The short answer
Five Bay Area firms worth shortlisting for AI product research are Akraya, Punchcut, EchoUser, Neuron, and Portigal Consulting, and each one fits a different need. Akraya fits enterprise AI teams that need UX research and design delivered continuously across an AI roadmap, specialists in how people work with AI, and hard-to-reach participants. Punchcut fits AI design for devices, voice, and multimodal experiences, EchoUser fits teams that want research, design, and testing from one boutique agency, Neuron fits enterprise and B2B design work, and Portigal Consulting fits teams building their own research practice.
Choosing a research partner for an AI product is harder than choosing one for a website or an app, because AI features change from one session to the next and can be confidently wrong.
The questions that matter also go well beyond usability. Your team needs to know if people trust the output, if they catch its mistakes, and if they keep using it after the first bad answer.
Most UX research firms can run a usability test, but far fewer are set up for the questions AI products raise, the participants those studies need, or the pace AI teams ship at. This guide is for UX research managers and leads comparing partners for that work.
It covers five Bay Area firms worth considering, what each one is best for, how to choose between them, and the questions and red flags that separate a strong AI research partner from a general one.
A note on this guide: Akraya wrote it, and we're on the list. We've described every firm, including ours, by what it offers and who it fits best, using information from each firm's own website, so you can decide which one fits your team.
Here's the short version, and each firm is covered in more detail further down.
Bay Area UX research firms for AI teams, compared
| Firm | Based in | Main focus | Best for |
|---|---|---|---|
| Akraya | San Francisco, running research in the US, UK, Romania, India, Japan, and Australia | UX research and design for AI products, delivered as rapid studies and embedded researchers under one managed team | Enterprise AI teams that need research and design delivered across an AI roadmap, AI specialists, and hard-to-reach participants |
| Punchcut | San Francisco | Design and innovation, with a user research team | AI design for devices, voice, and multimodal experiences |
| EchoUser | San Francisco | UX research, design, and testing | Complex products that need research and design together |
| Portigal Consulting | Bay Area | Research consulting, training, and coaching | Building your own team's research practice |
| Neuron | San Francisco | UX and UI design for enterprise and B2B software, with usability testing | Enterprise and B2B design work, including AI interfaces |
Before comparing firms, it helps to be clear about what AI research needs to answer, because that's what separates the partners on this list.
A usability test tells you if people can find and use a feature, but AI products raise a second set of questions that decide if people keep using the feature at all.
We break those questions down into six factors in what decides if people keep using your AI product, and a strong AI research partner should be able to answer all six:
01
Trust: do people believe the AI's output enough to act on it?
Users who trust it too little redo the work by hand, and users who trust it too much stop checking it. A strong partner measures trust and reliance with behavioral methods, beyond asking people how they feel.
02
Understanding: can users tell how it works, and when it's wrong?
Studies should include scenarios where the AI gives a wrong answer, so you learn if users catch it and which signals on screen help them.
03
Control: can users correct or override it when it slips?
Users need a way to fix one part of an answer without starting over, so correction flows need their own testing.
04
Reliability: does it hold up across real users and sessions?
An AI feature can work for the pilot team and still fail for users in another market. Catching that takes recruiting across languages, cultures, and edge cases.
05
Predictability: do users know what it will do the next time?
If the same request gets a different kind of answer each time, people can't build a routine around it. Longitudinal studies show where that variation wears down their confidence.
06
Adoption: will they keep using it, or drop it after one bad answer?
Adoption is where the other five factors show up. Research that follows users over time shows who drops off, when, and which factor caused it.
If your AI takes actions on a user's behalf, add one more question to the list: can users rely on those actions, including the ones that are hard to undo?
With those questions in mind, here's how each firm approaches the work.
Each entry covers where the firm is based, what it offers, who it fits best, and what's worth asking about on a first call.
No. 1
Where we're based: San Francisco, and we run research with local teams in the US, UK, Romania, India, Japan, and Australia.
What we offer: UX research and UX design for enterprise product teams, with deep specialization in AI. Our work on AI products covers:
Akraya also offers AI and data services and AI-led product engineering, so the same partner can help turn research findings into model evaluation and product work.
It's all delivered as UX research as a service, in two ways under one managed team. Rapid research answers a focused question with findings in 2 to 3 weeks, and an embedded researcher works inside your product team for the length of your program, managed by Akraya's PMO. You can run both at the same time.
Best for: Enterprise AI teams that need research and design delivered continuously across an AI roadmap, by a partner that owns the work from scoping to findings. Learn more about our UX research for AI products.
