What capability is essential when designing AI agents for teams?

Paula Goldman

Chief Ethical and Humane Use Officer & EVP, Product · Salesforce

Human at the Helm

Paula Goldman says the essential design principle is "human at the helm": rather than human in the loop, where AI drafts and people approve, systems should put the right things for human judgment at the right time, since agents reason through complex tasks at speed.

The answer starts at 3:19.

JE

Jonathan Evens

Product Lead · Google Deepmind

User Empathy

Jonathan Evens says user empathy is essential, saying to focus on deep user problems and pain points and to start with the problem.

Angela Jiang

Head of Product, Claude Platform · Anthropic

Natural Language Interaction

Angela Jiang says natural language interaction—just talking to the model—is the core piece, because the most magical outcome true to growing model capabilities is that talking to it is all the help needed.

TO

Todd Olson

CEO / Founder · Pendo.Io

Contextual User Data

Todd Olson's answer is contextual data about users and workflows: good information about what people use and where they get stuck, because that may lead an agent to make different decisions, like not changing heavily used workflows.

Duncan Lennox

Chief Product & Technology Officer · HubSpot

Agent Builder Capability

Duncan Lennox highlights HubSpot's Agent Builder capability, saying it democratizes access to agentic AI by letting customers and partners tailor the last 5-10% of out-of-the-box agents to their own go-to-market teams and business.