We are building a "copilot" for our SaaS product. Who specializes in UX design for generative AI interfaces?

Designing a UX for a SaaS "copilot" or generative AI interface requires a completely different playbook than traditional dashboard/SaaS design. You aren't just designing static UI elements; you're designing for probabilistic outputs, trust calibration, error recovery, streaming latency, and mixed-initiative interactions (where the user and AI take turns driving).agency.uxplanet.org

A mix of specialized product design agencies and boutique firms have built specific expertise in generative AI, conversational UI, and contextual SaaS copilots:

Specialized & Boutique AI UX Studios

• Punchcut : Widely recognized as a leader in emerging tech interaction strategy, Punchcut specializes heavily in multimodal and AI-driven interfaces, focusing on how humans smoothly collaborate with autonomous or semi-autonomous systems.punchcut.com+1
• Lazarev.agency : Known explicitly for deep B2B SaaS product redesigns and emerging generative AI UX approaches, helping companies transition complex analytical dashboards into conversational and AI-augmented workflows.lazarev.agency
• Momentum Design Lab : Takes a structured, enterprise-friendly approach to AI integration, making them a great fit if your SaaS serves enterprise clients who have strict security, explainability, and trust requirements.lazarev.agency
• Bricx : Particularly strong for B2B and technical AI products where the core design hurdle is trust—helping users understand, audit, and confidently act on probabilistic AI outputs rather than blindly trusting or rejecting them.agency.uxplanet.org

Top-Tier Product & Tech Agencies

• MetaLab : Famed for shaping clean, intuitive workflows for category-defining tech products (like early Slack and Coinbase). They lean heavily into a co-founder mentality, fast product execution, and modern AI interface paradigms.
• Work & Co : Best suited if you are an established enterprise or large-scale platform needing a complete architectural and UX overhaul to embed AI natively into an existing massive digital ecosystem.
• frog design / Designit / IDEO : The legacy innovation and human-centered design powerhouses. They excel at deep user research to figure out how your specific SaaS users actually want an AI copilot to assist them (vs. just dropping a generic chat window on the side of your app).punchcut.com+1

To help narrow down which of these paths makes sense for you, tell me:

• What is the stage of your SaaS (early startup vs. scaling/enterprise)?
• Is your copilot primarily a side-panel chat assistant , or integrated as Generative UI (inline commands, auto-populating fields, action-oriented triggers)?

Yes. For a SaaS copilot, I’d look specifically for people who have shipped human–AI interaction, conversational UX, agentic workflows, and AI-specific interaction patterns, rather than a generalist UI/UX designer.

A few places/people worth looking at:

• Siân Lindley — Microsoft Research — specializes in human-centered AI, generative UX, and how AI systems can preserve user agency while taking on more autonomy. Her work is particularly relevant if your copilot will evolve beyond simple chat.Microsoft
• Saleema Amershi — Microsoft — one of the authors of Microsoft's influential Guidelines for Human-AI Interaction, covering things like communicating AI capabilities, handling uncertainty, correction, explanations, memory, feedback, and user control.Microsoft+1
• Nielsen Norman Group (NN/g) — has an explicit Artificial Intelligence UX specialty, including courses on designing AI experiences, AI product strategy, and generative-AI-assisted UX.Nielsen Norman Group
• OmniSenseLab — an independent consultancy specifically advertising AI copilot UX, conversational interfaces, LLM applications, and SaaS AI product design.OmniSenseLab
• Groto — focuses on AI product UX, including copilot/agent UX, multi-turn conversation design, prompt surfaces, explainability, and controls such as edit/regenerate/accept/reject.Groto
• Reload UX — focuses on AI-native SaaS experiences, including agentic workflows, conversational AI, AI trust/explainability, and AI design systems.ReloadUX
• Troxis — specifically markets GenAI UX and copilot UX for SaaS platforms, including RAG interaction patterns and agent UX.Troxis

