he joined eDesk in 2018 as CTO and took over as CEO in October 2025 after a spell as Managing Director.
About
Gareth Cummings is CEO of eDesk, the AI-led platform that consolidates customer conversations across marketplaces, social channels and direct storefronts for online sellers. He trained at Trinity College Dublin, where he won the Kenneth Mullican Memorial Medal for academic distinction, and has spent more than two decades building the commercial instincts of a founder across Irish software.
His first technical role was as a web developer, where sitting in on customer meetings planted the commercial thread that would run through every subsequent job. He joined Globoforce (now Workhuman) as a systems lead in its thirty-person era, then spent five years at Sentenial before moving to Brite:Bill as Global Head of Delivery, building billing software for Comcast, T-Mobile and AT&T.
He joined eDesk in 2018 as CTO and held that role for seven years, shipping the platform’s first generative AI layer in October 2018, built on eDesk’s own e-commerce dataset four years before ChatGPT reframed the category. In November 2024 he stepped up to Managing Director, and in October 2025 took over as CEO. Under his leadership eDesk handles more than 14 million conversations a month for over 5,000 online sellers and underpins more than $25 billion in annual eCommerce transactions, with AI automation growth running above 100 percent year on year.
Career25 years
- 2018 to now eDesk CTO (2018 to 2025), Managing Director (2024 to 2025), CEO since October 2025
- 2014 to 2018 Brite:Bill Global Head of Delivery
- 2009 to 2014 Sentenial Head of Service Delivery and Technical Operations Manager
- 2006 to 2009 Globoforce now Workhuman
Show earlier roles: Various
- 2001 to 2006 Various
How Gareth thinks3 convictions
-
AI works when the dataset is narrow enough to train on
eDesk shipped its first generative AI layer in October 2018, four years before ChatGPT reframed the category for mainstream audiences. Cummings and his team had a structural advantage: a narrow e-commerce dataset where roughly ten query categories cover 70 percent of inbound volume. That constraint gave the team enough signal density to train a useful language model on modest compute, at a time when general-purpose customer support platforms faced a far harder problem. The lesson is that AI feasibility is often a data-scope decision long before it is a model-architecture decision.
-
Pre-sales is the most under-answered question in e-commerce
Pre-sales queries, the questions buyers ask before they commit, typically account for 30 to 60 percent of inbound support volume at an e-commerce business. They are routinely left unanswered because most support teams are trained for post-sale problem solving. Cummings frames this as the single highest-return behavioural change any online retailer can make. eDesk data shows that answering a pre-sales question within 15 minutes significantly lifts conversion, and his standing challenge to clients is simple: focus on pre-sales for a week and measure the difference.
-
Trust signals are about to become machine-readable
As consumer agents start transacting on behalf of their owners, trust stops being a soft brand attribute and starts being a filter. Shoppers will tell their agent to transact only with brands above a review threshold, or whose response times clear a specific bar. The brands that win in the agentic commerce world are the ones whose support responsiveness and customer satisfaction are already measurable, consistent and public today. Cummings sees this migration as the next phase of what eDesk has been building toward since 2018: customer support as the commerce signal that agents read.