Here's how we handle the questions every AI team should ask a research partner:
01
Who does the research
PhD-level researchers vetted against a big tech hiring bar, matched to the problem instead of assigned by availability. AI Interaction Engineers lead AI studies with a working knowledge of LLM limitations and model behavior, and Human Factors and Quantitative Psychology specialists measure trust, comprehension, and reliance.
02
How we test AI
Scenario testing that includes wrong outputs, simulations, longitudinal studies, and decision-path analysis, which traces what users looked at and what they believed the AI had done. For conversational AI, our conversation designers review transcripts across model versions with conversation quality metrics built for each use case, catching the frustration and confusion automated metrics miss.
03
How we measure trust
Our quantitative researchers break trust, engagement, and satisfaction into measurable parts using psychometrics, then build the surveys, data pipelines, and dashboards that track them from launch onward, so product teams see the trend without waiting for a new readout.
04
Confidential and unreleased models
Studies can run inside your own systems so prompts, outputs, and recordings stay in your environment, and they can also run unbranded, so participants don't know which company is behind the product.
05
Who takes part
Recruiting across languages, cultures, user types, and edge cases, plus people with disabilities from our accessibility panels, which cover 14 disability categories. Participant recruiting and logistics are also part of the engagement.
06
How we use AI
AI speeds up parts of the work, like analysis and heuristic evaluation, but our researchers lead every session, the synthesis, and the findings your team takes to leadership.
07
What happens if a researcher leaves
A dedicated PMO manages the researchers, the schedule, and delivery, and keeps the work moving if anyone on the team changes, so your team gets findings without managing more people.
Worth asking about: Which researchers would be matched to your product, and how a first rapid study could answer one real AI decision your team is facing now.
No. 2
Where they're based: San Francisco.
What they offer: A design and innovation company with a user research team that runs contextual inquiry, usability testing, and concept validation. Its AI practice covers agentic AI, multimodal and voice interfaces, and connected devices.
Best for: AI experiences that need design work, especially voice, multimodal interfaces, and connected devices.
Worth asking about: How research and design are balanced in the engagement, and how they'd measure trust and reliance in your AI feature.
No. 3
Where they're based: San Francisco.
What they offer: A full-service UX research and design agency covering research, design, and testing, including interviews, diary studies, usability testing, and human factors.
Best for: Teams that want research, design, and testing from one boutique agency.
Worth asking about: Their experience with AI products specifically, since AI-focused services aren't listed on their site.
No. 4
Where they're based: Bay Area.
What they offer: Consulting led by Steve Portigal, covering in-depth user research on complex questions, help maturing a research practice, research skills training, facilitation, and coaching.
Best for: Research leaders who want to strengthen their own team's practice, skills, and research culture.
Worth asking about: How the work would scale if you need studies run across several product teams at once.
No. 5
Where they're based: San Francisco.
What they offer: A boutique UX and UI design agency for enterprise and B2B software, offering product strategy, UX and UI design, and DesignOps. Its capabilities include usability testing and AI design, such as designing interfaces for AI products and adding generative AI to existing platforms.
Best for: Enterprise and B2B software teams that mainly need design, including AI interfaces, with user testing to support it.
Worth asking about: How much of the engagement is dedicated to research, and which research methods they'd use beyond usability testing.
The firms on this list are built for different jobs, so the fastest way to narrow the list is to start with the kind of help your team needs most.
You need research and design delivered across an AI roadmap
Look for a partner that can embed researchers in your teams, run rapid studies in parallel, design and evaluate AI conversations, and own delivery of the whole program, which is where Akraya fits.
You need one focused AI question answered fast
A rapid study scoped to one decision, like how users respond to a correction flow, gets you findings in weeks, and Akraya's rapid research delivers them in 2 to 3 weeks.
You want a boutique agency for research and design
EchoUser is a boutique UX research and design agency, and Neuron is a design agency that includes user testing in its work.
Your AI lives in a device, voice, or multimodal experience
Punchcut focuses on design for these experiences, and if you need human factors research on fit, comfort, and long wear, Akraya runs that research for wearables.
You want to grow your own team's research practice
Portigal Consulting focuses on research maturity, training, and coaching for research teams.
Once you have a shortlist, ask each firm for a short research plan on a real problem your team has, instead of a capabilities deck. A strong plan shows you most of what you need to know, and here's what to look for in it.
01
A clear research question tied to a decision
The plan should name the product decision the study will inform, because research that isn't tied to a decision tends to end up as a deck nobody acts on.
02
Named researchers with AI experience
You should know who will design the study, run the sessions, and present the findings, and what AI products they've researched before.
03
Methods built for AI
Look for scenario testing with wrong outputs, longitudinal studies, and validated measures of trust and reliance alongside interviews and usability testing.
04
A recruiting plan for the people your AI will reach
That includes domain experts, users in other languages and markets, newer users, and people with disabilities, with screening that keeps out professional test-takers.