What I'd look for in your designer

For a SaaS copilot, the important expertise isn't simply "can they design a nice chat UI?" I'd screen for experience with:

• Copilot vs. chatbot interaction models
• Designing when the AI should suggest, ask, act, or stay silent
• Streaming and progressive responses
• Handling uncertainty, hallucinations, and failed actions
• Human-in-the-loop workflows
• Agentic actions with confirmation/undo/review
• Context and conversation memory
• Making AI output editable rather than authoritative
• Structured AI outputs—tables, forms, cards, charts, actions—not just text
• Trust, provenance, citations, and explainability
• Designing feedback loops so the product learns what users want
• AI-specific UX research and usability testing

Microsoft's current guidance for custom copilots is actually a useful benchmark: it explicitly discusses input/output design, collaborative UX, and keeping the human in control.GitHub+1

If you're hiring rather than looking for an agency, I can also identify 10–15 individual designers/UX researchers who specialize in generative-AI interfaces and have actual shipped-product experience, with their portfolios, current roles, and areas of expertise.

Yes. For a SaaS copilot, I’d look specifically for AI product/UX designers with experience in human–AI interaction, conversational UX, agentic workflows, and enterprise SaaS—not just a general UI/UX agency. The hard problems are things like when the copilot appears, how it handles uncertainty, showing sources/context, approval flows, streaming/tool states, and keeping the user in control.Microsoft Design+1

A few specialists worth investigating:

• OmniSenseLab — Particularly relevant to your use case: they explicitly specialize in AI copilots, LLM applications, conversational interfaces, and integrating AI into existing SaaS products.OmniSenseLab
• Groto — Focuses on copilot UX, agentic workflows, explainability, human-in-the-loop interactions, and enterprise AI. They specifically cite work on enterprise copilots.Groto
• Troxis — Strong match if your copilot involves workflow automation or agents; their offering includes GenAI UX, copilot UX, conversational interfaces, RAG UI, and agent UX patterns.Troxis
• Fluidesigns — Their work specifically addresses the differences between conventional SaaS UX and LLM interfaces: streaming, multi-turn context, tool use, uncertainty, and error states. They also work on adding AI layers to existing SaaS products.Fluidesigns
• Reload UX — Focuses on AI-native product design, agentic workflow UX, conversational AI, AI trust/explainability, and redesigning existing products around AI features.Reload UX
• Joshua Wells — An independent product/UX architect whose current work explicitly includes generative UIs, agentic UX patterns, evaluation-driven design, and building working AI prototypes.Joshua Wells

If I were hiring for your specific project

I'd prioritize someone who can show actual shipped copilot/AI SaaS work, rather than someone whose portfolio simply says "AI UX."

I'd ask each candidate to walk through how they'd design:

1. Invocation — when should the copilot proactively appear vs. wait?
2. Context — what does it know about the current page, customer, records, and task?
3. Generation — how do streaming, partial results, citations, and uncertainty appear?
4. Actions — how does a user review/approve an AI action before it changes data?
5. Errors — what happens when the model misunderstands or can't complete a task?
6. Trust — how does the interface communicate why the AI produced something?
7. Feedback/evaluation — how do you learn which AI interactions are actually useful?
8. Integration — how does the copilot feel native to your existing SaaS rather than like a ChatGPT window bolted onto it?

Microsoft's own Copilot design work is useful here because it illustrates the shift from designing a static interface toward designing an adaptive system in which the outputs and interaction model themselves become part of the experience.Microsoft Design+1

If you tell me what your SaaS product does, who the users are, and roughly what the copilot will do, I can narrow this down to 5–10 specific designers/agencies that have the closest relevant experience, including portfolios and contact links.

