— Created the course: AI Customer Research Analysis for Designers, Product People and Researchers
— Featured on Maven "AI-Powered Super IC" top courses
— 4 Weeks to cutting your analysis time in half

Founder - AI Analysis Course

AI x Customer Research Newsletter

The only newsletter that digs deep into AI tools, workflows, considerations and relevant news to help you adopt AI for faster Customer Research.

Founder / Researcher

User Research Studio

🛠️ Bringing processes for trustworthy insights to fast-moving startup teams
🌏 Supporting EU, UK and US startups/scale-ups
🚀 Training teams + individuals to 3-10x your research impact

From pre-seed to scale-up, I advise early teams on optimal Customer Research and Product Discovery processes that take the least time for the most impact.

Lean product discovery processes for product teams with limited time and staff to dedicate to research. 

A few ways I've helped startup teams:
🎁 Supported the insights-driven development of a new product. It generated 3x the profit of product #1
⭐️ Pushed a pre-product startup to drop their first MVP and launch one instead that got customer satisfaction ratings >85% and proved product market fit by rapidly reaching first 1000 paid users
💰 Helped a startup launch its first subscription fee after 3+ years of a free product. Estimated risk of lost customers and potential new ones gained before price change
🌏 Helped teams avoid bad investment decisions in new markets and achieve growth by doubling down in the right places
🏃 Trained product teams to run 3-4x as much valuable customer discovery as before in the same time
 

I’m passionate about empowering founders and product teams to run their own trustworthy research for better products.

How I work: 
- Team advisor
- Workshop training sessions 
- 1:1 coaching

Founder, Customer Discovery Advisor

Relive

Customer Discovery Advisor

ManyPets

Clayground Ceramics Berlin

Sold in 2023. Built a highly demanded studio space and community in a craft niche. The studio was developed around lean research, identifying key insights pre-investment that enabled us to differentiate and thrive despite increasing competition and a pandemic. I used the business as a place to experiment in business model development and service design, as well as prove that customer research can make a dramatic difference for businesses of all sizes and industries.

Co-Founder

Soundtrack Your Brand

User Research Advisor

For the Spotify Business product that then became "Soundtrack", I built the customer insights discipline from within the Product team while partnering with Growth and Marketing. 

I put discovery and validation processes in place so they could:

⭐️ Launch a new service tier with significantly higher conversion 
⭐️ Nearly double their customer base
⭐️ De-risk pricing and offer changes that discreetly found the right price 
⭐️ Supported the product team to make faster decisions (not slow them down with research) 
⭐️ Supported a major brand transition

Supported the C-level to constantly challenge assumptions and find the right balance between listening to customer requests and innovating for their underlying needs.

Head of User Research

Dreams Sustainable

Led the customer insights strategy and hands-on work for a team pre-product. Worked with them to establish foundational understanding of a new target audience, identify first best-fit users, narrow focus to the most important customer problems to solve, and clarify potential MVP feature set.

Dreams

Led UX Research on a contract-basis for early stage fintech startup with a product but no UXR team. Helped the team define and pinpoint their best target audiences, find barriers to engagement, and prepare for risks of launching a subscription fee for the first time with minimal damage.

Later supported their first Head of UXR by providing training and mentorship to include diverse team members in research, enable more people to lead product discovery projects, and improve frequency of internal insights-sharing for full-team impact.

Fjord

Worked with large Swedish companies to innovate in their domains, and solve user problems better than their competitors. Was responsible for UX Research, Service Design, UX Design and client management for companies across banking and retail industries.

Senior Interaction Designer

Keeping designers, marketers, and product people ahead with faster high-quality AI-assisted customer insights—from YouTube and Volvo to Supermetrics and Breyta.ai.

I partner with teams for short- and mid-term collaborations, focused on driving faster customer clarity and smarter growth decisions:

🎯 Advisory Roles – I work as a short-term or fractional partner, helping teams use AI to uncover customer insights faster, sharpen positioning, and find hidden growth opportunities.

🚀 Training & Workshops – I design custom AI trainings and hands-on workshops to equip your team with faster insight systems—so you can learn what customers want and move smarter, not slower.

🔍 Consulting Projects – Tailored work to solve specific customer insight challenges—like refining messaging, accelerating product validation, or building AI-powered feedback loops.

⚡ Strategic Sprints – Fast, high-impact projects where we apply AI to customer research and insight generation in focused, action-driven phases (perfect for teams moving fast).

🤝 Team Discounts on AI Content – Clients I work with get my premium newsletter for FREE plus major discounts on my courses to continuously keep their teams ahead.

