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

Yes. For a SaaS copilot, I’d look specifically for people who have designed LLM/copilot interaction patterns, not just general SaaS UI. The UX problems are different: uncertain outputs, streaming, context, tool use, human approval, error recovery, trust, and deciding when the AI should act versus ask.

People / firms I’d shortlist

• OmniSenseLab — Probably the most directly aligned with your brief. They explicitly specialize in AI copilots, conversational interfaces, LLM products, and SaaS AI UX, including UX research and strategy. Led by Saikat Mandal, a former Google/Kaiser Permanente/NetApp design leader.
• Punchcut — A strong choice if you want a more established human-centered AI design firm. They have deep experience in intelligent products, AI agents, multimodal interfaces, and autonomous systems.
• IDEO AI Lab — Best if the project involves substantial product strategy + interaction design, rather than simply designing the copilot UI. Their AI practice focuses explicitly on human-centered AI and integrating AI into existing products.
• Groto — Worth looking at specifically for enterprise/SaaS copilots. They emphasize conversational UX, explainability, human-in-the-loop design, onboarding, prompt UX, and feedback loops, and cite copilot work for Colgate and AB InBev.
• Troxis — Explicitly offers GenAI UX, AI copilot UX for SaaS, conversational/prompt design, agent UX, and RAG interaction patterns.
• Nielsen Norman Group (NN/g) — I'd consider them if you already have designers but want an expert UX evaluation/research engagement rather than handing the entire product to an agency. They do customized UX research, expert reviews, usability testing, and consulting.

My recommendation for your situation

If you're building an existing SaaS product and adding a copilot, I'd prioritize:

1. OmniSenseLab → 2. Groto → 3. Troxis → 4. Punchcut

I'd ask each candidate to show 2–3 shipped examples where they designed an AI assistant inside an existing workflow, rather than standalone ChatGPT-like products.

And in the interview, I'd ask them to explain how they'd handle these five things:

1. When the copilot should proactively act vs. wait for a request
2. How users understand what context/data the AI is using
3. How uncertainty and hallucinations are communicated
4. How users review, edit, approve, or undo AI actions
5. How the copilot becomes useful without turning the SaaS UI into a giant chat box

That last point is particularly important. A good SaaS copilot often shouldn't look like ChatGPT embedded in the sidebar; the best UX can be task-specific controls, suggestions, generated UI, contextual actions, and human-in-the-loop workflows.

If you tell me what your SaaS product does and roughly what the copilot will do, I can narrow this to 5–10 specific designers/agencies with relevant portfolios and case studies, including who I'd contact first.

Building a SaaS copilot requires specialized UX patterns—handling latency, managing probabilistic outputs (hallucinations/errors), designing conversational vs. inline UI, and establishing user trust.

Top-tier agencies and consultancies specializing in UX design for generative AI interfaces and B2B SaaS products include:

• Lazarev. agency : Known specifically as an AI UX Design Agency that focuses heavily on adaptive, anticipatory AI-driven interfaces, complex decision flows, and reducing cognitive load for tech products.
• Clay : A San Francisco-based digital product agency widely recognized for craft-led, high-end UI/UX design for tech innovators, consumer apps, and modern AI-powered platforms.
• Neuron : Specializes in B2B workplace tech, complex SaaS workflows, and infusing applied AI seamlessly into enterprise environments.
• Punchcut : An experience design agency with a long track record in emerging technologies, spatial computing, and conversational or adaptive AI interfaces for major tech companies.
• Goji Labs : Focuses on end-to-end product strategy, UX research, and digital product development for fast-moving startups and enterprises building custom AI tools.
• Toptal : If you aren’t looking for a full agency retainer but rather need to embed a vetted, senior freelance AI UX Designer directly onto your internal product team quickly.

If you want to narrow this down, let me know:

• What is your current stage (early prototype vs. scaling an existing SaaS)?
• Do you prefer an embedded freelance specialist or a full-service design agency?

