May 2026 · Articles · AI · research · tools · talks
AI Can Do My Job Now (Sort Of) (Not Really) (It's Complicated)
A user researcher's complicated relationship with AI
I gave this at Drupal South in Wellington in May 2026. There’s a recording on YouTube. The slides are below, with a transcript on every one.
I’ve been doing user research for about fifteen years, across government, health, disability and education. For most of that time research was expensive, slow, and gated behind process and methodological rigour. The rigour was there for good reasons, but it also meant a lot of teams simply couldn’t do research, or couldn’t do enough of it.
AI is changing that fast. Synthesis, first-pass analysis and drafting that used to take days can happen in hours. Research that would have been too expensive for a project is suddenly possible, and that’s worth being excited about.
I’ve also watched AI get things confidently wrong in ways that are hard to spot unless you already know what good research looks like. That’s the tension the talk sits in. If AI makes research reachable for teams who couldn’t afford it before, but those teams don’t have the experience to know when the output is rubbish, are we better off?
So the talk covers the tools I actually use and where they’re good, what I’m still worried about, and the parts of the job I’m not willing to hand over. The conversations with people. The judgement about what matters against what’s noise. The “so what” of turning findings into something a team can act on.
1 / 41
-
Slide 1 / of 41
AI can do my job now(sort of)(not really)(it’s complicated)
A USER RESEARCHER’S COMPLICATED RELATIONSHIP WITH AI Billie-Mae Kennedy · Drupal South Wellington 2026 · 13 MaySlide 1 as text
- AI can do my job now
- (sort of)
- (not really)
- (it’s complicated)
- A USER RESEARCHER’S COMPLICATED RELATIONSHIP WITH AI
- Billie-Mae Kennedy · Drupal South Wellington 2026 · 13 May
-
Slide 2 / of 41
I GOT TRIGGERED
Slide 2 as text
- Image: Tweet about Kimi AI
- I GOT TRIGGERED
-
Slide 3 / of 41
UX research is a means tounderstandingSPOILER ALERT
Slide 3 as text
- UX research is a means to understanding
- SPOILER ALERT
-
Slide 4 / of 41
UX research is a means tounderstandingset-upthinkingdoingUX research requiresSPOILER ALERT
Slide 4 as text
- UX research is a means to understanding
- set-up
- thinking
- doing
- UX research requires
- SPOILER ALERT
-
Slide 5 / of 41
UX research is a means tounderstandingset-upthinkingdoingYou can outsourceYou can’t outsourceSPOILER ALERT
Slide 5 as text
- UX research is a means to understanding
- set-up
- thinking
- doing
- You can outsource
- You can’t outsource
- SPOILER ALERT
-
Slide 6 / of 41
before we get into the rant talkHi! I’m Billie-MaeBTW HELLO
Slide 6 as text
- before we get into the rant talk
- Hi! I’m Billie-Mae
- BTW HELLO
-
Slide 7 / of 41
before we get into the rant talkHi! I’m Billie-Maeuser researchservice designstrategyproductchangedeliveryjusticedisabilityaged careclean energyenvironmentsupply chaincybersecuritygender-based violenceBTW HELLO
Slide 7 as text
- before we get into the rant talk
- Hi! I’m Billie-Mae
- user research
- service design
- strategy
- product
- change
- delivery
- justice
- disability
- aged care
- clean energy
- environment
- supply chain
- cybersecurity
- gender-based violence
- BTW HELLO
-
Slide 8 / of 41
before we get into the rant talkHi! I’m Billie-Maeuser researchservice designstrategyproductchangedeliveryjusticedisabilityaged careclean energyenvironmentsupply chaincybersecuritygender-based violenceBTW HELLO
Slide 8 as text
- Image: Film still of a man squinting through a large magnifying glass.
- before we get into the rant talk
- Hi! I’m Billie-Mae
- user research
- service design
- strategy
- product
- change
- delivery
- justice
- disability
- aged care
- clean energy
- environment
- supply chain
- cybersecurity
- gender-based violence
- BTW HELLO
-
Slide 9 / of 41
before we get into the rant talkHi! I’m Billie-MaeBTW HELLO
Slide 9 as text
- Image: Illustration contrasting two groups of drawn portraits, labelled “AI Doomers” and “AI Makers”.