💌 If one of these is the accelerator help you need, email * hi@userresearchstudio.com *



____

A few current/former clients: Lufthansa Airlines, Relive, Aumio, ManyPets, Yubo, Spotify Business/Soundtrack, Dreams Nordic, Google, Scandinavian Airlines, Hemköp, Nordea, Dunkin'​ Donuts, AT&T.

⚠️ Hey Marketers, Researchers, PMs: 𝗚𝗣𝗧-5 𝘀𝘁𝗶𝗹𝗹 𝘀𝘂𝗰𝗸𝘀 𝗮𝘁 𝘃𝗲𝗿𝗯𝗮𝘁𝗶𝗺 𝗾𝘂𝗼𝘁𝗲𝘀. Watch out.

If you work in 𝗩𝗼𝗶𝗰𝗲 𝗼𝗳 𝗖𝘂𝘀𝘁𝗼𝗺𝗲𝗿 or 𝘄𝗿𝗶𝘁𝗲 𝗺𝗮𝗿𝗸𝗲𝘁𝗶𝗻𝗴 𝗺𝗲𝘀𝘀𝗮𝗴𝗲𝘀 that rely on exact customer language…

Maybe you were as hopeful as I was that ChatGPT would finally make it easy to pull out real customer quotes.

Unfortunately…
A bigger, shinier model doesn’t mean more accurate quotes.

For all the benchmarks showing improved results on math tasks and Humanity’s Last Exam…

…the improvements don’t translate here.

𝗪𝗵𝗲𝗻 𝗚𝗣𝗧-5 “𝗰𝗹𝗲𝗮𝗻𝘀 𝘂𝗽” 𝘄𝗵𝗮𝘁 𝗽𝗲𝗼𝗽𝗹𝗲 𝘀𝗮𝘆, 𝗶𝘁 𝗰𝗵𝗮𝗻𝗴𝗲𝘀 critical 𝗺𝗲𝗮𝗻𝗶𝗻𝗴.

Sometimes subtly, sometimes a WHOLE lot.

Bad quotes = bad customer understanding = bad messaging.

𝗜 𝘀𝗽𝗲𝗻𝘁 𝘁𝗵𝗲 𝘄𝗲𝗲𝗸𝗲𝗻𝗱 𝗿𝘂𝗻𝗻𝗶𝗻𝗴 𝗺𝘆 𝘂𝘀𝘂𝗮𝗹 𝘀𝘁𝗿𝗲𝘀𝘀 𝘁𝗲𝘀𝘁𝘀 𝘁𝗼 𝗲𝘅𝗽𝗹𝗼𝗿e 𝘁𝗵𝗲 𝗻𝗲𝘄 𝗺𝗼𝗱𝗲𝗹:

📑 Loads of customer interview transcripts
1️⃣ First, using simple single prompts 
(clear role, context, 1-line task, request for verbatim quotes with references)
2️⃣ Then, my step-by-step workflow (it gets accurate, thorough results from all major LLMs)

𝗛𝗲𝗿𝗲'𝘀 𝘄𝗵𝗮𝘁 𝗵𝗮𝗽𝗽𝗲𝗻𝗲𝗱:
Without a precise process…
Results are not any better than before. Bummer. 
🙄 All that hype for nothing.

They’re still full of paraphrases of multiple statements that might as well be invented.
(Because the user did not actually say them.)

When I ran my tight, heavily tested analysis workflow - built for any qualitative research, not just VoC - the results were as good as they always are.

No, GPT-5 isn’t magic for those of us running transcripts and surveys through ChatGPT.

It’s the same deal as always with ChatGPT: 
You need to ace communicating what you need and *force it* to color inside the lines to get trustworthy results.

〰️
👋 𝗜𝗳 𝘆𝗼𝘂’𝗿𝗲 𝘀𝘁𝗶𝗹𝗹 𝘀𝘁𝗿𝘂𝗴𝗴𝗹𝗶𝗻𝗴 𝘁𝗼 𝗴𝗲𝘁 𝗿𝗲𝘀𝘂𝗹𝘁𝘀 𝘆𝗼𝘂 𝗰𝗮𝗻 𝘁𝗿𝘂𝘀𝘁 𝗳𝗿𝗼𝗺 𝗔𝗜 𝗮𝗻𝗮𝗹𝘆𝘀𝗶𝘀...

September could *finally* be your turning point.

I’m re-running my AI Customer Research Analysis course.
⭐️ Every rating from June's cohort was 10/10!
🆕 September will have the latest LLMs and live breakdowns of what works now.
🤝 Every participant gets personalized 1:1 support from me
🪑 Limited spots so I can give you loads of feedback

🔗 Learn more → Link in comments + my profile

Caitlin Sullivan

AI for Customer Clarity ⚡️ Guidance to figure out customers 3x faster. Training Fortune 500 and growth-ready teams to use AI.