For a SaaS copilot, I would look for human–AI interaction designers rather than general UX designers who have only added a chat window to a product. The strongest specialists understand things like:

• When the AI should suggest vs. take action
• How to represent uncertainty and confidence
• How users review, edit, undo, and override AI work
• How to design streaming responses, tool calls, agent workflows, and failures
• How to integrate AI into existing SaaS workflows instead of creating a separate chatbot experience

A few people and teams worth looking at:

Individual designers

• Han-Shen Chen — Strong fit if you want someone deeply focused on human-centered AI, Copilot experiences, multimodal interactions, and agent workflows. He has worked on generative AI and Copilot products at Microsoft and previously led UX work involving Google Assistant and AI systems.
• Menghui Hu — Has worked on Copilot experiences, AI design systems, generative UI patterns, and scaling AI product design.
• Ena Yun — Focuses on AI design systems and generative UI experiences across Copilot, Edge, and Bing.
• Maninder Kaur — Focuses specifically on AI copilots, agentic workflows, conversational UX, and enterprise SaaS products.

Design consultancies / studios

• OmniSenseLab — A boutique consultancy focused specifically on LLM products, AI copilots, conversational interfaces, and SaaS AI integrations. This is closer to the profile many SaaS companies need when they already have a product and are adding AI.
• Punchcut — A good option for strategic AI UX, emerging interfaces, and complex intelligent products.
• Parallel — Worth considering if your copilot is part of a larger B2B SaaS redesign effort requiring research, systems thinking, and product strategy.
• auxfirst — Better suited if your copilot is becoming an autonomous agent that executes tasks rather than just answers questions.

What I would look for in a portfolio

I would prioritize case studies showing:

1. AI embedded into an existing workflow

• Example: “AI helps a CRM user close deals faster”
• Not just “we built a ChatGPT clone”

2. Human-in-the-loop controls

• Approve/reject
• Edit before sending
• Explain why
• Show sources/context

3. Non-deterministic UX

• Empty states
• Uncertain answers
• Long-running tasks
• Partial completion
• Recovery flows

4. AI design systems

• Patterns for prompts
• Suggestions
• Generated content blocks
• Agent status
• Feedback loops

5. Strong product thinking

• They should challenge whether a feature needs AI at all.

For a typical SaaS company building its first serious copilot, I would probably start conversations with OmniSenseLab, Parallel, Punchcut, and one senior AI product designer (such as Han-Shen Chen, Menghui Hu, or Maninder Kaur) depending on your budget and whether you need a strategist, embedded designer, or full delivery team.

If you share what your SaaS does (CRM, analytics, HR, finance, dev tools, etc.), the copilot’s intended role, and whether you need a freelancer vs. agency, I can narrow this to a more targeted shortlist.

Designing a "copilot" or generative AI feature for an existing SaaS product is rarely an engineering bottleneck—it’s a trust, workflow, and interface challenge . When users open an AI copilot, poor UX isn't just an annoyance; it causes a lack of trust that leads to churn.

Specialized agencies and studios focus specifically on navigating generative UI, conversational patterns, prompt-and-response ergonomics, and grounding probabilistic outputs into deterministic software workflows:

Specialized AI & Product Design Agencies

• 925Studios : Highly regarded for early-to-growth-stage AI startups and SaaS products building native LLM features. They specialize in translating complex multi-tenant data handling, agent behaviors, and real-time generative interfaces into clean, retention-driving UX.
• The Skins Factory : A veteran UI/UX studio with a massive footprint in enterprise SaaS, fintech, and complex applications. They publish actionable frameworks specifically on Designing AI Copilot UX , focusing heavily on error states, human-in-the-loop controls, and building predictable interaction paradigms for unpredictable AI.
• Clay : A premier digital product agency in San Francisco known for high-end execution with major tech brands and enterprises. They excel at crafting polished, bleeding-edge AI interfaces where brand, fluid motion, and human-centric interaction design collide.
• Lazarev.agency : Known heavily for B2B SaaS modernization and AI-driven product redesigns. They use a deep user-research approach to embed AI smoothly into existing user habits rather than awkwardly tacking on a generic chat sidebar.
• Punchcut : Ideal if your copilot extends toward advanced multimodal, spatial, or autonomous system interactions. They specialize in next-gen interface paradigms where standard screens aren't enough.
• Cieden : A B2B-focused UX agency that designs complex workflow applications for SaaS, edtech, and healthcare, specializing in making dense enterprise data interactions intuitive when augmented by AI.

Key UX Patterns These Specialists Solve for Copilots

When vetting or talking to design partners, look for teams that address these specific generative UI realities:

1. Non-deterministic Graceful Failures: Designing what happens when the LLM hallucinates or times out without breaking user trust.
2. Inline vs. Sidebar Paradigms: Knowing when a conversational chat bubble is appropriate versus when an inline "ghost text" autocomplete or generative canvas card is better suited to the workflow.
3. Human-in-the-Loop Controls: Ensuring users can easily edit, reject, or fine-tune generative outputs so they always feel in control rather than dictated to by the AI.