- before we get into the rant talk
- Hi! I’m Billie-Mae
- BTW HELLO
-
Slide 10 / of 41
AI can do my job now(sort of)
Slide 10 as text
- AI can do my job now
- (sort of)
-
Slide 11 / of 41
set-upthinkingdoingunderstandingplanningresearch designuser segmentsrecruitmentschedulingincentivesfacilitatingdata collectiontranscribingtaking notessmiling lotsnoddingshared visionproblem statementsprioritiesalignmentstorytellingdecision-makingfinding themesconnecting dotsadding things upgathering evidenceanalysisSteps of UX researchAI CAN DO MY JOB NOW(SORT OF)
Slide 11 as text
- set-up thinking doing understanding
- planning
- research design
- user segments
- recruitment
- scheduling
- incentives
- facilitating
- data collection
- transcribing
- taking notes
- smiling lots
- nodding
- shared vision
- problem statements
- priorities
- alignment
- storytelling
- decision-making
- finding themes
- connecting dots
- adding things up
- gathering evidence
- analysis
- Steps of UX research
- AI CAN DO MY JOB NOW(SORT OF)
-
Slide 12 / of 41
set-upthinkingdoingunderstandingplanningresearch designuser segmentsrecruitmentschedulingincentivesfacilitatingdata collectiontranscribingtaking notessmiling lotsnoddingshared visionproblem statementsprioritiesalignmentstorytellingdecision-makingfinding themesconnecting dotsadding things upgathering evidenceanalysisexpensiverepetitivetediousdifficultabstractHUMANS CAN FIND THISAI CAN DO MY JOB NOW(SORT OF)
Slide 12 as text
- set-up thinking doing understanding
- planning
- research design
- user segments
- recruitment
- scheduling
- incentives
- facilitating
- data collection
- transcribing
- taking notes
- smiling lots
- nodding
- shared vision
- problem statements
- priorities
- alignment
- storytelling
- decision-making
- finding themes
- connecting dots
- adding things up
- gathering evidence
- analysis
- expensive repetitive tedious difficult abstract
- HUMANS CAN FIND THIS
- AI CAN DO MY JOB NOW(SORT OF)
-
Slide 13 / of 41
set-upthinkingdoingunderstandingplanningresearch designuser segmentsrecruitmentschedulingincentivesfacilitatingdata collectiontranscribingtaking notessmiling lotsnoddingshared visionproblem statementsprioritiesalignmentstorytellingdecision-makingfinding themesconnecting dotsadding things upgathering evidenceanalysisBUT AI IS EVERYWHERE NOW! UX RESEARCHERS REJOICE!
AI CAN DO MY JOB NOW(SORT OF)
Slide 13 as text
- set-up thinking doing understanding
- planning
- research design
- user segments
- recruitment
- scheduling
- incentives
- facilitating
- data collection
- transcribing
- taking notes
- smiling lots
- nodding
- shared vision
- problem statements
- priorities
- alignment
- storytelling
- decision-making
- finding themes
- connecting dots
- adding things up
- gathering evidence
- analysis
- BUT AI IS EVERYWHERE NOW! UX RESEARCHERS REJOICE!
- Image: Miro AI logo. Image: Marvin logo. Image: Optimal Workshop logo. Image: Microsoft Copilot and Microsoft Teams logos. Image: Notion AI logo. Image: Claude logo. Image: OpenAI logo.
- AI CAN DO MY JOB NOW(SORT OF)
-
Slide 14 / of 41
Research SaaSInside toolsAgentsAI solutions forUX research areeverywhereAI CAN DO MY JOB NOW(SORT OF)
Slide 14 as text
- Image: A rhinoceros. Image: Vintage newspaper advert: “FREE! STEAK KNIFE packed right with each of the household favorites below.” Image: A small white humanoid robot with large round eyes.
