Oxford grad Arif explains how Causal uses synthetic personas to optimize retail discounts and discusses AI's impact on employment.
Summary
Arif, founder of Causal, explains how his platform helps retailers identify which customers require discounts to purchase versus those who would pay full price, noting that approximately half of discount spend is wasted. He details the use of anonymized transaction data to create synthetic customer personas and simulate counterfactual outcomes without violating GDPR. Arif shares the origin story of the startup, which pivoted from subscription churn reduction to discount optimization after recognizing the significant impact on profit margins. He advises against raising venture capital unless necessary, citing punitive terms and the difficulty of pitching complex AI models to defensive finance directors. Additionally, Arif expresses alarmist views on AI, predicting that most people will not have jobs in 10 years and discussing the degradation of learning in education systems.
Key takeaways
- 1 Causal helps retailers identify which customers require discounts to purchase versus those who would pay full price, noting that approximately half of discount spend is wasted on customers who would have bought anyway.
Unidentified speaker (speaker_1) · 1:34
Episode sections
- 0:00 – 1:19
Introduction and Background
The host introduces Arif, a 24-year-old Oxford graduate who ranked first in his class of 261 students. Arif discusses his background in history and his previous career in management consulting before founding his business.
- 1:19 – 2:27
Causal Product and Problem Statement
Arif explains that Causal helps retailers identify which customers require discounts to purchase versus those who would pay full price. He notes that approximately half of discount spend is wasted on customers who would have bought anyway.
- 2:27 – 6:44
Methodology: Synthetic Personas and Simulation
Causal uses anonymized transaction data (basket size, frequency) to create synthetic customer personas. By simulating a counterfactual universe where the customer did not receive an offer, the system predicts purchase behavior without violating GDPR.
- 6:44 – 8:35
Origin Story: From University Analytics to Cancer Research Math
The idea evolved from a university course matching tool to leveraging mathematical concepts from cancer research. Arif’s co-founder, a Cambridge mathematician, applied simulated environments used for personalized cancer treatments to retail discount optimization.
- 8:35 – 10:44
Pivot to Discounts and Subscription Churn
The startup initially focused on subscription churn reduction before pivoting to discounts, a shift influenced by the host's father. Arif highlights that discounts impact profit margins significantly more than subscription retention for companies with thin margins.
- 10:44 – 13:36
Early Hustle and Customer Acquisition
Arif describes the early days of registering the company and cold emailing. He cites a story of a Deliveroo employee who gained a CEO meeting by physically delivering a package to the office, emphasizing the need for aggressive, unconventional outreach.
- 13:36 – 19:28
Fundraising and Pitching Challenges
Arif advises against raising venture capital unless necessary, citing punitive terms. He describes the difficulty of pitching complex AI models to finance directors who are defensive about cost-saving measures, while customer-facing teams fear increased workload.
- 19:28 – 23:23
AI Impact on Employment and Education
Arif expresses alarmist views on AI, predicting that most people will not have jobs in 10 years. He discusses how AI is already replacing 90% of his daily tasks and raises concerns about the degradation of learning in education systems where teachers and students both use AI.
- 23:23 – 23:57
AI content saturation and truth verification
Speaker_1 discusses the difficulty of distinguishing truth from AI-generated content on social media, questioning whether truth matters when bad AI content is easily spotted but good AI content is not.
- 23:57 – 24:15
Causal product mechanics and origin
Speaker_0 expresses surprise at Causal's use of synthetic personas and mathematical simulations, asking how the concept was developed.
- 24:15 – 24:42
Startup idea iteration and adaptation
Speaker_1 explains that startup ideas typically change significantly in the first six to eight months because the initial concept rarely fits market demand perfectly, requiring adaptation.
Claims worth verifying
Arif graduated number one out of 261 people in his History class at Oxford.
Unidentified speaker (speaker_0) ·0:28
How to verify: Oxford University graduation records or Arif's LinkedIn profile. For recent events, verify 2026-08-04 through 2026-11-04. Use another period only when the speaker explicitly names it.
Approximately half of retail discount spend is wasted on customers who would have paid full price.
Unidentified speaker (speaker_0) ·1:34
How to verify: Academic papers on retail pricing elasticity or Causal's published case studies/whitepapers. For recent events, verify 2026-08-04 through 2026-11-04. Use another period only when the speaker explicitly names it.
Causal uses anonymized data including basket size, purchase frequency, and recency to create synthetic personas.
Unidentified speaker (speaker_1) ·2:27
How to verify: Causal's privacy policy or technical documentation on their website. For recent events, verify 2026-08-04 through 2026-11-04. Use another period only when the speaker explicitly names it.
The concept of simulating customer behavior was adapted from cancer research mathematics.
Unidentified speaker (speaker_1) ·8:15
How to verify: Causal's website 'About' section or press releases detailing the co-founder's background. For recent events, verify 2026-08-04 through 2026-11-04. Use another period only when the speaker explicitly names it.
Some companies spend 20% of their overall revenue on discounts.
Unidentified speaker (speaker_1) ·9:44
How to verify: Retail industry reports (e.g., McKinsey, BCG) on promotional spend as a percentage of revenue. For recent events, verify 2026-08-04 through 2026-11-04. Use another period only when the speaker explicitly names it.
Causal generated over £500,000 in revenue in nine months from a previous university analytics venture.
Unidentified speaker (speaker_0) ·17:43
How to verify: Companies House filings for the previous entity or Arif's LinkedIn profile. For recent events, verify 2026-08-04 through 2026-11-04. Use another period only when the speaker explicitly names it.
Primary sourceThe previous university analytics venture charged £6,000 per seat for a two-week summer school.
