Making a Mind Live from SXSW: When AI Agents Leave the Lab
Summary
Everyone's talking about AI agents, but what actually happens when companies deploy them to do real work? This live recording of the "Making a Mind" podcast explores the gap between how we think agentic AI should work and how it actually behaves in the wild. Cognitive scientist and podcast host Dr. Danielle Perszyk sits down with Michael Giannangeli, head of product, agentic AI for Amazon Nova, plus industry voices from Amanda Doerr, vice president of core shopping at Amazon and more who've recently launched agents into production. Together, they'll dissect real implementation stories—the successes, the surprises, and the moments where human expectations collided with AI reality. We'll explore the fundamental questions emerging as AI agents become collaborators: Where do humans want control versus autonomy? How do we build trust with systems whose reasoning we can't see? And how does cognitive science research translate into product decisions—from tool-calling architecture to the design choices that determine when agents act independently versus ask for help? This isn't a demo. It's an honest conversation about what we're learning as agentic AI moves from theory to practice—and what it reveals about both artificial and human intelligence.
- Format: Panel
- Category: Panel
- Event Type: panel
- Presented By: n/a
Contributors
- Danielle Perszyk
- Michael Giannangeli
- Amanda Doerr
Raw Event JSON
Large nested arrays are compacted for page readability. Full payload remains in dataset files.
Open JSON payload
{
"id": "PP1150544",
"track": "Tech & AI",
"focus_area": null,
"category": "Panel",
"event_id": "PP1150544",
"event_type": "panel",
"format": "Panel",
"genre": null,
"subgenre": null,
"name": "Making a Mind Live from SXSW: When AI Agents Leave the Lab",
"presented_by": null,
"publish_at": "2026-02-02T14:17:00.000-06:00",
"reservable": true,
"reservable_id": "PP1150544",
"reserved": false,
"date": "2026-03-18",
"end_time": "2026-03-18T12:30:00.000-05:00",
"start_time": "2026-03-18T11:30:00.000-05:00",
"message": null,
"image": null,
"credentials": [
{
"name": "Platinum Badge",
"type": "platinum"
},
{
"name": "Innovation Badge",
"type": "innovation"
}
],
"contributors": [
{
"id": 89012,
"company": "Amazon",
"credential_types": [
"innovation"
],
"entity_id": 2251307,
"name": "Danielle Perszyk",
"speaker_types": [],
"title": "Cognitive Scientist & Head Of Human-Computer Interaction, Amazon Agi",
"image": {
"alt_text": "Danielle Perzszyk",
"credit": "",
"url": "https://images.sxsw.com/nTV2iWgPliKb2GYEXmkG6cuXos8=/0x0:3215x3215/600x600/images.sxsw.com/48/36a489b1-f0e5-4c6d-9b2b-5afff80a9979/pcid-1637331"
},
"speaker_type": "Speaker",
"details": "Danielle Perszyk, PhD, cognitive scientist and member of the technical staff, Amazon AGI SF Lab,\rleads the human-computer interaction (HCI) team at the Amazon AGI SF Lab, which is\rdeveloping foundational capabilities for practical AI agents that can act in both digital and\rphysical environments. She received her PhD from Northwestern, where she studied language\revolution and the development of social cognition. Prior to Amazon, she was at Adept and\rGoogle.",
"links": {},
"type": "person"
},
{
"id": 89011,
"company": "Amazon",
"credential_types": [
"innovation"
],
"entity_id": 2251306,
"name": "Michael Giannangeli",
"speaker_types": [],
"title": "Head Of Product, Agentic AI - Nova",
"image": {
"alt_text": "Michael Giannangeli",
"credit": null,
"url": "https://images.sxsw.com/HAuQTO5YCbs-byNigmahhcsO7Dw=/139x0:755x616/600x600/images.sxsw.com/48/5cf3ab8a-9e04-4ebf-8c5b-599c1f7d8f9d/pcid-1637326"
},
"speaker_type": "Speaker",
"details": "Michael leads Agentic AI for Amazon’s Nova foundation models, focused on building the capabilities for intelligent systems to reason, act, and scale in the real world. Before that, he helped grow Alexa from 1M to 100M+ customers, culminating in the launch of Alexa+. Along the way, he led the Echo and Echo Dot product lines to become the #1-selling product on Amazon.com, and founded Alexa’s computer vision initiative to make Alexa more personalized and adaptive. Before Amazon, Michael earned an MBA from MIT, worked as an economic consultant, interned on Wall Street and Capitol Hill, and played NCAA basketball.",
"links": {},
"type": "person"
},
{
"id": 89010,
"company": "Amazon",
"credential_types": [
"innovation"
],
"entity_id": 2251305,
"name": "Amanda Doerr",
"speaker_types": [],
"title": "Vice President, Core Shopping",
"image": null,
"speaker_type": "Speaker",
"details": "Amanda leads Amazon's Core Shopping organization, overseeing the end-to-end customer experience across homepage, navigation, product pages, reviews, and Alexa Shopping worldwide. She's at the forefront of AI-powered shopping innovation, leading teams that help customers discover, evaluate, and purchase products both within Amazon's store and across the web through agentic features like Buy for Me, which enables direct purchases from brand websites. Since joining Amazon in 2010, Amanda has driven product innovation across Kindle, Supply Chain Optimization Technology (SCOT), and Shopping. Prior to Amazon, she spent a decade at Microsoft leading global product launches for Windows and the Entertainment & Devices group.",