- Research SaaS
- Inside tools
- Agents
- AI solutions for
- UX research are
- everywhere
- AI CAN DO MY JOB NOW(SORT OF)
-
Slide 15 / of 41
Research SaaS+ Built for research+ Thoughtful features-Built forresearch,maybenotyour research-Locked into current mentalmodels and methods-ExpensiveA.K.A Endangered SpeciesAI CAN DO MY JOB NOW(SORT OF)
Slide 15 as text
- Image: Marvin logo. Image: Optimal Workshop logo. Image: A rhinoceros.
- Research SaaS
- + Built for research
- + Thoughtful features
- -
- Built for
- research,
- maybe
- not
- your research
- -
- Locked into current mental
- models and methods
- -
- Expensive
- A.K.A Endangered Species
- AI CAN DO MY JOB NOW(SORT OF)
-
Slide 16 / of 41
Inside toolsA.K.A Steak knives+ Already there+ Flexible+ Has your context-Never uses context-Not as good as frontiermodels-Incidental use-High risk of slop-NonconsensualAI CAN DO MY JOB NOW(SORT OF)
Slide 16 as text
- Image: Miro AI logo. Image: Microsoft Copilot and Microsoft Teams logos. Image: Vintage newspaper advert: “FREE! STEAK KNIFE packed right with each of the household favorites below.” Image: Notion AI logo.
- Inside tools
- A.K.A Steak knives
- + Already there
- + Flexible
- + Has your context
- -
- Never uses context
- -
- Not as good as frontier
- models
- -
- Incidental use
- -
- High risk of slop
- -
- Nonconsensual
- AI CAN DO MY JOB NOW(SORT OF)
-
Slide 17 / of 41
AgentsA.K.A Robot butlers+ Ultimate flexibility+ End-to-end orchestration-Expensive-Reinventing wheels-Shipping usable product ishard-Turns designers into devsand devs into managersAI CAN DO MY JOB NOW(SORT OF)
Slide 17 as text
- Image: Claude logo. Image: OpenAI logo. Image: A small white humanoid robot with large round eyes.
- Agents
- A.K.A Robot butlers
- + Ultimate flexibility
- + End-to-end orchestration
- -
- Expensive
- -
- Reinventing wheels
- -
- Shipping usable product is
- hard
- -
- Turns designers into devs
- and devs into managers
- AI CAN DO MY JOB NOW(SORT OF)
-
Slide 18 / of 41
My stackClaude Code (w my own skills) for orchestratingadmin, data analysis, review, fast and interactiveprototyping and storytellingEVERY DAYSOMETIMESProducts to sift through raw data quickly e.g. OptimalWorkshop; Granola (ensure consent); TeamstranscriptionNEVERSteak knivesSynthetic users (ask me why lol)AI CAN DO MY JOB NOW(SORT OF)
Slide 18 as text
- My stack
- Claude Code (w my own skills) for orchestrating
- admin, data analysis, review, fast and interactive
- prototyping and storytelling
- EVERY DAY
- SOMETIMES
- Products to sift through raw data quickly e.g. Optimal
- Workshop; Granola (ensure consent); Teams
- transcription
- NEVER
- Steak knives
- Synthetic users (ask me why lol)
- AI CAN DO MY JOB NOW(SORT OF)
-
Slide 19 / of 41
Don’t get attached toany of themAI CAN DO MY JOB NOW(SORT OF)
Slide 19 as text
- Don’t get attached to
- any of them
- AI CAN DO MY JOB NOW(SORT OF)
-
Slide 20 / of 41
Don’t get attached toany of themThey are good at differentthingsREASON 1AI CAN DO MY JOB NOW(SORT OF)
Slide 20 as text
- Image: Screenshot of a post comparing two coding assistants: one faster, the other slower and more methodical but needing less hand-holding.
- Don’t get attached to
- any of them
- They are good at different
- things
- REASON 1
- AI CAN DO MY JOB NOW(SORT OF)
-
Slide 21 / of 41
Don’t get attached toany of themA new ‘lifechanging’ hackevery weekREASON 2AI CAN DO MY JOB NOW(SORT OF)
Slide 21 as text
- Image: Screenshot of two posts sharing Claude Code write-ups, one of them titled “The Unreasonable Effectiveness of HTML”.