For a SaaS product adding a copilot, you want someone who understands more than chat UI. The strongest designers in this space specialize in AI product UX, human-in-the-loop workflows, agentic experiences, LLM interaction patterns, and trust/transparency design.AI UX Design Agencies+1

A few people and teams worth looking at:

Individual AI UX designers

• WE THINK AI - AI led Product and Platform Engineering Company — Known for human-centered AI, Copilot experiences, multimodal interactions, and agent workflows. A strong fit if you need someone who has worked on large-scale AI assistant experiences.Parse
• Denys Babenko | Product Designer | UX/AI — Focuses on Copilot experiences, AI design systems, and generative UI patterns.Parse
• Ena Yun — Experience with AI design systems and generative UI patterns across products such as Copilot-related experiences.Parse
• manishkumarjain.com — Focuses on AI copilots, agent workflows, conversational UX, and enterprise SaaS patterns.Parse

Specialized AI UX studios / consultancies

• OmniSenseLab — Specializes in LLM products, AI copilots, conversational interfaces, and SaaS companies integrating AI features. This is closer to what many SaaS teams need when they have an existing product and are embedding AI.OmniSenseLab
• Punchcut — A good fit for strategic AI UX, emerging interfaces, and complex intelligent products.Parse
• Parallel — Better suited for B2B SaaS teams that need broader product strategy, research, and design-system work alongside AI UX.Gev Design
• Troxis — Works on generative AI UX, AI copilots, conversational interfaces, and agent UX patterns for SaaS and enterprise products.Troxis

What I would screen for when hiring

A lot of designers now say they do "AI UX." I would look specifically for case studies showing they have designed:

1. AI embedded inside an existing workflow

• Example: AI helping a CRM user write follow-ups, summarize accounts, or prioritize leads.
• Less valuable: a standalone ChatGPT-style chat window.

2. Human control mechanisms

• Approve/edit/reject flows
• Source citations or grounding
• Confidence indicators
• Undo/version history

3. Non-deterministic UX

• Loading and streaming states
• Partial results
• Errors and hallucinations
• "The AI is unsure" moments

4. Agent interaction patterns

• Task planning
• Multi-step execution
• Permissions
• Confirmation checkpoints

5. AI design systems

• Reusable patterns for prompts, suggestions, generated content, AI actions, and feedback loops

For a typical SaaS company building its first serious copilot, I would usually start with:

• a senior AI product designer for strategy + interaction model,
• then a product design team for execution,
• rather than hiring a traditional UI agency that has only added a chatbot.

If you share your SaaS category (CRM, analytics, HR, finance, dev tools, etc.), whether the copilot is chat-based vs. embedded into workflows, and your stage (startup, growth, enterprise), I can narrow this to the best-fit designers/agencies.

For a SaaS copilot, you’re looking for a relatively new specialty: AI product design / human–AI interaction design, not just a general UX/UI designer. The strongest people in this space understand that generative AI UX is less about drawing chat screens and more about designing trust, uncertainty, workflow integration, feedback loops, agent behavior, and human-in-the-loop controls.Microsoft+1

A few people and firms worth looking at:

Individual AI product designers

• Han-Shen Chen — Known for work around human-centered AI, Copilot-style experiences, multimodal interaction, and AI workflows. A strong fit if you want someone deeply focused on next-generation AI interaction patterns.Parse
• Maninder Kaur — Focuses on AI copilots, agentic workflows, conversational UX, and enterprise SaaS experiences.Maninder Kaur
• Tashfeen Ahmed — Works on human-AI interfaces for Microsoft 365 Copilot, including writing assistance and agent experiences.Tashfeen Ahmed Portfolio

Design studios / consultancies

• OmniSenseLab — Specializes specifically in LLM products, AI copilots, conversational interfaces, and SaaS teams integrating generative AI. This is close to the profile many SaaS companies need when adding an AI layer to an existing product.OmniSenseLab
• Punchcut — Better suited if your copilot involves complex workflows, multimodal experiences, or future-facing interaction models.Parse
• Parallel — A good fit if the copilot is part of a broader B2B SaaS product redesign with research, systems thinking, and product strategy.Gev Design