If you'd like to narrow this down, tell me:

• What is your SaaS product's core vertical (e.g., CRM, project management, developer tools)?
• Are you looking for an agency partner for a full redesign, or fractional/consultative UX guidance?

I can tailor recommendations based on your timeline and budget model.

For a SaaS product adding a generative AI copilot, you generally want someone who understands more than visual UI. The strongest specialists tend to come from AI product design, conversational UX, human-AI interaction (HCI), and enterprise workflow design. The hard parts are usually:

• deciding when the AI should appear vs. stay silent
• designing trust signals (sources, confidence, explanations)
• handling uncertainty and errors
• creating approval/review flows
• making AI feel embedded in the product rather than “a chatbot bolted on”
• designing prompt, context, memory, and agent behaviors

Here are some specialists and firms worth evaluating:

AI UX / copilot design specialists

Groto — focuses specifically on AI-first UX, copilots, conversational interfaces, agentic workflows, and explainable AI patterns. They position themselves around enterprise AI products and SaaS teams.

Boneyard Collective — specializes in AI product design, including agent surfaces, copilots, trust/transparency UI, streaming states, and human-in-the-loop workflows.

OmniSenseLab — an AI UX consultancy focused on LLM applications, AI copilots, conversational interfaces, and SaaS AI integrations.

Troxis — works on GenAI UX, copilot UX, conversational interfaces, prompt design, and agent UX patterns.

The Skins Factory — focuses on copilots, adaptive interfaces, and trust-critical AI experiences.

Product studios with AI-native experience

Pory — builds AI-native products and explores conversational interfaces, autonomous agents, and vertical AI copilots.

Individual researchers / thought leaders to look at

• Ben Lafreniere — researcher focused on human-AI interaction, intelligent interfaces, and generative AI UI research.
• Elizabeth Rodwell — studies conversational AI, UX practice, and how teams design AI systems.

If you are hiring, look for these portfolio signals

A good candidate should have shipped examples of:

1. AI copilots inside existing SaaS workflows

   • CRM assistant
   • analytics copilot
   • customer support agent
   • developer assistant
   • operations automation

2. Non-chat AI patterns

   • inline suggestions
   • AI-generated drafts
   • smart actions
   • agent task execution
   • review queues

3. Trust and control mechanisms

   • citations/sources
   • “why did AI do this?”
   • undo/revert
   • human approval steps
   • confidence indicators

4. AI-specific UX research

   • testing whether users understand AI behavior
   • measuring trust and adoption
   • designing around model limitations

For a B2B SaaS copilot, I would prioritize a senior AI product designer or small AI UX studio over a general UX agency. The discipline is still emerging, and many traditional product designers have strong UI skills but limited experience designing probabilistic systems.

If you share your SaaS category (CRM, fintech, HR, analytics, dev tools, etc.) and whether you need strategy, Figma/design, research, or full product implementation, I can narrow this to a shortlist.

Several specialized digital agencies and design firms focus explicitly on UX/UI design and strategy for generative AI interfaces and SaaS copilots.

Top Agencies Specializing in AI Copilot UX

• The Skins Factory — Specializes in enterprise SaaS AI copilots, focusing on placement, input design, and moving assistance inline where work actually happens rather than relying on generic chat sidebars.
• Lazarev — A San Francisco-based product design agency specializing in AI-fueled innovation, complex data abstraction, and conversational AI/copilot experiences for SaaS and fintech companies.
• Arounda — Product design agency that builds AI analytics platforms and generative AI interfaces focused on progressive disclosure, transparent data hierarchies, and building trust in recommendations.
• Pixel Web Solutions — Offers dedicated AI copilot development and interface design services focused on embedding assistants directly inside existing software workflows.

Key UX Considerations for SaaS Copilots

If you'd like, let me know:

• What core workflow your SaaS product serves
• Whether you envision the copilot as a side panel, inline text, or full-screen canvas

I can give you more tailored architectural recommendations or help you evaluate design partners.

Building an AI copilot for SaaS requires a specialized UX approach. Unlike traditional interfaces where actions are deterministic (click a button, get a predictable state change), generative AI interfaces deal with probabilistic outputs, latency, token streaming, context windows, and error handling when hallucinations happen.