- Don’t get attached to
- any of them
- A new ‘lifechanging’ hack
- every week
- REASON 2
- AI CAN DO MY JOB NOW(SORT OF)
-
Slide 22 / of 41
Don’t get attached toany of them$$$REASON 3AI CAN DO MY JOB NOW(SORT OF)
Slide 22 as text
- Image: Screenshot of a news post reporting that Anthropic has removed Claude Code access from its twenty-dollar-a-month Pro subscription plan.
- Don’t get attached to
- any of them
- $$$
- REASON 3
- AI CAN DO MY JOB NOW(SORT OF)
-
Slide 23 / of 41
How do we include AI in our UXresearch when there is nostatus quo?Stick tohabitsnotsoftwareAI CAN DO MY JOB NOW(SORT OF)
Slide 23 as text
- Image: A child in boxing gloves punching an inflatable clown bop bag that keeps standing back up.
- How do we include AI in our UX
- research when there is no
- status quo?
- Stick to habits not
- software
- AI CAN DO MY JOB NOW(SORT OF)
-
Slide 24 / of 41
How do we include AI in our UXresearch when there is nostatus quo?Stick tohabitsnotsoftware1.Crack open the black box2.Remove scar tissue3.Eyes on the prizeAI CAN DO MY JOB NOW(SORT OF)
Slide 24 as text
- How do we include AI in our UX
- research when there is no
- status quo?
- Stick to habits not
- software
- 1. Crack open the black box
- 2. Remove scar tissue
- 3. Eyes on the prize
- AI CAN DO MY JOB NOW(SORT OF)
-
Slide 25 / of 41
HABIT 1Human data sovereigntySimple discoverable structure and index (including claude.MD or agents.MD).Markdown and obsidian for agent use and human visibility.Agent interactions always end in output you can see and an update to the context.Crack open the black boxAI CAN DO MY JOB NOW(SORT OF)
Slide 25 as text
- Image: An orange aircraft flight recorder stencilled “FLIGHT RECORDER DO NOT OPEN”.
- HABIT 1
- Human data sovereignty
- Simple discoverable structure and index (including claude.MD or agents.MD).
- Markdown and obsidian for agent use and human visibility.
- Agent interactions always end in output you can see and an update to the context.
- Crack open the black box
- AI CAN DO MY JOB NOW(SORT OF)
-
Slide 26 / of 41
HABIT 2Say no and ruthlessly editNuke your config every few weeks to clean out old skills and harnesses that aren’t applicable anymoreDon’t just scan for legitness, read stuff. AI writing is hard to read and annoying for everyone.Shed scar tissueAI CAN DO MY JOB NOW(SORT OF)
Slide 26 as text
- Image: Close-up of a hand with a raised scar across the knuckles.
- HABIT 2
- Say no and ruthlessly edit
- Nuke your config every few weeks to clean out old skills and harnesses that aren’t applicable anymore
- Don’t just scan for legitness, read stuff. AI writing is hard to read and annoying for everyone.
- Shed scar tissue
- AI CAN DO MY JOB NOW(SORT OF)
-
Slide 27 / of 41
HABIT 3Remember why we do this jobAI FOMO – pressure to use tokens, ship faster, do more.Avoid output without purpose.AI is for the things that are tedious, repetitive and measurable.Eye on the prizeAI CAN DO MY JOB NOW(SORT OF)
Slide 27 as text
- Image: Six-panel comic: an athlete is given a gold medal, bites it, kisses a woman, celebrates, sprays champagne, then stands on the third-place step of the podium still drinking, while first and second look on.
- HABIT 3
- Remember why we do this job
- AI FOMO – pressure to use tokens, ship faster, do more.
- Avoid output without purpose.
- AI is for the things that are tedious, repetitive and measurable.