Unidentified speaker (speaker_1) ·18:11
How to verify: Marketing materials or archives of the summer school program. For recent events, verify 2026-08-04 through 2026-11-04. Use another period only when the speaker explicitly names it.
Arif outsources 90% of his daily tasks to AI agents.
Unidentified speaker (speaker_1) ·20:40
How to verify: Arif's public social media posts or interviews detailing his daily workflow. For recent events, verify 2026-08-04 through 2026-11-04. Use another period only when the speaker explicitly names it.
Arif predicts most people will not have jobs in 10 years due to AI.
Unidentified speaker (UNKNOWN) ·21:18
How to verify: Transcript of this podcast episode. For recent events, verify 2026-08-04 through 2026-11-04. Use another period only when the speaker explicitly names it.
Cold emails have an 80% spam landing rate and a 0.5% response rate.
Unidentified speaker (speaker_1) ·11:49
How to verify: Industry email marketing benchmarks (e.g., Mailchimp, HubSpot reports). For recent events, verify 2026-08-04 through 2026-11-04. Use another period only when the speaker explicitly names it.
Speaker_1 claims that much of the content on Instagram is AI-generated.
Unidentified speaker (UNKNOWN) ·23:33
How to verify: Review academic studies or industry reports on the prevalence of AI-generated media on Instagram during the verification period (2026-08-04 to 2026-11-04). For recent events, verify 2026-08-04 through 2026-11-04. Use another period only when the speaker explicitly names it.
Speaker_1 claims that bad AI content is noticeable, but good AI content is not.
Unidentified speaker (UNKNOWN) ·23:40
How to verify: Conduct a qualitative survey or experiment asking participants to identify AI-generated images/text on social media platforms to assess detection rates of high-quality vs. low-quality AI outputs. For recent events, verify 2026-08-04 through 2026-11-04. Use another period only when the speaker explicitly names it.
Speaker_1 claims that startup ideas change significantly in the first six to eight months.
Unidentified speaker (speaker_1) ·24:15
How to verify: Analyze longitudinal data from startup accelerators or pitch deck archives to track the frequency of pivot or iteration events within the first 8 months of a company's formation. For recent events, verify 2026-08-04 through 2026-11-04. Use another period only when the speaker explicitly names it.
Do your own research
- Verify Causal's customer base and case studies to assess the validity of the '50% wasted discount spend' claim.
- Review Causal's privacy policy and data handling procedures to confirm GDPR compliance and the use of synthetic personas.
- Analyze the financial impact of discount optimization on thin-margin retailers to validate the pivot rationale.
- Research venture capital term sheets to evaluate the 'punitive terms' claim and assess alternative funding options.
- Investigate independent studies on AI's impact on employment rates and educational outcomes to counter Arif's alarmist predictions.
Full transcriptexpand
- 0:00
- Unidentified speaker (speaker_0): So the background to this is kind of a bit funny because you slid into my dad's DMs on LinkedIn.
- 0:05
- Unidentified speaker (speaker_1): That's right.
- 0:05
- Unidentified speaker (speaker_0): Which I thought was quite funny. Um I'd my dad didn't know what sliding in meant, which was funny, even funnier. So you're 24 and you have your own business called Cosal.
- 0:16
- Unidentified speaker (speaker_1): Causal, yeah.
- 0:18
- Unidentified speaker (speaker_0): Right. Which we'll get on to, but first I kind of wanted to like set the scene about who you are. So you were you went to Oxford, you studied history.
- 0:28
- Unidentified speaker (speaker_0): When you were out there, you were like number one out of 261 people.
- 0:32
- Unidentified speaker (speaker_1): That's right, yeah.
- 0:33
- Unidentified speaker (speaker_0): And then you got involved and started to do like business thing. Well, what did you start doing? So
- 0:38
- Unidentified speaker (speaker_1): after that, I sort of took the traditional route of working in consulting. I think most people think about, especially from Oxford, think about careers. I'm gonna either become an investment banker, a lawyer, or a consultant, um. or
- 0:49
- Unidentified speaker (UNKNOWN): an
- 0:50
- Unidentified speaker (speaker_1): accountant. That's like the the bingo. Um, so I picked my one, which was consulting. So I did that for a couple of years and then I started my own business.
- 0:57
- Unidentified speaker (speaker_0): Nice. Yeah.
- 0:58
- Unidentified speaker (UNKNOWN): I'm
- 0:58
- Unidentified speaker (speaker_0): trying to think of the three forms of one the university I went went to. I think it was like worked at Google, worked at social content creator, waitress, not as long as quite as high eyebrow pro as that. Um that's really interesting that you kind of came up with this. So our I've been trying to wrap my head around COSL.
- 1:18
- Unidentified speaker (speaker_1): Yeah.
- 1:19
- Unidentified speaker (speaker_0): Could you explain it to me in really simple terms? Firstly, what how is this like a market in itself? So, what do you do?
- 1:27
- Unidentified speaker (speaker_1): Yeah, yeah. So, in essence, what we do is we realise that companies, especially retail companies, spend lots and lots of money on discounts.
- 1:34
- Unidentified speaker (speaker_0): Oh, yeah.
- 1:34
- Unidentified speaker (speaker_1): The problem with discounts is that lots of them go to people who are willing to pay. full price. Actually half of them roughly. So think about like Black Friday coming up, right? Like you're probably not buying the laptop you thought about because you're waiting or expecting it to go on discount.
- 1:49
- Unidentified speaker (speaker_0): How did you know?
- 1:52
- Unidentified speaker (speaker_1): So so for companies they spend lots and lots of money on these discounts and a portion of them is wasted or half of them is wasted on people who would be willing to pay full price. The problem that companies have is they don't know who is willing to pay full price and who isn't. They can't
- 2:09
- Unidentified speaker (UNKNOWN): do
- 2:09
- Unidentified speaker (speaker_1): this. So what we do is we we figure that out and we we tell them.