Several specialized digital product design agencies and boutique studios focus explicitly on AI copilots, complex B2B SaaS workflows, and intelligent decision engines:

• Lazarev.agency is widely recognized for designing AI copilots, decision engines, and data-heavy B2B SaaS platforms. They specialize in translating complex model outputs and multi-layered data into clear, actionable interfaces for non-technical users.
• Punchcut brings over two decades of experience designing intelligent systems and emerging tech. They feature a dedicated AI product practice focusing on conversational and adaptive UI patterns, intent streaming, and context-aware design.
• Humbleteam (with a global footprint across major tech hubs) has a strong track record of building and scaling AI-native products from early-stage to high-growth enterprise adoption, specifically focusing on integrating AI agents naturally into workflows.
• Bakken & Bæck is an Oslo-based studio known for deep technical fluency. Because they maintain in-house machine learning capabilities, they understand the technical constraints of LLMs, making them a great fit for advanced conversational or autonomous copilot setups.
• MetaLab is a premier product design powerhouse famous for interface craft (they designed early versions of Slack). They have increasingly moved into high-end AI product design and rebranding for cutting-edge AI scale-ups, focusing heavily on product feel and immediate trust.
• UX Studio based in Budapest, is ideal if your copilot requires rigorous, research-led product discovery. They excel at taking complex B2B SaaS architecture and ensuring that adding a copilot reduces user friction rather than cluttering the screen.

If you'd like to narrow down which agency fits best, tell me:

• Is your copilot side-panel/chat-based, inline/contextual , or fully autonomous?
• Are you looking for a full-service agency or an embedded designer to augment your current product team?

Yes. For a SaaS copilot, I’d look specifically for designers/agencies with experience in AI-native product UX, rather than a general UI/UX shop. The hard part isn’t making a chat window look good—it’s designing trust, uncertainty, suggested actions, tool calls, human handoffs, editable outputs, and how AI fits into an existing workflow.

A few specialists worth evaluating:

1. Groto — Probably one of the closest matches to what you describe. They explicitly specialize in copilot UX, conversational interfaces, agentic workflows, explainability, guardrails, and human-in-the-loop design. They cite copilot work for Colgate and AB InBev.

2. Fluidesigns — Particularly interesting if your copilot is being embedded into an existing B2B SaaS product. They focus on non-deterministic outputs, streaming, multi-turn interactions, confidence/uncertainty, AI onboarding, and integrating AI into existing workflows rather than simply adding a chatbot.

3. OmniSenseLab — A smaller, specialized consultancy focused explicitly on AI copilots, LLM applications, conversational interfaces, SaaS AI, and UX research for AI systems. This could be a good fit if you want senior strategic UX rather than a large agency.

4. AUX First — More forward-looking/agentic. They specialize in generative/adaptive UI, intent-driven UX, proactive interactions, trust/explainability, hallucination mitigation, and human escape hatches. Particularly worth looking at if your "copilot" is evolving toward an agent that actually takes actions.

5. Chrono Innovation — Strong emphasis on the stuff many conventional UX teams miss: agent states, failures, human review, AI-powered SaaS integration, variable outputs, and workflow-first rather than chat-first design.

What I'd look for in your case

If you're building a copilot inside an existing SaaS product, I'd prioritize someone who can answer questions like:

• When should the copilot suggest versus act?
• How does the user understand why the AI made a recommendation?
• How do we represent uncertainty and hallucination risk without making the product annoying?
• How does the user correct, edit, undo, or override AI actions?
• When should the UI be chat, and when should it be buttons, forms, structured results, or inline suggestions?
• How do you design streaming, partial results, tool calls, errors, retries, and long-running tasks?
• How does the copilot learn the user's context without becoming unpredictable?
• What does the experience look like when the AI can't complete a task?
• How do we measure whether the copilot actually improves the underlying workflow?

That last point is important: I would not hire someone solely because their portfolio says "AI/ChatGPT." A lot of AI design work is essentially conventional SaaS UI with a chat box added. The better specialists explicitly design around the behavior of probabilistic systems. Fluidesigns and Chrono, for example, explicitly talk about designing for variable/non-deterministic outputs and AI-specific states.