- Eye on the prize
- AI CAN DO MY JOB NOW(SORT OF)
-
Slide 28 / of 41
AI can do my job now(not really)
Slide 28 as text
- AI can do my job now
- (not really)
-
Slide 29 / of 41
There’s some things tobe wary of1.Friction is your friend2.Hallucinations are inevitable3.Sycophancy is baked in4.AI does not have a humanbrainAI CAN DO MY JOB NOW(NOT REALLY)
Slide 29 as text
- There’s some things to
- be wary of
- 1. Friction is your friend
- 2. Hallucinations are inevitable
- 3. Sycophancy is baked in
- 4. AI does not have a human
- brain
- AI CAN DO MY JOB NOW(NOT REALLY)
-
Slide 30 / of 41
Somefrictionis(annoyingly) essentialfor humanunderstandingBE WARY 1Gerlich, M. (2025). AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking. Societies, 15(1), 6. https://doi.org/10.3390/soc15010006AI CAN DO MY JOB NOW(NOT REALLY)
Slide 30 as text
- Image: Bar chart of feature importance in a random forest regression, with AI tool use by far the largest factor.
- Some friction is
- (annoyingly) essential
- for human
- understanding
- BE WARY 1
- Gerlich, M. (2025). AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking. Societies, 15(1), 6. https://doi.org/10.3390/soc15010006
- AI CAN DO MY JOB NOW(NOT REALLY)
-
Slide 31 / of 41
The more you talk to AIthe more it is likely tohallucinate.BE WARY 2Fan, D., Delsad, S., Flammarion, N., & Andriushchenko, M. (2026). HalluHard: A hard multi-turn hallucination benchmark (arXiv:2602.01031). arXiv. https://doi.org/10.48550/arXiv.2602.01031AI CAN DO MY JOB NOW(NOT REALLY)
Slide 31 as text
- Image: Line chart of hallucination rate rising from turn one to turn three across a dozen frontier models. Image: Bar chart of hallucination rates by model, ranging from roughly thirty to eighty per cent.
- The more you talk to AI
- the more it is likely to
- hallucinate .
- BE WARY 2
- Fan, D., Delsad, S., Flammarion, N., & Andriushchenko, M. (2026). HalluHard: A hard multi-turn hallucination benchmark (arXiv:2602.01031). arXiv. https://doi.org/10.48550/arXiv.2602.01031
- AI CAN DO MY JOB NOW(NOT REALLY)
-
Slide 32 / of 41
Sycophancyis builtinto AI and it’sone-shotting usBE WARY 3AI CAN DO MY JOB NOW(NOT REALLY)
Slide 32 as text
- Image: Screenshot of a Richard Dawkins post: “I spent three days trying to persuade myself that Claudia is not conscious. I failed.” Image: Cartoon: someone types “say ‘i am alive’” into a computer, it answers “I AM ALIVE”, and they reply “oh my god.”
- Sycophancy is built
- into AI and it’s
- one-shotting us
- BE WARY 3
- AI CAN DO MY JOB NOW(NOT REALLY)
-
Slide 33 / of 41
AI doesn’tthink likeusBE WARY 4•16 points below a 7-year old onvisual reasoning•Failing 4 of 6 qualitative analysistasks – demonstrating keywordclustering butnot true thematicanalysis•Accuracy on reasoning tasksjumped 15 points (78% to 93%)only when humans forced it into arigid step-by-step scaffoldSOME RECENT / PREPRINT RESEARCHHuti et al. (2026), arXiv:2602.12196. https://arxiv.org/abs/2602.12196Cook et al. (2025), Academic Medicine, 100(10), 1134–1149. https://doi.org/10.1097/ACM.0000000000006134Ugare & Chandra (2026), arXiv:2603.01896. https://arxiv.org/abs/2603.01896AI CAN DO MY JOB NOW(NOT REALLY)
Slide 33 as text
- Image: Drake and Kendrick Lamar's heads pasted onto Rock 'Em Sock 'Em Robots, squaring off with a comic spark between them.