- 2:12
- Unidentified speaker (speaker_0): Right. How do you or under GDPR or whatever? You can't like just look under the bonnet anymore and say, right, this per Connie lives here and you know she loves a Black Friday deal, da da da, she'd be the right person. So how do you like go about
- 2:25
- Unidentified speaker (UNKNOWN): like
- 2:26
- Unidentified speaker (speaker_0): determining this?
- 2:27
- Unidentified speaker (speaker_1): Yeah, so so we work with an anonymized data. So we look at like the we don't look at sort of you as a person or where you live or who you are, but we look at sort of what you buy, um, the basket size, how frequently you bought, when's the last time you bought, um so these sort of things that are inherently data you you
- 2:47
- Unidentified speaker (speaker_1): we get or companies get as you work through or look through a website. Uh-huh. That's what we work with.
- 2:52
- Unidentified speaker (UNKNOWN): Right.
- 2:52
- Unidentified speaker (speaker_0): And then you go to like these retailers and are like.
- 2:56
- Unidentified speaker (speaker_1): So so they have they send out coupons, right? Like personalized coupons, and we say, okay, we can tell you which ones or for which customers we can withhold those coupons. The problem with that, and these customers who you withhold them, they'll pay full price. You're not gonna lose any order volumes from these people. The problem with that is actually very difficult to do because what you want to do is you
- 3:16
- Unidentified speaker (speaker_1): want to give an offer to someone who wouldn't buy um without an offer.
- 3:20
- Unidentified speaker (speaker_0): Okay.
- 3:21
- Unidentified speaker (speaker_1): But would buy with an offer. Whereas a lot of people buy with both. And the way you do that or you test that is you have to give someone both an offer and not an offer at the same time, right?
- 3:30
- Unidentified speaker (speaker_0): Oh.
- 3:31
- Unidentified speaker (speaker_1): And the problem is you can't do that because there's only one you and there's only one universe.
- 3:34
- Unidentified speaker (speaker_0): Yeah.
- 3:35
- Unidentified speaker (speaker_1): So what we do is we simulate another universe and we recreate you. through your data and then we give you the other side of the equation. So in the the real world you may have got an offer. In the simulated world, you don't have an offer, and then we compare the outcome and we say, okay, Connie, you are someone who reacts in this way, and therefore we should send you an offer.
- 3:54
- Unidentified speaker (speaker_0): Really? Yeah. That's pretty cool.
- 3:57
- Unidentified speaker (UNKNOWN): That's
- 3:57
- Unidentified speaker (speaker_1): what all we do, yeah.
- 3:58
- Unidentified speaker (speaker_0): But how do you do that, right? I'm not being funny, but how do you do that? Like, is it is it maths? Is it is it anthropology? Do you s do you f like is
- 4:07
- Unidentified speaker (UNKNOWN): it
- 4:07
- Unidentified speaker (speaker_0): different countries tend to
- 4:09
- Unidentified speaker (UNKNOWN): love
- 4:09
- Unidentified speaker (speaker_0): a deal? Yeah,
- 4:10
- Unidentified speaker (speaker_1): exactly. It's an interesting thing. Like, how do you build a model or or more generally, how do you simulate human behavior? Yeah.
- 4:15
- Unidentified speaker (speaker_0): Or
- 4:16
- Unidentified speaker (speaker_1): how can you create a a synthetic person that reacts and behaves in the same way as a normal person?
- 4:21
- Unidentified speaker (speaker_0): Yeah.
- 4:21
- Unidentified speaker (speaker_1): And how can you put that into an AI agent and how do you train that, right? So
- 4:26
- Unidentified speaker (speaker_1): typical like foundation, like your claws or your open AIs, they would train it on um lots and lots of techs. So basically the whole internet of techs, and then we'll use that to correlate. Um so what we do is we look at
- 4:39
- Unidentified speaker (speaker_1): we train it based on companies sort of own past transactions. And we say, Okay, here are your types of customers and here's how they move. So if you're selling, I don't know, specialty coffee, it's like oh there are these people who buy
- 4:55
- Unidentified speaker (speaker_1): large amounts of coffee as their first order, then they don't buy again, and they prefer your Guyana coffee or whatever. It's this segment in our simulated environment that isn't
- 5:07
- Unidentified speaker (UNKNOWN): that
- 5:07
- Unidentified speaker (speaker_1): is responsive to an offer and wouldn't pay without an offer. Or these are the ones you just should send offers to, and from this, we look at sort of your passion. past transaction history and say, okay, you're someone who fits this bill, therefore you should get an offer.
- 5:19
- Unidentified speaker (UNKNOWN): God,
- 5:19
- Unidentified speaker (speaker_0): that's really interesting. But is there I'm not being funny, and but like, is there anyone that wouldn't accept a discount? Because for me, I can't who are the people that wouldn't expect a discount, accept a discount?
- 5:30
- Unidentified speaker (speaker_1): So it's not that they would, it's the problem isn't that they wouldn't accept a discount. They would accept a the discount, and that's the problem. Right. The thing you're trying to solve is these are the people who'd accept, who'd buy. both with and without a discount.
- 5:40
- Unidentified speaker (speaker_0): Fine.
- 5:41
- Unidentified speaker (speaker_1): And in theory, as a company, you don't want to give these people discounts.
- 5:44
- Unidentified speaker (speaker_0): No. Is it rich people?
- 5:46
- Unidentified speaker (speaker_1): It's yeah.
- 5:47
- Unidentified speaker (speaker_0): That's what I would assume. I would assume that you would be like rich people would target.