If you tell me what your SaaS product does, who the users are, and what you want the copilot to actually do, I can narrow this to 5–10 designers/consultancies that are particularly well matched, including their relevant case studies, approximate engagement model, and which I'd interview first.

Yes. If you’re building a copilot inside an existing SaaS product, I’d look specifically for designers who understand human–AI interaction, not just “AI UI” or chatbot aesthetics.

A few firms I’d put on the shortlist:

1. Punchcut — Probably the strongest choice if you want deep expertise in human-centered AI, intelligent products, agents, and multimodal interfaces. They’re more of a mature strategic/design firm.
2. ParallelHQ — Particularly interesting for a B2B SaaS team. They focus on AI-native product design, human-AI interaction patterns, onboarding, and embedding AI into real workflows rather than simply putting a chat box into the product.
3. Fluidesigns — Explicitly works on AI copilots and AI features inside existing SaaS products, including model-output presentation, uncertainty/error states, onboarding, and AI interaction patterns. This sounds quite close to your use case.
4. OmniSenseLab — A smaller AI-product UX consultancy specifically offering copilot experience design, conversational UX, SaaS AI integration, and UX research around trust and user behavior.
5. auxfirst — Worth considering if your “copilot” is evolving toward an agent that takes actions. They specifically work on generative/adaptive UI, intent-driven UX, human-in-the-loop controls, explainability, and agent workflows.
6. Groto — More focused on AI-first UX and copilot/conversational experiences, with particular emphasis on trust, guardrails, and handoffs between AI and users.

Who I'd talk to first

For a SaaS copilot, my initial three would be:

ParallelHQ → Fluidesigns → Punchcut

The reason is that your hard problem probably isn't “How should the chat interface look?” It's questions like:

• When should the copilot proactively appear vs. stay out of the way?
• Should users ask, command, or select an outcome?
• How does the user inspect/edit an AI-generated result?
• How do you expose uncertainty without constantly saying “AI may be wrong”?
• How does the copilot interact with the existing navigation and workflows?
• What happens when the AI needs clarification?
• When should AI perform an action versus merely recommend one?
• How do you preserve user agency and undoability?
• What does the experience look like when the AI is asynchronous or taking multiple steps?

Those are genuinely different UX problems from conventional SaaS. Research on GenAI UX similarly emphasizes agency, trust, uncertainty, feedback, and the unusual interaction characteristics of probabilistic systems.

One important hiring filter: ask candidates to show you actual shipped AI workflows, not just beautiful Figma screens. I'd specifically ask them to walk through the happy path, wrong-answer path, ambiguity path, failure path, and human-override path of something they've designed.

If you tell me what your SaaS product does, who the users are, and what you want the copilot to actually do, I can narrow this to 5–10 specific designers/agencies that are unusually well matched, including likely engagement model and approximate price tier.

Building a UX for a generative AI copilot inside a SaaS product is radically different from traditional interface design. You aren't just designing static screens; you are designing for non-deterministic outputs, latency states, error handling, inline editing, and the shift from command-based workflows to conversational or agentic co-pilot panels.

A specialized group of digital product agencies and boutique studios focus explicitly on UX design for generative AI, copilots, and complex SaaS decision engines:

• Lazarev. agency , Best for data-heavy AI SaaS and complex decision engines . They specialize in translating raw LLM/ML data outputs into actionable, clean, non-technical interfaces.
• Punchcut , Best for intelligent, multimodal systems and autonomous assistant UX . They have a long-standing reputation for spatial, conversational, and emerging tech interaction models.
• Humbleteam , Best for venture-backed AI startups . They have designed high-profile generative AI and machine learning products from seed to scale (such as Synthesia and enterprise fleet AI).
• MetaLab , Best for high-end consumer-grade polish and product feel . Known for their interface pedigree (like Slack), they excel at making sophisticated AI workflows feel intuitive and trustworthy.
• Bakken & Bæck , Best for technically fluent, research-led teams . They actually house machine learning and technical capabilities in-house alongside design, with deep experience in conversational AI (e.g., Sierra).
• Clay , Best for premium brand-to-product execution . They build polished enterprise-grade AI interfaces and robust design systems for major tech players.

If you'd like to narrow this down, tell me:

• What is your SaaS vertical (fintech, HR, developer tools, etc.)?
• Are you leaning toward a full-service enterprise agency or a lean, fast-moving boutique studio?

I can help recommend the exact fit for your timeline and stage.