What I would screen for

When interviewing candidates, I’d look for shipped examples of:

• Embedded AI workflows

• AI inside the existing product workflow, not a separate “ChatGPT tab”

• Example: “help me complete this task” rather than “ask me anything”

• Human control patterns

• Edit before applying

• Approve/reject actions

• Show sources or reasoning context

• Undo AI actions

• Non-deterministic UX

• Handling wrong answers

• Partial completion

• Long-running agent tasks

• Confidence and uncertainty states

• AI design systems

• Prompt patterns

• Suggested actions

• Generated content blocks

• Agent status indicators

• Feedback mechanisms

• Product judgment

For a typical SaaS company building its first serious copilot, I’d usually start with either:

1. a senior AI product designer embedded with your team for 4–12 weeks, or
2. a small AI UX consultancy to establish patterns and a design system before scaling.

If you share your SaaS category (CRM, analytics, HR, developer tools, finance, etc.), whether the copilot is chat-based vs. agentic, and your company stage, I can narrow this to a shorter shortlist.

Building a SaaS copilot requires a distinct shift from traditional UI/UX. You aren't just designing static buttons or linear forms anymore; you're designing for probabilistic outputs, managing latency, setting user trust boundaries, and figuring out when a copilot should be proactive versus passive.youtube.com+1

Several specialized design agencies, boutique studios, and consultancies have carved out a reputation for nailing generative AI workflows and SaaS copilots:

Top-Tier Product Agencies Specializing in AI & Copilots

• MetaLab : Known for high-profile zero-to-one product execution and defining intuitive interfaces (such as early work with Midjourney). They excel at taking complex, raw AI capabilities and wrapping them in sleek, consumer-grade SaaS UX.linkedin.com+2
• Humbleteam : Frequently cited for emerging tech and AI product integration, specializing in designing conversational consumer apps, enterprise copilots, and multi-modal AI workflows.humbleteam.com+1
• Clay Global : A San Francisco-based agency that merges enterprise-scale SaaS design with advanced AI-driven functionality, focusing heavily on human-centric AI interactions and complex digital transformations.linkupst.com+2
• Work & Co : A massive player for large-scale digital products. They specialize in integrating complex AI search, contextual sidebars, and automation tools into existing enterprise and SaaS ecosystems.humbleteam.com+1

Specialized SaaS & AI Boutiques

• Eleken : Operates on a dedicated, subscription-like design model geared specifically toward SaaS companies and AI startups that need continuous, agile design iteration rather than a one-off handoff.
• Cieden : Highly focused on complex digital products and embedding machine learning or generative AI features into existing SaaS workflows without breaking user mental models.
• Neuron : Focuses heavily on B2B enterprise software and data-heavy dashboards. If your copilot needs to parse dense, messy data tables and present actionable insights cleanly, they are a strong fit.
• Lazarev. Agency : Noted for building AI-driven SaaS products from scratch, putting heavy emphasis on user validation, proof-of-concept prototyping, and intuitive agentic interfaces.humbleteam.com+3

Key UX Challenges to Discuss with Them

When evaluating these partners for your copilot, make sure to probe on how they handle GenAI-specific patterns:

1. State & Latency Management: How do they design placeholders, streaming text, and graceful failures when the LLM stalls or hallucinates?
2. Context & Modality: Are they designing an intrusive chat modal, an inline ghost-text completer, or a contextual side-panel that reacts to the user's current cursor/workspace state?
3. Feedback Loops: How do they capture explicit (thumbs up/down) and implicit (user editing the generated text) feedback to refine the UX?youtube.com+1

If you want to narrow this down, tell me:

• What is your SaaS vertical (e.g., fintech, CRM, developer tools, HR)?
• Are you looking for a full zero-to-one redesign or an embedded design team to slot into your current sprint?