- 5:53
- Unidentified speaker (UNKNOWN): Yeah,
- 5:53
- Unidentified speaker (speaker_1): it's it's it's rich people, it's people who are less sensitive to discounts, um, generally wealthier people, right? So, like the best example is like delivery. And delivery, you may get like the app notification which has like seven pounds off if you use this code.
- 6:06
- Unidentified speaker (speaker_0): Yeah.
- 6:07
- Unidentified speaker (speaker_1): And they they use a similar idea. And they
- 6:10
- Unidentified speaker (speaker_1): do, yeah. So I get discounts because it knows I need the discount. But I'm sure if you're someone who works in a in a company and your company's paying for all of your expenses as your dinners if you're staying past 8 p.m.,
- 6:22
- Unidentified speaker (UNKNOWN): they
- 6:23
- Unidentified speaker (speaker_1): don't get a discount.
- 6:24
- Unidentified speaker (speaker_0): No.
- 6:24
- Unidentified speaker (speaker_1): Because they know like that person is buying regardless. of the discount.
- 6:28
- Unidentified speaker (speaker_0): That's really interesting.
- 6:30
- Unidentified speaker (speaker_1): Yeah.
- 6:30
- Unidentified speaker (speaker_0): Gosh.
- 6:31
- Unidentified speaker (UNKNOWN): I
- 6:31
- Unidentified speaker (speaker_0): wonder what category I'd fit into.
- 6:34
- Unidentified speaker (UNKNOWN): Oh,
- 6:34
- Unidentified speaker (speaker_1): use causal.
- 6:35
- Unidentified speaker (speaker_0): We'll use causal
- 6:36
- Unidentified speaker (UNKNOWN): and I'll find out.
- 6:37
- Unidentified speaker (speaker_0): So take me back to like day one, right? So you you come up with this idea.
- 6:42
- Unidentified speaker (speaker_1): Kind of.
- 6:43
- Unidentified speaker (speaker_0): Kind of okay.
- 6:44
- Unidentified speaker (speaker_1): Kind of. So the idea didn't start like this. So the idea sort of evolved. If you ask me what the idea was. four months ago or even when I quit my job,
- 6:54
- Unidentified speaker (UNKNOWN): it
- 6:54
- Unidentified speaker (speaker_1): was very different. So it started out as university analytics. The idea was to help students find the best matching university course.
- 7:02
- Unidentified speaker (speaker_1): So it was like based on your
- 7:05
- Unidentified speaker (speaker_1): profile or the things you like as subjects, your GCSCs, your A levels, or whatever, which courses can we find that are good for you? And the idea here was that the system. is biased towards rich people because they have the money for tutors, they have the advice, they go to better schools, and therefore they surface the courses that are right or good fits for them. Whereas
- 7:25
- Unidentified speaker (speaker_1): what we can do is we can actually determine probabilistically your likelihood of getting into this course and on top of that, the how likely you you'll will enjoy it. And therefore we'd surface courses um that someone otherwise wouldn't have thought of.
- 7:39
- Unidentified speaker (speaker_1): So that was what the idea was a while ago. And it worked, I went to some schools. They they they they use it and then we and then I was like okay I need to find someone I need I need a technical person.
- 7:51
- Unidentified speaker (speaker_0): Right. I
- 7:51
- Unidentified speaker (speaker_1): need someone to really build this and really make the algorithm really good.
- 7:54
- Unidentified speaker (speaker_0): I'll throw my hat in the ring.
- 7:57
- Unidentified speaker (speaker_1): So I reached out I reached out to an ol an old friend of mine, uh he's at Cambridge, he's a mathematician he's a postdoc there, he's an AI researcher. And he was interested and he joined, but his sort of skill set was a bit too let's say good for this problem.
- 8:14
- Unidentified speaker (speaker_0): Right.
- 8:15
- Unidentified speaker (speaker_1): Like he was so his background was in creating these simulated environments for cancer treatments.
- 8:19
- Unidentified speaker (speaker_0): Wow.
- 8:19
- Unidentified speaker (UNKNOWN): Where
- 8:19
- Unidentified speaker (speaker_1): it's like two people could have the same cancer, but
- 8:23
- Unidentified speaker (UNKNOWN): it
- 8:23
- Unidentified speaker (speaker_1): doesn't mean that they should have the same treatment because everybody is unique. And therefore, how can you simulate what the treatment should be based on AI and based on data so that you could give the right treatment to the right person?
- 8:34
- Unidentified speaker (speaker_0): Wow.
- 8:35
- Unidentified speaker (speaker_1): So
- 8:36
- Unidentified speaker (UNKNOWN): he
- 8:36
- Unidentified speaker (speaker_1): took that idea, we took that sort of mathematical sort of concept and we applied it. Can we apply this elsewhere? Um he's still doing the cancer research. Um but um can and we sort of moved around and we started with subscriptions. Can we help companies out? identify which people are churning and send targeted offers to stop them from churning. And then
- 8:56
- Unidentified speaker (speaker_1): that kind of worked and then we moved towards discounts thanks to in in no small part by your dad who really pushed this idea where
- 9:03
- Unidentified speaker (speaker_0): oh hold on dad I didn't know that.
- 9:07
- Unidentified speaker (speaker_1): But
- 9:07
- Unidentified speaker (speaker_0): that's the interesting thing with the subscription because I I I like told myself I'm only allowed one one like in terms of television subscriptions that I have, basically because Mick and I did another an episode where I went through my bank statement and it was horrible and I was spending so much money on subscriptions that I had to really cut out.
- 9:24
- Unidentified speaker (UNKNOWN): Yeah.
- 9:24
- Unidentified speaker (speaker_0): But I think that's something that people always like, we'll go back to television. But discounts is quite an interesting like concept, I think. And I didn't realise how like linked this is to businesses they always offer dis usually always offer some sort of incentive or discounts.
