Em Shreve

Em Shreve builds tools and learning programs that help engineers and non-engineers put AI to work. As Head of AI Enablement at Automattic at the time of the 2026 AI Engineer World’s Fair talk, Shreve described an approach built around dedicated time to learn, real projects, and continued support from colleagues. That work follows an engineering career across WordPress.com, WooCommerce, Tumblr, and Pocket Casts, alongside open-source contributions that make WordPress capabilities accessible to agents.

From WordPress contributor to AI enablement

Shreve’s career began with Google Summer of Code projects for WordPress in 2009 and 2010. At Automattic, the work progressed from internal tooling to WordPress.com’s social features and partnerships, followed by WooCommerce Core from 2015 to 2020. Shreve subsequently worked on Tumblr, Pocket Casts, and Tumblr migration before moving into Applied AI in 2025.

Open-source contributions continued alongside those roles. Shreve contributed to WordPress releases 4.4 and 4.7 and to WooCommerce Core. Participation in the WordPress AI team extends that product engineering experience into infrastructure for agents: giving software a consistent way to discover what a site can do and invoke those capabilities.

Making WordPress capabilities usable by agents

The WordPress Abilities API gives plugins and other parts of WordPress a common way to describe operations, their inputs and outputs, and their permission requirements. Shreve contributed its REST API controllers and JavaScript client package. The client lets browser code discover available abilities, retrieve an individual ability, and execute it—for example, fetching a list of pages or creating a draft post. These contributions connect the collaborative project’s registered capabilities to the interfaces that use them.

Shreve also created the WP Ability Toolkit, a development and testing environment for those integrations. Its assistant can navigate the WordPress dashboard, reload pages, and guide developers through creating additional abilities. Bundled Claude Code skills help scaffold plugins, validate ability definitions, and manage a local WordPress environment. The toolkit provides a practical development loop: create a capability, test how an agent uses it, and extend what the assistant can do.

Giving people time and support to change their work

Shreve’s approach to AI fluency starts with a concrete organizational commitment: two weeks of dedicated, in-person learning, divided between workshops and real project work. Engineers, cross-functional groups, and non-technical teams—including finance and leadership—apply the tools to tasks they need to complete. The argument is that tool access alone does not provide the time or experience people need to change their habits.

Several parts of the approach make that learning useful beyond the initial program:

• Projects grounded in participants’ work: The programs combine instruction with building. Examples included a Zendesk connector, a shared skills directory, and a tool built by a support teammate. The support example illustrates why practical AI learning extends beyond engineering roles.
• Graduates who help colleagues: Guides, local meetups, and AI ride-alongs spread participants’ experience through demonstrations of actual workflows. An ongoing help channel, together with Shreve’s office hours, workshops, and internal tooling, gives people ways to continue learning after the intensive program ends.
• Follow-up over time: Surveys at 72 hours, 30 days, and 90 days provide repeated opportunities to assess participants’ experience after the program, rather than relying only on their immediate reaction.

Bringing image tools into development

Shreve’s Nano Banana image plugin addresses another practical obstacle: producing icons, diagrams, and cleaned-up screenshots while working in Claude Code. Shreve built the plugin to connect the coding assistant to Google’s Gemini image model, allowing users to generate and edit images through natural-language requests within the coding workflow. Like the WordPress toolkit, it makes a capability available where someone is already doing the work.