If you haven’t watched it yet, Aneta Kmiecik picked some top prompting tips from my Maven Lightning Lesson 👇

uxportfolio.co | Helping designers build their portfolios

Many designers are asking the same question: “Will AI replace me?” It’s a wrong question. What you should be asking is "How do I become irreplaceable?

Fewer junior design roles, “restructuring” due to AI, and managers transitioning to ICs. The design market is messy and hyper competitive right now. We all feel it. Designers are stressed because no one knows the future of the industry. At the same time, we’re all overwhelmed by the number of AI tools and tactics we need to learn.

The way we design is changing, so what can we do to survive? 👀

✅ The only certain things in your design career are your skills
They’re lifelong, so invest in them. Learn AI now and become a Super IC so you can become “unfireable”.

That’s why in this carousel, I’m sharing 14 ways to use AI in your design workflow. I want to help you learn right away, without waiting.

All insights come from Maven Lightning Lessons taught by AI design practitioners with real world experience.


These are my favourite free lessons ⬇️

1️⃣ AI Customer Insights: Prompt Fixes + Top LLMs for Analysis by Caitlin Sullivan: https://shorturl.at/eol0K

2️⃣ Rapid Product Discovery with AI-Powered Interviews by John Whalen, PhD: https://shorturl.at/nQYRB

3️⃣ Create AI-Powered Prototypes as a Product Designer by Xinran Ma: https://shorturl.at/WpGzY


If you want more, consider joining AI courses on Maven for designers.
I loved this one ➡️ https://bit.ly/42N3JGH

🎁 Use my code ANETAxMAVEN to save on any AI design Maven course (check the comments)


What’s your way of using AI in a design workflow? Let me know! 😊⬇️


#MavenPartner #Maven #UXDesigner #AIDesigner #AI

14 Ways to use  AI in your design workflow

People choose transcript tools like they're picking lottery numbers. 𝗨𝘀𝗲 𝗔𝗜 𝘁𝗼 𝗰𝗼𝗺𝗽𝗮𝗿𝗲 𝘁𝗵𝗲𝗺 𝗶𝗻 <15𝗺𝗶𝗻 👇

In all the conversations I have with my course students, clients and newsletter subscribers, one thing comes up A LOT:

✦ AI transcription tools ✦

Getting good transcripts is *kiiiiind of* essential for running great AI analysis.

Bad transcripts >> bad analysis.
(There’s a lot more to the process, but you’re more or less screwed if the data input is bad).

There’s a piece here that rests on *us* though.

𝗠𝗼𝘀𝘁 𝗽𝗲𝗼𝗽𝗹𝗲 𝘁𝗿𝘆𝗶𝗻𝗴 𝘁𝗿𝗮𝗻𝘀𝗰𝗿𝗶𝗽𝘁𝗶𝗼𝗻 𝘁𝗼𝗼𝗹𝘀 𝗮𝗿𝗲𝗻'𝘁 𝗿𝘂𝗻𝗻𝗶𝗻𝗴 𝘀𝘆𝘀𝘁𝗲𝗺𝗮𝘁𝗶𝘇𝗲𝗱 𝘁𝗲𝘀𝘁𝘀.

They aren’t running processes to compare transcripts the same way every time, and with concrete metrics. 

I think 𝗮 𝗹𝗼𝘁 𝗼𝗳 𝗽𝗲𝗼𝗽𝗹𝗲 𝘁𝗵𝗶𝗻𝗸 𝘁𝗵𝗮𝘁 𝘀𝗼𝘂𝗻𝗱𝘀 𝗿𝗲𝗮𝗹𝗹𝘆 𝗵𝗮𝗿𝗱 - time-consuming and complicated.

But it doesn't need to be!

In fact…
✦ You can go from generating the transcripts in 2+ tools to final results compared in <1 hour
✦ It can take <15 minutes if you have transcripts ready to compare

Understanding which transcription tool is more reliable is really just a prompt away.


Here’s a 👀 sneak peek 👀 into my July newsletter edition -
𝗧𝗵𝗲 𝗾𝘂𝗶𝗰𝗸 𝗲𝘃𝗮𝗹𝘂𝗮𝘁𝗶𝗼𝗻 where 𝗖𝗹𝗮𝘂𝗱𝗲 𝘁𝗲𝗹𝗹𝘀 𝘆𝗼𝘂 𝘄𝗵𝗶𝗰𝗵 𝘁𝗿𝗮𝗻𝘀𝗰𝗿𝗶𝗽𝘁 𝘁𝗼𝗼𝗹 𝗶𝘀 𝘄𝗼𝗿𝘁𝗵 𝘁𝗿𝘂𝘀𝘁𝗶𝗻𝗴.