- 9:44
- Unidentified speaker (speaker_1): Yeah, and for some companies it's huge. Like for some companies they spend twenty percent of their overall revenue on discounts. They probably make ten percent what's called profit off their you know, think of like fast fashion. Yeah. They sell lots, it's cheap, it's very discounted and the margins are very small. Yeah. Or grocery stores. So if you could say to a company we're gonna cut
- 10:04
- Unidentified speaker (speaker_1): your discount spend by twenty, thirty, forty percent, and we're not gonna change the order volume, then that impact on the pro on the profit is much, much larger, right? Like you could double potentially uh a company's profit. Whereas if we whereas when we're focusing on subscriptions, it's like, yeah, maybe we can make sure make make people retain their subscriptions, but and
- 10:24
- Unidentified speaker (speaker_1): it's less compelling.
- 10:25
- Unidentified speaker (UNKNOWN): No,
- 10:26
- Unidentified speaker (speaker_0): I think it is less compelling, to be honest. Well, I would say that. Right. So what right, so you have this I so you have this idea what do you do next? Because when people say to me, Well I started a business, like I would just imagine you sit there with a laptop and coffee. Like what do you actually do on like day one? Which I I'd
- 10:43
- Unidentified speaker (UNKNOWN): be quite
- 10:43
- Unidentified speaker (speaker_0): intrigued with.
- 10:44
- Unidentified speaker (speaker_1): So first day I was like we were like let's get registered. So I was like uh going on company house company's house filing, doing all of that boring stuff. But it makes it feel official, right? There's a occasion to it. And I think when you start a business, you need two things, right? You need something to sell and someone to sell it to. So like if you have like so what is the thing that you want
- 11:04
- Unidentified speaker (speaker_1): to sell? And then the harder part is how do you sell it? So then like you want to make a website, you want to make some emails, like an e an email address, um, and then you just need to start reaching out to people.
- 11:15
- Unidentified speaker (speaker_1): having conversations. Hey, I've started this thing, it could be useful to you. I don't know if it is. Are you free for a chat? And most people don't respond.
- 11:26
- Unidentified speaker (UNKNOWN): Um
- 11:26
- Unidentified speaker (speaker_1): very few people respond. Like um you so some people do, your dad has responded, other people respond, and then it goes from there. They know people who know people, and then slowly if the idea is good and if you're selling something that people think is potentially valid. valuable, that will spread and it will spread faster. If you're if it's not a very good idea, then people won't respond.
- 11:46
- Unidentified speaker (speaker_0): How do you get your first customer when no one's ever heard of EA?
- 11:49
- Unidentified speaker (speaker_1): Um so the first customer for us was um a cold email. Um, and they say cold called emails are pro like generally very hard because no one responds to a cold called email and it lands. in spam eighty percent of the time. But you're working with very low percentages. It's like you're l working with like half a percent response rate.
- 12:10
- Unidentified speaker (speaker_0): Yeah.
- 12:10
- Unidentified speaker (UNKNOWN): So
- 12:10
- Unidentified speaker (speaker_1): then it's like, okay, if I just send ten thousand, then someone will respond.
- 12:14
- Unidentified speaker (speaker_0): Yeah.
- 12:15
- Unidentified speaker (speaker_1): Um so there's that. But like they say like you should do things that
- 12:19
- Unidentified speaker (UNKNOWN): it's
- 12:19
- Unidentified speaker (speaker_1): very hustly, like there's no
- 12:21
- Unidentified speaker (UNKNOWN): way around
- 12:21
- Unidentified speaker (speaker_1): it. Like it's very like um so I heard the story of of of this guy who was uh um selling Deliveroo shopping. So he built the the shopping element within Deliveroo. And he was like, Yeah, we wanted to sell to Oliver bonus or some clients. Um and they were like, um, how can we get to the
- 12:41
- Unidentified speaker (speaker_1): CEO to have that conversation? So he just got a Deliveroo delivery bag, put items from that company in it, just rocked up to the to the
- 12:49
- Unidentified speaker (UNKNOWN): office.
- 12:49
- Unidentified speaker (speaker_1): No invitation, no nothing. He says, I have a delivery roo for the CEO. Like, yeah, can we take a given this? No, I need the code, so I need to go to him personally. Then he went, he goes to find the CEO one-on-one, and he he takes the item out of the box and he's
- 13:02
- Unidentified speaker (UNKNOWN): like,
- 13:02
- Unidentified speaker (speaker_1): This is what the delivery partnership could look like with you.
- 13:05
- Unidentified speaker (speaker_0): Wow. And
- 13:05
- Unidentified speaker (speaker_1): then he got the meeting and then it worked. So it's like um that kind of thing. It's not fun. Some people love it, some people just love that, like that. I suppose that adrenaline. you get from just rocking up to an office unannounced, seeing what happens. Um I don't love it, not gonna lie, but
- 13:21
- Unidentified speaker (UNKNOWN): But
- 13:22
- Unidentified speaker (speaker_0): you must have so you're quite in a fortunate position because you've got these friends, very academic, very clever friends, so you trust them, right? You trust this mathematician, yeah, etc. etc. And then you do your cold reaching out and blah blah blah. How do you start then raising money for it?
- 13:36
- Unidentified speaker (speaker_1): Yeah, so I think raising money is a whole is
- 13:38
- Unidentified speaker (UNKNOWN): a whole
- 13:39
- Unidentified speaker (speaker_1): different ecosystem.
- 13:40
- Unidentified speaker (speaker_0): Right.
- 13:41
- Unidentified speaker (speaker_1): I think don't raise
- 13:43
- Unidentified speaker (UNKNOWN): like
- 13:43
- Unidentified speaker (speaker_1): money if the if you don't have to, because they could be quite um punitive, that's and quite quite mean to you. So if you want to raise uh from from venture capitalists, um there's a lot of them out there. Um and they they go off signals and vibes.