"links": {},
"type": "person"
}
],
"venue": {
"accessible": false,
"age_policy": null,
"floor": "3",
"id": "V0455",
"indoor_outdoor": "INDOOR",
"name": "Salon E",
"parent_id": "V0403",
"parent_venue_name": "JW Marriott",
"venue_entry_info": null,
"formats": null,
"location": {
"address": "110 E 2nd St.",
"lat_lon": [
30.2645713,
-97.743198
],
"city": "Austin",
"postal_code": "78701",
"state": "TX",
"name": "JW Marriott"
},
"root": {
"accessible": null,
"age_policy": null,
"floor": null,
"id": "V0403",
"indoor_outdoor": null,
"name": "JW Marriott",
"parent_id": null,
"parent_venue_name": null,
"venue_entry_info": null,
"formats": null
},
"events": "[omitted 5 venue events]"
},
"description": "Everyone's talking about AI agents, but what actually happens when companies deploy them to do real work? This live recording of the \"Making a Mind\" podcast explores the gap between how we think agentic AI should work and how it actually behaves in the wild. Cognitive scientist and podcast host Dr. Danielle Perszyk sits down with Michael Giannangeli, head of product, agentic AI for Amazon Nova, plus industry voices from Amanda Doerr, vice president of core shopping at Amazon and more who've recently launched agents into production. Together, they'll dissect real implementation stories—the successes, the surprises, and the moments where human expectations collided with AI reality. We'll explore the fundamental questions emerging as AI agents become collaborators: Where do humans want control versus autonomy? How do we build trust with systems whose reasoning we can't see? And how does cognitive science research translate into product decisions—from tool-calling architecture to the design choices that determine when agents act independently versus ask for help? This isn't a demo. It's an honest conversation about what we're learning as agentic AI moves from theory to practice—and what it reveals about both artificial and human intelligence.",
"experience_level": "Beginner",
"accessibility": [
"Closed captioned"
],
"accessible_venue": false,
"add_bcl_url": false,
"add_slido": false,
"age_policy": null,
"american_sign_language": false,
"apple_url": null,
"audio_description": false,
"caption_url": null,
"closed_captioned": true,
"cpe_credit": false,
"has_subtitles": false,
"hash_tags": [],
"high_sensory_experience": false,
"live": false,
"long_description": "Everyone's talking about AI agents, but what actually happens when companies deploy them to do real work? This live recording of the \"Making a Mind\" podcast explores the gap between how we think agentic AI should work and how it actually behaves in the wild. Cognitive scientist and podcast host Dr. Danielle Perszyk sits down with Michael Giannangeli, head of product, agentic AI for Amazon Nova, plus industry voices from Amanda Doerr, vice president of core shopping at Amazon and more who've recently launched agents into production. Together, they'll dissect real implementation stories—the successes, the surprises, and the moments where human expectations collided with AI reality. We'll explore the fundamental questions emerging as AI agents become collaborators: Where do humans want control versus autonomy? How do we build trust with systems whose reasoning we can't see? And how does cognitive science research translate into product decisions—from tool-calling architecture to the design choices that determine when agents act independently versus ask for help? This isn't a demo. It's an honest conversation about what we're learning as agentic AI moves from theory to practice—and what it reveals about both artificial and human intelligence.",
"meeting_url": null,
"meeting_url_live": false,
"mentorly_url": null,
"mobile_audio_url": null,
"open_captioned": false,
"playlist_tag": null,
"rebroadcast": false,
"recommended_ids": [],
"search_conversions": null,
"short_program": false,
"slido_url": null,
"sort": "MAKING A MIND LIVE FROM SXSW: WHEN AI AGENTS LEAVE THE LAB",
"source": "Official",
"squad_up_url": null,
"stream_embed": null,
"stream_id": null,
"stream_url": null,
"strobe_warning": false,
"subdiscipline": null,
"summit": null,
"summit_display_name": null,
"tags": [
{
"id": 1343,
"name": "Technology",
"created_at": "2017-10-11T17:42:23.253-05:00",
"updated_at": "2024-09-13T07:02:24.613-05:00",
"sort": "TECHNOLOGY",
"alpha": "T"
},
{
"id": 2114,
"name": "AI",
"created_at": "2020-01-17T14:51:44.091-06:00",
"updated_at": "2026-02-17T20:42:14.828-06:00",
"sort": "AI",
"alpha": "A"
},
{
"id": 2548,
"name": "Podcast",
"created_at": "2025-01-22T12:49:10.822-06:00",
"updated_at": "2025-01-22T12:49:10.822-06:00",
"sort": "PODCAST",
"alpha": "P"
}
],
"talent_attending": null,
"title_only": false,
"track_display_name": "Tech & AI track sponsored by IBM",
"trailer_id": null,
"trailer_url": null,
"video_on_demand": null,
"vimeo_id": null,
"vod": false,
"xr_project_type": null,
"youtube_id": null,
"related_sales_client": null
}