05
Clear rules for confidentiality and AI use
The firm should explain how it protects unreleased models and data, and which parts of its own work use AI and how researchers check that output.
06
Findings your team can act on
Each finding should say what's wrong, who it affects, how serious it is, and what to change, so product and engineering can prioritize it.
Some warning signs come up again and again when teams evaluate AI research partners, and any one of these is worth a follow-up question before you sign.
AI is only a label on their website
They can't walk you through a real AI study, the participants involved, or a trust problem they uncovered.
Nobody can tell you who does the work
The people pitching can't name who will run your sessions and analyze the findings.
Every conversation is about screens
An AI partner should talk about workflows, decisions, and risks, beyond layouts and visual design.
They treat AI output as fixed
AI responses change with the prompt, the context, and the model version, so a partner that tests one response once will miss most of what matters.
Big claims from small samples
A handful of interviews can generate ideas, and it can't tell you how trust holds up across thousands of users.
Vague answers about participants
They can't explain how they find hard-to-reach users or screen out professional test-takers.
Use these questions to compare firms side by side, and the answers will tell you quickly which partners have done this work before.
01
What AI products have you researched, and what was the hardest trust or adoption problem you found?
02
Who exactly will run our studies, and what is their AI research experience?
03
How do you tell the difference between users liking the AI and users being able to rely on it safely?
04
How would you test an agent that takes actions on our users' behalf?
05
How do you protect unreleased models, prompts, and customer data during a study?
06
How do you recruit participants in other languages, markets, and with disabilities?
07
Which parts of your own work use AI, and how do your researchers check it?
08
How quickly can you have a study in field, and what happens if our needs grow?
In most enterprise teams, the UX research manager or lead chooses the partner, and product leadership signs off once the choice is made. That sign-off goes faster when you bring three things:
01
The AI decision the research will inform
For example, a launch, a rollout to a new market, or a change to how the AI handles mistakes.
02
What a wrong decision would cost
Put it in terms leadership tracks, like rework, launch delays, or users who stop using the feature.
03
A small first study scoped to that decision
Leadership sees the value before committing to more, and your team gets a low-risk way to see how a partner works.
AI teams rarely need one study and then nothing. Every model update, new market, and new feature raises fresh questions, but most research teams can't answer all of them at the pace product ships.
Akraya is built for that situation, in five ways.
Rapid studies answer focused questions in 2 to 3 weeks, and embedded researchers stay with your teams, learn your product, and track trust and adoption release after release.
A dedicated PMO manages the work, and Akraya owns delivering the research from scoping to findings, so your team keeps control of the research strategy and gets answers it can act on.
AI teams need more than findings, so our teams also design the experience. We design AI conversations and onboarding, evaluate how AI models hold a conversation across versions, and build the frameworks that track trust and engagement after launch.
Akraya also offers AI and data services and AI-led product engineering, so the same partner can help turn findings into model evaluation and product work.
Our PhD-level researchers are vetted against a big tech hiring bar and matched to your problem. They specialize in how people work with AI, from trust and reliance to agents that act for users.
We recruit across languages, markets, edge cases, and people with disabilities, and we can run confidential studies inside your own environment so unreleased models stay protected.
Akraya has worked inside enterprise organizations for 25 years, and our UX research teams support Fortune 500 enterprises building software, AI products, and wearables.
Get the AI research your roadmap needs, from scoping to findings
Tell us about the AI product your team is supporting and the decision coming up next. We'll scope a first study with the right researchers and participants, and you can start with a single study with no long-term commitment.
Talk to our UX research teamFive Bay Area firms worth considering for AI product research are Akraya, Punchcut, EchoUser, Neuron, and Portigal Consulting. The right choice depends on what you need: ongoing research and design with AI specialists across a roadmap, design for devices and multimodal AI, a boutique research and design agency, enterprise design work, or help building your own research practice.
It should research more than interface usability. Look for experience measuring trust and reliance, testing if users can spot and correct wrong outputs, studying behavior across many sessions, testing agents that act on a user's behalf, recruiting hard-to-reach participants, and running confidential studies on unreleased models.
Pricing depends on the scope and the engagement model. A single study is priced differently from an ongoing research program, and costs change with the number of participants, how hard they are to recruit, and the methods involved. Ask each firm for a scoped proposal tied to a real research question so you can compare like for like.
Many enterprise teams do both: they keep research strategy and synthesis in-house and bring in a partner to run more studies, add specialist skills like AI interaction or accessibility research, and handle recruiting. A partner lets your team take on more research without waiting months to hire.
It varies by firm and by how hard the participants are to recruit. At Akraya, a rapid study delivers findings in 2 to 3 weeks.