- 13:59
- Unidentified speaker (UNKNOWN): Oh.
- 14:00
- Unidentified speaker (speaker_1): Because they don't really know, like they don't really understand. deep down how it works. And there's a lot of people who could say a lot of things but don't actually like back it up. So how if you're um in a VC, how do you differentiate between those two people? But it's similar, you have to outreach, you have to go to events, you have to meet other founders, you have to sort of build out a network, and hopefully
- 14:19
- Unidentified speaker (speaker_1): someone um will will We'll meet you and then you have a meeting and then you have to convince them to give you a lot of money.
- 14:25
- Unidentified speaker (speaker_0): Gosh, what's the hardest moment so far?
- 14:28
- Unidentified speaker (speaker_1): The hardest moments are the pitches like with with to clients or to customers that like take a lot of time to build up and to get towards. And then you go in and you have your 30 minutes or your hour, and then it sounds great. And you people are very like can be very British about these things where it's like they don't like it, but to your face it's like the best thing.
- 14:48
- Unidentified speaker (UNKNOWN): Oh yeah, it's brilliant.
- 14:49
- Unidentified speaker (speaker_1): Brilliant, like fantastic. And um then they and then you have it and then you leave the meeting, it's like this great, we got this, this is huge for us, and then a week and a half later it's like there's some excuse. Like though now's not the right time, or we have X going on, we have Y going on, so
- 15:04
- Unidentified speaker (speaker_1): let's let's touch base later, you know. So so that's that's the sort of up and down because you have the thrill of thinking you did very well. There's also all the time and preparation that goes into that.
- 15:13
- Unidentified speaker (speaker_0): Uh-huh.
- 15:14
- Unidentified speaker (speaker_1): And then the the failure, right?
- 15:16
- Unidentified speaker (speaker_0): God. Okay.
- 15:19
- Unidentified speaker (speaker_0): When you do your pictures, how do you sort of like illustrate your idea? Because it's quite complicated, right? I'm still trying to wrap my head around it. And how do you go about that?
- 15:27
- Unidentified speaker (speaker_1): Yeah, it's hard. Um I think the the hard part is it's quite complicated and they sort of want to know how it works under the hood, but they sort of don't want to know.
- 15:36
- Unidentified speaker (speaker_0): Yeah, I
- 15:36
- Unidentified speaker (UNKNOWN): know that
- 15:37
- Unidentified speaker (speaker_0): feeling.
- 15:37
- Unidentified speaker (speaker_1): Because it's like, yeah, so how does how do you actually do it? And it's like there's a simulated thing and that sometimes works, but
- 15:43
- Unidentified speaker (UNKNOWN): um
- 15:45
- Unidentified speaker (speaker_1): or you go very deep in the math and they just don't don't don't want to hear that. So it's like how do you convince them that we could actually do what we say we can? Yeah. Given that we're new, we don't have any case studies.
- 15:56
- Unidentified speaker (speaker_0): Yeah.
- 15:57
- Unidentified speaker (speaker_1): Um, we don't have like
- 16:00
- Unidentified speaker (UNKNOWN): um
- 16:00
- Unidentified speaker (speaker_1): there's very little proof or credibility to go on, right? So you have to sort of put it in in their terms and ideally it's addressing a problem they already have.
- 16:12
- Unidentified speaker (speaker_0): Yeah.
- 16:12
- Unidentified speaker (speaker_1): So it's like if you're going to a company, like and they don't discount, then it's not gonna work. But like if you're putting if you're going to a company and they're discounting a lot and it's that's already on their agenda that they need to cut discounts,
- 16:24
- Unidentified speaker (speaker_0): uh-huh,
- 16:25
- Unidentified speaker (speaker_1): then it's like, okay, you could do it in internally as your team and it'll take a lot of time. Or you could work with us.
- 16:32
- Unidentified speaker (UNKNOWN): Uh
- 16:32
- Unidentified speaker (speaker_0): huh.
- 16:32
- Unidentified speaker (speaker_1): We are hopefully some clever people who could probably do a better job and we have this cool way of doing it. I don't really understand how it works. But I s see uh sounds like they know what they're doing.
- 16:41
- Unidentified speaker (UNKNOWN): Um,
- 16:41
- Unidentified speaker (speaker_1): let's give them a shot. And that's sort of the the way. Um but
- 16:46
- Unidentified speaker (speaker_0): I can imagine this is quite difficult because I like can imagine if you go into a financial finance directors, they're quite defensive.
- 16:51
- Unidentified speaker (speaker_1): Yeah.
- 16:52
- Unidentified speaker (UNKNOWN): I
- 16:52
- Unidentified speaker (speaker_0): can imagine that they're like, well, why wouldn't I I know what I'm doing and I know how to cost save and da da da. And then I feel like who are you saying this to really?
- 17:01
- Unidentified speaker (speaker_1): So the finance directors are more receptive for for us specifically, because
- 17:06
- Unidentified speaker (UNKNOWN): it's
- 17:06
- Unidentified speaker (speaker_1): like money saved, you know.
- 17:07
- Unidentified speaker (UNKNOWN): Oh, I thought they'd be
- 17:08
- Unidentified speaker (speaker_0): defensive. Oh
- 17:09
- Unidentified speaker (speaker_1): Um and in theory it sounds great. The people that actually have to use it so the like the the people that manage the customers,
- 17:18
- Unidentified speaker (speaker_0): uh-huh.
- 17:19
- Unidentified speaker (UNKNOWN): Uh
- 17:19
- Unidentified speaker (speaker_1): or the people that send out the offers, like these people, they they see this and it's like
- 17:24
- Unidentified speaker (UNKNOWN): like
- 17:25
- Unidentified speaker (speaker_1): part of our thinking is like, yeah, this this means I have now have more work to do.
