Mel McVeigh

Love Letters on Cassettes Tapes - Spotify AI DJ Product Review

Mel McVeigh
Love Letters on Cassettes Tapes - Spotify AI DJ Product Review

Spotify AI DJ Product Feature Review

Hey Internet, I am your Spotify AI DJ, here to spin some Tuesday tunes.

Maybe it is just me… maybe I am officially moving to the contrarian luddite phase of my product career. 

I say, bring it on cos it is time product folks start to question more and more the why or what we do as opposed to the how or when based on frameworks, feature lists, growth loops and spreadsheets.

As a VP Product with many friends also working at Spotify, it is the one app I adore but increasingly adore less so. It feels bloated with so many features that do almost the same thing but slightly different. 

So before I start, this is just an observational conversation and musings. Not based on metrics, performance and OKRs and intentionally so. This is based on customer insight, behaviour and just some irksome feeling in my gut that it just doesn’t feel right.

(In more detail — this is based on the personal preferences as well as observations on how my peers, friends, network use the app and talk about the comparisons and usage of Apple + Amazing Music or even SoundCloud / Mixcloud and … so consider it guerilla user research, with a sprinkle of intuition to kickstart a conversation).

My product hat hunch is, that all these new personalisation features, especially those created with AI, are ideated around the need to drive engagement and Time-in-App. My guess also is it’s working at a metrics level for engagement as they are rolling out the feature in 50 countries.

So we will assume it is working for the business but…

Even so, it still makes me wonder what customer problem these features are solving and do I, as a customer really need them… and will I use the app more as a result.

So let’s begin with… the…

Why — the Customer Opportunity?

I am not sure what customer opportunity Spotify are solving for? Maybe to replicate the joy of radio, maybe to optimise cost of delivery and scale, maybe experiment with AI, Open AI and Voice. 

But maybe, there are asking the wrong question in their insights about what customers want. 

I get the impression it is less about the actual customer experience than optimising the hell out the app for engagement metrics, OKRs and targets, revenue, subscribers and churn with new features, hooks and reasons to come back. Optimising and experimenting their way out of actual true delight and customer innovation.

Why this feature?

So, I am late to this feature that I want to discuss - Spotify’s AI DJ, it only appeared in my app in December 2023 (but I did say also I am an VP Product — luddite / contrarian).

My first response when I saw this feature was WHY? And why am I so irked by it? I tried it for 2 mins and had to stop. I have not used it since.

Why this feature… borrowed from Spotify PR — A ‘stunningly realistic voice’ and ‘never before has listening felt so completely personal to each and every user’ or to maximise the tech from their purchase of Sonantic, a tech Voice AI start up to further personalise the listening experience.

And the customer need and opportunity we were solving for?

Let us think about these quotes for a second. 

As a customer, I want to listen to music + podcasts + mixes as albums, playlists, single tracks. I want to listen to ones that I choose, or that is recommended to me or has been made for me by my friends, family, colleagues, people I admire.

And definitely, I am more than willing for technology to recommend tracks for me based on my listening habits. AI for curation and recommendation done well is magical.

But, do I need an AI simulating a human voice to speak to me, just like a mate who knows what I like? Do I need to listen to a ‘stunningly realistic voice’ or do I want to listen to a human voice? If all these Spotify editors are in the background pulling together content for the AI DJ, why not just have a human do it?

And do I need a ‘never before has listening felt so completely personal to each and every user’ that is unique to me that I can’t share the experience, I can’t be part of a community, I can’t comment. That is not personal, it is just me and my machine riffing together, alone. 

As product people, we need to ask yourselves that just cos we can make it a product feature with AI Voice, do we need to have a synthetic voice replace that what humans used to do, to pretend and mimic and potentially replace the connection and storytelling human connections provide. 

Yup, cos as a customer that is what I asked for! NOT!!! (note sense of irony)

I am a big fan of music and audio personalisation. Personalisation when done well with a sense of curation can be poetic, useful and unexpected. I love my Discover Weekly and my ever changing Daylist. I love the recommendations each week and hour for new music. 

The AI DJ is basically doing the same thing with an extra element — AI Voice. Personally, the is a leap too far for me. This is what i mean by the same features optimised and tweaked for performance, and incrementally adjustment without actually creating something new.

As a result, my Spotify engagement has gone down as have many people I have spoken to. It was a jolt and a shock, and a really desire to focus on my listening experience that way I used to love it with the reasons it gave me joy and where possible removed all element of AI running voice interference.