Here's the exact prompt:

➡️ “Act as a master of transcript accuracy and comparison. 
Run Python: jiwer.compare to compare the words transcribed in the [#] transcripts uploaded.”

(Yep, it’s one of the shortest prompts I’ve ever written, but it works.)

☑️ Upload transcripts from 2+ tools
☑️ Paste this prompt
☑️ Get instant WER scores (Word Error Rate = rate of discrepancy between the transcripts)

Basically, using WER scores helps you see how transcripts from new platforms you're testing compare with a solid transcript (your "control) - or with each other.

Fun findings I've had from this one:

The most trusted tool of the bunch I tested recently performed worst.

The winner was a tool I’d written off based on previous tests.

Without this system, I’d be recommending some tools and not others, based on incomplete evidence.

Systematic evaluations matter.

—

𝗗𝗼 𝘆𝗼𝘂 𝘀𝘁𝗶𝗹𝗹 𝘄𝗮𝗻𝘁 𝘁𝗼 𝗸𝗻𝗼𝘄 𝗮𝗯𝗼𝘂𝘁…
- the tools I tested and how they performed
- the results of my tests
- the full output I got from this prompt
- the exact test protocol to follow step-by-step
(and more)?

✦ That’s all in my July newsletter going out TOMORROW ✦
(It’s free now!)

Link in comments 👇 / my profile

𝗠𝗼𝘀𝘁 𝘁𝗲𝗮𝗺𝘀 𝘀𝘁𝗶𝗹𝗹 𝗱𝗼 𝗮𝗻𝗮𝗹𝘆𝘀𝗶𝘀 𝘄𝗶𝘁𝗵 𝗔𝗜 𝗮𝗹𝗹 𝘄𝗿𝗼𝗻𝗴 → I’m teaching key parts of my course LIVE at 𝗘𝗫𝗟 2025 (the only chance to learn it in person) 👏

I’ve run my 4-week course on Maven multiple times this year with people and teams at Spotify, Instacart, Typeform and the Mayo Clinic.

But then I got a little challenge from Ben Labay.

Could I run the course IN PERSON at #EXL2025?

So I’m bringing my course to Austin - for the first time offline.
(It will probably be the last time, too.)

Here’s how we’ll do this -

It will be a fast-paced, hands-on session made for anyone doing customer research. We’ll:
• Implement workflows that actually get solid, trustworthy AI analysis
• Understand what causes LLMs to hallucinate, and prevents it
• Decide what to automate in analysis processes - and what to own


Plus…the speaker lineup is bonkers. It includes Elena Verna, John Cutler, and moderation by fellow Berlin resident Nils Stotz. ⚡️

I think EXL 2025 is going to be *special*.

If you’re attending, please come say hi.

And come get better at using AI for analysis so your product changes are based on something real (just discovered much faster).

The deets are here 👇
https://lnkd.in/dPCytC-G

👋 Question for you: Would you rather get a reply 7 days late or instantly from AI pretending to be me? 🙃

Here’s why I ask:

I’m a team of one.
With a small kid.
And a very full plate.

So I prioritize *hard*.

Most messages in my inbox don’t get a reply for days.

Sometimes…it’s weeks. 🫣
And I truly hate that.

It’s not that I don’t care.
I just don’t have enough hours.
If I replied to everyone in real time… 

I’d never get any actual work done.
(I’d also never see my kid, more importantly).

So far, I’ve refused to let AI reply for me.
I have never sent emails or messages that AI composed.

But I’m starting to wonder…

Which feels worse?
⌛️ A REALLY late reply from me
🤖 An instant reply from AI that sounds like me

There’s no right answer. 
I’d really love your take in the comments.
👇

Taplio

Caitlin Sullivan's Linkedin Analytics

What is Caitlin talking about?

's Best Posts (last 30 days)

Want to drive more opportunities from LinkedIn?

Taplio

.css-1jgreyo{background-image:linear-gradient(284deg, rgba(39,43,72,1) 0%, rgba(33,47,183,1) 100%);color:transparent;-webkit-background-clip:text;background-clip:text;}Caitlin Sullivan's Linkedin Analytics

What is Caitlin talking about?

's Best Posts .css-qqfgvy{font-size:var(--chakra-fontSizes-md);}(last 30 days)

Want to drive more opportunities from LinkedIn?

Caitlin Sullivan's Linkedin Analytics

's Best Posts (last 30 days)