- 17:28
- Unidentified speaker (speaker_0): Yeah.
- 17:29
- Unidentified speaker (speaker_1): Right? Like they at the end of the day it's not in their sort of like
- 17:34
- Unidentified speaker (speaker_1): they're not thinking about how I can save the company
- 17:37
- Unidentified speaker (UNKNOWN): s
- 17:38
- Unidentified speaker (speaker_1): X amount. of money. They're thinking about how can I grow my sales. So that's that's a that's a challenge.
- 17:43
- Unidentified speaker (speaker_0): Your previous business report you made over
- 17:47
- Unidentified speaker (speaker_0): if I'm reading this right, over 500,000 in revenue in nine months. How did you achieve that?
- 17:53
- Unidentified speaker (UNKNOWN): So
- 17:54
- Unidentified speaker (speaker_1): with that, that was um the university analytics thing, but in this earlier version, it was a summer school with some data in it. So it was a summer school where people came around the world um to Oxford for two weeks.
- 18:07
- Unidentified speaker (speaker_0): Oh cool.
- 18:08
- Unidentified speaker (speaker_1): And um it was like
- 18:11
- Unidentified speaker (speaker_1): six grand a seat.
- 18:12
- Unidentified speaker (speaker_0): Wow.
- 18:13
- Unidentified speaker (speaker_1): And that we hit like a hundred people, so that's how how we did it. And that was that was like we were selling a very commoditized, like it wasn't an AI anything. It was literally like seats in a university for two weeks, basically a glorified holiday.
- 18:27
- Unidentified speaker (speaker_1): So how can you convince parents to spend a lot of money to send their kids from all around the world to a company with no credibility, has never done this before, no photos?
- 18:36
- Unidentified speaker (UNKNOWN): You
- 18:36
- Unidentified speaker (speaker_1): just have to get calling. Like you go, we we traveled around the world to different schools. So we went to I went to the Cayman Islands, the US,
- 18:44
- Unidentified speaker (speaker_1): Hong Kong, Vietnam, and Thailand. And then we would just say, like, hey, like, can we do a presentation?
- 18:51
- Unidentified speaker (speaker_1): to your school about applying to the UK? Oh. And because we were like on behalf of like because we were at Oxford, we could say that we we could help with this. Yeah. And that because they like the fact that we flew in.
- 19:05
- Unidentified speaker (UNKNOWN): So
- 19:05
- Unidentified speaker (speaker_1): then they're like, oh well, these guys are here, we might as well run quickly get an assembly together. And then you just go and you present, and then some of the some of the kids bring it to the parents and some of the parents sign them up.
- 19:14
- Unidentified speaker (speaker_0): Wow. So obviously AI is very prevalent at the moment. I'd be inter and you obviously have your consultancy background as well. I'd be interested to hear how at Oxford what their ideas of like AI, what your personal ideas, AI and tech.
- 19:28
- Unidentified speaker (speaker_1): Yeah, I think it's it's quite polarizing. I think there's a lot of fear currently.
- 19:32
- Unidentified speaker (speaker_0): Yeah.
- 19:32
- Unidentified speaker (speaker_1): About like, is it gonna kill us all? And the sort of heads of these AI labs are sort of fueling this fire. Um, I think my opinion, and sort of let's say other people at Oxford, I think it's very split. My personal opinion is like, I think we could all die, yeah, sure. Like, I could so see that happening.
- 19:52
- Unidentified speaker (speaker_1): The
- 19:53
- Unidentified speaker (UNKNOWN): I
- 19:53
- Unidentified speaker (speaker_1): think there's this idea of like, how clever are these AI agents at the moment? Like, is it real
- 20:00
- Unidentified speaker (UNKNOWN): like
- 20:00
- Unidentified speaker (speaker_1): thinking? Is it really actually intelligence? And I would say that it is, right? Like, there's this like hunt for AGI or like artificial general intelligence. But like, when does something become intelligent? Is I think like almost a philosophical question. Because like the power of these agents, if you're um good at using it, even
- 20:20
- Unidentified speaker (speaker_1): now, is so much more like I could like 10x my productivity. Um, and therefore it can. do so much more than a team of say analysts or or and for that reason it's like wow this is so much smarter than me. Like I'm I to a large extent I'm outsourcing most of my at
- 20:40
- Unidentified speaker (speaker_1): least 90% of everything I do to it in terms of daily tasks sending out emails. The only thing I do is check. So I'm in essence a claw claude man a claud manager and I just make manage lots and lots of different agents that just pings me as things get done. And I just review and send. And for it to do all of these tasks
- 21:00
- Unidentified speaker (speaker_1): automated is so so helpful and very very worrying because it's almost it's already smarter than I think most people. Like I think it's smarter than any individual, to be honest. Um, And what that means, I think the big what worry is that means for hiring. And I don't, and I think
- 21:18
- Unidentified speaker (UNKNOWN): in
- 21:18
- Unidentified speaker (speaker_1): 10 years most people won't have jobs, to be honest. Like, I I I'm quite maybe a bit alarmist on this, but I also think it means
- 21:28
- Unidentified speaker (UNKNOWN): um
- 21:29
- Unidentified speaker (UNKNOWN): what
- 21:30
- Unidentified speaker (speaker_1): what do you do in a society that doesn't need jobs?
- 21:34
- Unidentified speaker (speaker_0): Well, I don't know, I don't know what if we do in society that don't need jobs, but I feel like it will be like survival of the fittest, really, in terms of like who ends up getting a job because of AI will replace quite a lot of jobs. There's some jobs that AI will never be able to do. I find myself not thinking as much
- 21:51
- Unidentified speaker (speaker_0): when I use AI and I won't I didn't grow up with AI, so I'm wondering like all these like kids at school, if they don't understand essays and on whatever, and they would just put into the chat GBT, write me your essay on history, blah blah blah blah.