Cos you know, what I love more… is the personal playlists people make me or I make them. I love mashing up tracks at a party with people competing to ensure the dance floor stays full with their own tracks, scouring the playlists for just the right tune. I love my friends recommending podcasts to listen to and even recently, discovering new podcasts and mediation lists based on visiting a random wellness centre in Bali. 

Each moment, unexpected serendipitous and woven into a story and life experience I can share — with other humans but augmented by great technology.

That is a level of personalisation AI can’t replicate, nor should it. I have never shared an AI playlist, I have never played an AI playlist with my friends at a party cos… while I love AI done well…

I want AI to augment my life but I sure as hell do not want it to replace it.

Community and conversation is a blend of curation and personalisation wrapped into one. Whether you are a DJ, just massively into music or simply like listening in the background. It is the simple act of sharing something that matters to you with someone else or storing it for your repeat pleasure when in the mood. A personal gift just for me or a gift I make for me and my friends.

A gift that is shared, discussed, collaborated, remixed and replayed. Like cassette tapes / mix tapes wrapped in envelopes and sent in the mail with long letters explaining all the tracks. Or songs sent to you each week to add to you own playlist. Or a DJ making you a personalised playlist as a gesture not just sharing one of the many he made.

Spotify’s AI DJ is a Solo Human to Big Tech paradigm like many AI agents. It makes you think you have connection but it has removed the human element of that connection. Is this what the future holds as product people and customers, is this what we want?

So, as a product person, back to the customer problem, we are solving for. Do we need a 1–2–1 relationship with technology who knows too much about us and talks to us in dulcet creepy tones… or do we need technology to create environments for connection, community, collaboration conversation with my friends, family, peers, communities so I can chat, learn, discuss and explore.

So Spotify friends, how about some features that augment human behaviour and connection as opposed to creating features that could replace it?

I would love to see features that enhance the current shared playlists. I would love to be able to see what others are listening to, not just the rabbit hole of my own unique personalised lists and lists and lists based on previous listening behaviour. I want to be able to connect with more people and share cool songs within my network.

Why?

Because that is what I do already, and I would love tech to make it a little bit easier and a little bit more convenient.

To give some context and in the spirit of sharing… these are some of the the amazing playlists and remixes made for me in 2023… why they were made and the impact they had and why… we need more of this in 2024 and beyond in the Year of AI… 2024 needs to be the return of human connection and less mediation of life via screens.

My 2023 Top Wrapped…

Made by humans, Shared by humans, Remixed by humans.

TIM’S PLAYLIST 2023

Tim is an guy I used to work with. For 3 years, every Sunday night I get a track or two in my inbox. Links to songs he likes that I would never know about. This was to play, confuse, train and delight my algorithm with a little serendipity. By adding those tracks to my music listening experience taught the algorithm to seek new styles. My Discover Weekly has forever changed based on these tracks.

Nothing beats Tim’s continuous sending of tracks. I look forward to them every week with anticipation.

German rock — who knew I would like it so much.

Tim’s Hack the Algo 2022 + Tim 2021 playlists.

Feature Idea: So how about a feature that allows you to create ongoing playlists with friends and then for AI to recommend within that space of collaboration. Create the behaviour the encourages AI to learn… and adapt rather than just look at previous listening habits.

BERN’S PLAYLIST 1988

In 1988, I was teenager, my cousin and I used to write letters to each other. Mine were full of teenage angst, his were full of playlists and music. Lovingly made on cassette tapes and sent in the mail. Recently he found an old letter with the mixtape list. 

The cassettes are long gone — which saddens me greatly but the one playlist lives one.

Feature Idea: So instead of creating your own time capsule for 2025 — yes a new feature that came up on my app yesterday. How about a feature where I can create a playlist for someone that is time bound, they get a track a week or they can to experience it in a certain order. With a companion letter from your friend.

the original letters / mixtape list - circa 1988

IMPORTANT

My son made me this playlist. I thought it was just his tracks for himself that he played and added me to his playlist. At XMAS, he told me he made it for me.

Feature Idea: None really, this is what Spotify does already pretty well.

DiscoPrints — the modern take on the mix tape

And finally when playlists meet print. Austin, another mate of mine, decided to use the Spotify API to create Discoprints of Playlists. Prints of albums of your fave list or tracks. From A4 — A1 size printed lovingly as gifts.

I made three as XMAS presents this year.

Feature Idea: Integrating into the shop of not just artist merch but optimising Spotify Playlists.

So, Spotify, thank you for your AI DJ but I prefer the real deal. 

  • Just cos you can build a feature doesn’t mean you should

  • Augment not automate

Here’s to hoping for some more human to human features within the Spotify App coming soon.

THANKS FOR READING