Alea Abrams · AI Behavior & Conversation Systems Design

Designing the human
side of AI behavior.

10+ years in AI design

~2,000 people trained

I shape how intelligent systems behave in the moments that matter: when to answer, when to ask, when to act, when to explain, and when to slow down. My work sits between product design, model behavior, conversation design, and content systems, making AI clearer, more useful, and more accountable.

View selected work Get in touch

Here’s where I designed
how AI behaves.

Otto answering the one-word query “computer” by naming the readings it could mean and asking which one

Designing AI to Know When to Ask

Shaping how the organization handles ambiguous requests, moving from guessing wrong or giving a giant generic response to asking the right clarifying question.

10+enterprise customers demanded the feature back within weeks of its removal

Conversation Design Intent Understanding Organizational Advocacy Framework Development

Full case study →

The shipped AI Steps panel listing each step the agent took in plain language

Making AI’s Work Legible

Redesigning how autonomous and agentic AI communicates what it’s doing, moving from technical logs to plain language that builds trust.

0escalations to disable processing messages since the redesign shipped, teams had been turning the feature off

Agentic AI Trust Design User Mental Models Language Architecture

Full case study →

A shipped Otto answer to “what’s the parental leave policy”: it leads with the answer, names the policy that applies to where the user works, and cites its source

Making AI Behavior a Design Decision, Not a Model Default

Sustained advocacy that turned behavioral quality into architecture, from a designer’s opinion to the Otto PRD’s named source of truth.

~4×deflection: 11.4% vs. 2.75% for the legacy path, from early-access customer data

Advocacy Quality Rubric PRD Authorship Organizational Design

Full case study →

The Conversation design guidelines published in ServiceNow’s Horizon design system

Creating a Conversation Design Discipline at ServiceNow

Principles, training, patterns, and tools that shifted how a 1,000-person design org approaches AI behavior.

600+designers directly trained; enablement reached close to 2,000 people

Discipline Building Enablement Systems Thinking Thought Leadership

Full case study →

I design AI products for humans.

I’m Alea Abrams, a Staff Designer focused on model behavior and human-AI interaction, with over 10 years designing in the AI space.

My work sits between product design, model behavior, conversation design, and human-computer interaction. I shape how AI products understand intent, handle ambiguity, explain what they are doing, recover from uncertainty, and know when to ask, act, or pause.

At ServiceNow, I design enterprise AI experiences where user intent is complex, automation has consequences, and trust depends on repeated, reliable behavior. My work spans both the visible experience and the systems behind it: interaction patterns, conversation frameworks, behavior standards, evaluation criteria, content strategy, and cross-functional product decisions.

I’m especially interested in the messy middle between a person’s intent and a system’s response. That is where AI can be genuinely helpful. It is also where it can overstep, obscure uncertainty, or make people responsible for decisions they did not understand.

In short: I design AI products that know how to behave when the answer is not obvious.

Prompt Design Agentic AI UX Organizational Strategy Conversation Design AI Mental Models Designing Uncertainty

Pearls of wisdom.

Fun fact: Pearl is my middle name. A pearl forms because of friction, because something got in the way. I think good AI design does the same thing: it turns the right kind of friction into clarity, agency, and trust.

Design for the human, not the model

A prompt is only the surface of what someone needs. Good AI design accounts for the person, the moment, the context, and the consequence.

Humanity is not human-likeness

AI does not need to sound more human. It needs to work better for humans: clear, capable, honest, and appropriately restrained.

Friction is not failure

A question, confirmation, pause, or boundary can make an AI product better. Seamless is not always kind. Sometimes the most human thing a system can do is slow down.

Trust should be earned, not performed

Confidence, fluency, and speed can make AI feel trustworthy before it has earned that trust. Good design gives people evidence: what the system understood, what it is doing, why it is doing it, and how to recover if it gets something wrong.

Responsibility belongs in the product

Ethics cannot live only in principles, decks, or policy docs. It has to show up in defaults, constraints, escalation paths, confirmations, evaluation criteria, and the everyday behavior of the system.

Designing the human
side of AI behavior.

Here’s where I designed
how AI behaves.

Designing AI to Know When to Ask

Making AI’s Work Legible

Making AI Behavior a Design Decision, Not a Model Default

Creating a Conversation Design Discipline at ServiceNow

I design AI products for humans.

Pearls of wisdom.

Speaking & writing.

Looking for the next hard behavior problem.

Designing the human side of AI behavior.

Here’s where I designedhow AI behaves.

Designing AI to Know When to Ask

Making AI’s Work Legible

Making AI Behavior a Design Decision, Not a Model Default

Creating a Conversation Design Discipline at ServiceNow

I design AI products for humans.

Pearls of wisdom.

Speaking & writing.

Looking for the next hard behavior problem.

Designing the human
side of AI behavior.

Here’s where I designed
how AI behaves.