- 22:03
- Unidentified speaker (UNKNOWN): Yeah,
- 22:03
- Unidentified speaker (speaker_1): it's a big problem.
- 22:04
- Unidentified speaker (UNKNOWN): Will it
- 22:04
- Unidentified speaker (speaker_0): like dumb people down? I
- 22:06
- Unidentified speaker (UNKNOWN): think
- 22:06
- Unidentified speaker (speaker_1): so, yeah.
- 22:07
- Unidentified speaker (UNKNOWN): Like
- 22:07
- Unidentified speaker (speaker_1): with history, we both did history, right?
- 22:09
- Unidentified speaker (UNKNOWN): Like
- 22:09
- Unidentified speaker (speaker_1): I did it in the
- 22:10
- Unidentified speaker (UNKNOWN): pre
- 22:11
- Unidentified speaker (speaker_1): I suppose my last year I had AI, but like back then it wasn't as powerful. These days, like, what's the point of writing an essay? You could just like give it the books to read, just upload that as PDFs. This is not advice for for for students, and then just ask it to write. the essay in the style of your favorite historian. And then all of
- 22:31
- Unidentified speaker (speaker_1): the all of the writing would it would be fantastic. You'll get it first.
- 22:34
- Unidentified speaker (speaker_0): Yeah.
- 22:35
- Unidentified speaker (speaker_1): Um so like the so how do you even mark things? And I was speaking to teachers when I was doing the educating thing. It's like, oh the teacher sets the homework, b makes a worksheet using AI that the students then fill out with AI, only for the teacher to then mark market it with AI and give feedback. in AI, right? So it's like the teachers, teachers are using it just as
- 22:54
- Unidentified speaker (speaker_1): much as the students. Then the students are using it to answer the questions. So then there's no learning, right?
- 23:01
- Unidentified speaker (speaker_0): Yeah, no, there's not.
- 23:02
- Unidentified speaker (speaker_1): There's no learning, yeah.
- 23:03
- Unidentified speaker (speaker_0): I feel like it's quite
- 23:05
- Unidentified speaker (UNKNOWN): yeah.
- 23:05
- Unidentified speaker (speaker_0): So it's quite polarizing. So I'm see assuming some people love it. This is the future.
- 23:10
- Unidentified speaker (UNKNOWN): Yeah.
- 23:10
- Unidentified speaker (speaker_1): We're
- 23:10
- Unidentified speaker (speaker_0): in a declining population, we need AI, and other people are like, this is awful.
- 23:15
- Unidentified speaker (UNKNOWN): I
- 23:15
- Unidentified speaker (speaker_1): think so, yeah. And I think socially, society as a society, we don't know what to do with it yet. We haven't figured it out from an anthropological
- 23:23
- Unidentified speaker (UNKNOWN): like
- 23:23
- Unidentified speaker (speaker_1): way. It's like, what do you do with the fact that you don't need to learn anymore? Um, everything we we see or we read on the news, we don't know if it's true. Right? Like like
- 23:33
- Unidentified speaker (UNKNOWN): I
- 23:33
- Unidentified speaker (speaker_1): don't know, so much so much of the content you see on Instagram or whatever is AI generated. And you notice the bad ones, right? Like you
- 23:40
- Unidentified speaker (UNKNOWN): notice the
- 23:41
- Unidentified speaker (speaker_1): really AI slop. But like you don't notice the good ones. So then it's like, what is true? So then like
- 23:47
- Unidentified speaker (UNKNOWN): if
- 23:47
- Unidentified speaker (speaker_1): you push that, my thinking is like, does truth even matter? Like um
- 23:53
- Unidentified speaker (UNKNOWN): so
- 23:53
- Unidentified speaker (speaker_1): so I don't know.
- 23:55
- Unidentified speaker (speaker_0): We're on a bit of a tangent there. Yeah.
- 23:57
- Unidentified speaker (UNKNOWN): But
- 23:57
- Unidentified speaker (speaker_0): yeah, I don't know. I was just trying to think. Well causal Cosal sounds really, really interesting. I had no idea. about all these like simulations and you can create a synthetic person, do all this mathematicianal mathematician stuff.
- 24:11
- Unidentified speaker (UNKNOWN): I
- 24:11
- Unidentified speaker (speaker_0): don't even know how you c came up with this, to be honest.
- 24:14
- Unidentified speaker (UNKNOWN): Yeah,
- 24:15
- Unidentified speaker (speaker_1): it was strange. It was a strange iteration and a strange journey. Um I think if you do start a business, your your first six, to eight months, the idea changes a lot.
- 24:24
- Unidentified speaker (speaker_0): Yeah, I've seen that.
- 24:25
- Unidentified speaker (speaker_1): And that th that's it has to be because chances are the the idea
- 24:30
- Unidentified speaker (UNKNOWN): um
- 24:31
- Unidentified speaker (speaker_1): chances are the idea isn't exactly gonna fit with what people want.
- 24:35
- Unidentified speaker (speaker_0): No.
- 24:35
- Unidentified speaker (UNKNOWN): So
- 24:35
- Unidentified speaker (speaker_1): you're gonna have to adapt and you figure out as you go along.
- 24:38
- Unidentified speaker (UNKNOWN): I think
- 24:38
- Unidentified speaker (speaker_0): that's great advice. Arif, thank you so much for coming on.
- 24:41
- Unidentified speaker (UNKNOWN): Thank
- 24:41
- Unidentified speaker (speaker_1): you for having me.