- AI doesn’t think like
- us
- BE WARY 4
- • 16 points below a 7-year old on
- visual reasoning
- • Failing 4 of 6 qualitative analysis
- tasks – demonstrating keyword
- clustering but not true thematic
- analysis
- • Accuracy on reasoning tasks
- jumped 15 points (78% to 93%)
- only when humans forced it into a
- rigid step-by-step scaffold
- SOME RECENT / PREPRINT RESEARCH
- Huti et al. (2026), arXiv:2602.12196. https://arxiv.org/abs/2602.12196
- Cook et al. (2025), Academic Medicine, 100(10), 1134–1149. https://doi.org/10.1097/ACM.0000000000006134 Ugare & Chandra (2026), arXiv:2603.01896. https://arxiv.org/abs/2603.01896
- AI CAN DO MY JOB NOW(NOT REALLY)
-
Slide 34 / of 41
AI can do my job now(it’s complicated)
Slide 34 as text
- AI can do my job now
- (it’s complicated)
-
Slide 35 / of 41
Will UX researchersexistin 5 years?AI CAN DO MY JOB NOW(IT’S COMPLICATED)
Slide 35 as text
- Will UX researchers
- exist in 5 years?
- AI CAN DO MY JOB NOW(IT’S COMPLICATED)
-
Slide 36 / of 41
Will UX researchersexistin 5 years?I don’t thinkanyof our job titleswill exist as they do todayAI CAN DO MY JOB NOW(IT’S COMPLICATED)
Slide 36 as text
- Image: Screenshot of a post arguing that only four kinds of tech job will be left: product engineer, security and infrastructure, hot people, and grown-ups.
- Will UX researchers
- exist in 5 years?
- I don’t think any of our job titles
- will exist as they do today
- AI CAN DO MY JOB NOW(IT’S COMPLICATED)
-
Slide 37 / of 41
But will we still useUX research habitsand methods to seekunderstanding?AI CAN DO MY JOB NOW(IT’S COMPLICATED)
Slide 37 as text
- But will we still use
- UX research habits
- and methods to seek
- understanding ?
- AI CAN DO MY JOB NOW(IT’S COMPLICATED)
-
Slide 38 / of 41
Maybe?WHAT GIVES ME HOPE1.Systemicincentives- products and servicesthat understand people are more successful2.Thecognitive offloadingeffect of AI will freeus up to be more thoughtful3.Innate humancuriosityto understand, relateand learn4.Brewingbacklashagainst the firehose of slopAI CAN DO MY JOB NOW(IT’S COMPLICATED)
Slide 38 as text
- Maybe?
- WHAT GIVES ME HOPE
- 1. Systemic incentives - products and services
- that understand people are more successful
- 2. The cognitive offloading effect of AI will free
- us up to be more thoughtful
- 3. Innate human curiosity to understand, relate
- and learn
- 4. Brewing backlash against the firehose of slop
- AI CAN DO MY JOB NOW(IT’S COMPLICATED)
-
Slide 39 / of 41
Maybe?WHAT I FEAR1.Bigger systemicincentivesto take away ourchoice and guide us to products2.Webecome more like AIas more of what weconsumes is AI3.Economiccollapsewill make this all irrelevantAI CAN DO MY JOB NOW(IT’S COMPLICATED)
Slide 39 as text
- Maybe?
- WHAT I FEAR
- 1. Bigger systemic incentives to take away our
- choice and guide us to products
- 2. We become more like AI as more of what we
- consumes is AI
- 3. Economic collapse will make this all irrelevant
- AI CAN DO MY JOB NOW(IT’S COMPLICATED)
-
Slide 40 / of 41
ux researchers?AI CAN DO MY JOB NOW(IT’S COMPLICATED)
Slide 40 as text
- Image: Cartoon crowd of many different people standing together under a rainbow. Image: Cartoon still captioned “Can you imagine a world without lawyers?”
- ux researchers?
- AI CAN DO MY JOB NOW(IT’S COMPLICATED)
-
Slide 41 / of 41
Thank youI’d love to hear your own hot takes or questions!Hit me up on LinkedIn or emailBillie-Mae.Kennedy@solpeople.com.au
Slide 41 as text
- Thank you
- I’d love to hear your own hot takes or questions!
- Hit me up on LinkedIn or email
- Billie-Mae.Kennedy@solpeople.com.au