AIContentGen Chats with Francesco Magnocavallo, Chief Product Officer of Contents.com
Transcript of the podcast:
Scott (00:04): Hello, and welcome to Marketing AI CEO chats. And I’m Scott Sweeney and I’m here with my co-founder from AI content. Jen, John CA.
Francesco (00:14): Thanks, Scott. Great to be here again.
Scott (00:17): And today we’re speaking with Francesco Magnocavallo. I hope I said that, right? Francesco.
Francesco (00:25): Yeah. Scott, don’t worry. It’s a good
Scott (00:27): Line. <Laugh> and he is the chief product firstname.lastname@example.org. And today we’re gonna be talking AI marketing and learning more about what contents.com can do for you. And so thanks for joining us.
Francesco (00:44): Yeah. Thank you for having me. It’s great to be speaking with you for the public.
Scott (00:49): And so we’re gonna get right into it. We want a little bit of background first Francesco. You have a very, very interesting background looking through your LinkedIn. I saw that not only did you do some content for hears communications and luxury brands, but the dichotomy of spending 10 years in hip hop in Milan is really interesting. And I’m just wondering how all that melds into you and your AI journey and where you are now.
Francesco (01:21): Yeah, so I figure basically it’s always been about creating content. So I had the opportunity and the lack of meeting pioneers in NEP pop, like phase two. He was he’s been in Milan, been friends for years and learning, you know, very different kind of writing that was wild writing, a personal alphabet. Nobody really can replicate. And you know, that’s coming from your soul. It’s not even it’s very different. So those were crazy times getting to know this culture, this great culture, and then doing some systems. So basically you know, this was really multimedia <laugh> at mm-hmm <affirmative> and often they’re doing freestyle, just getting your soul out. And later through doing websites, I got into content and I’m fond of remembering the very early days of blogging, like years 2000, and 2001, the atmosphere was very different.
Francesco (02:21): Everybody was a developer then in blogging, it was easy to learn. You, you could study, you know, IBM stuff, materials just early days of content designs. That was great in learning your way around content. Great times then when blogs exploded, we launched the company, it was a clone of Google media and incorporated that to be companies publishing a gadget Giese model and all that stuff. And we eventually sold the company. So, I remember I launched something like one blog every month for five or six years. Those were crazy times. We had a thousand orders writing for us in all <laugh>. And then we sold the company. I did some consulting and ended up at Hurst where I spent six years doing digital strategy and launching the digital operation for six or seven websites. And I had, I think I had the privilege of launching Esquire magazine in Italy and our per so our is 150 years old brand. That’s something it’s, it’s even, you know, difficult to think about this because this there’s so much history in there. And the company now is, is got incredible infrastructure and strategy in New York. So it was very interesting times as well,
John (03:39): Francesco I’ve worked in the custom content industry as well, and that’s basically the custom content industry is the magazine industry either you know, public magazines or, or private corporate ones. And one of the things that I found, although I’m a digital guy, I was working with those magazine editors and writers. I found the discipline of that industry to be really interesting because, to me, it was as a digital guy, almost like writing an entire website in a month when you’re changing a magazine every month. How did that discipline affect you and how you think about content?
Francesco (04:18): Yeah, I think this is a great concept indeed because the freer the ecosystem and the more possibilities you have, the more discipline you need, the more strategy. So this is very true for digital as well. What I learned from the first magazine was really journalism in the sense of not really high brow, but high-quality journalism. There was some high-brow content. We published it in Esquire magazine, but, you know, doing really premium quality stuff. And that’s something I learned the hard way with print journalism, doing digital transformation, which is a very difficult activity. Sometimes was really rough to deal with people with very different languages, different, you know, a very different mindset, different professionals com everything was very the two sides, the digital and the print merging the two sides was an incredibly demanding task.
Scott (05:20): I can see that Francesco and I think in many ways AI content generation has a lot to do with digital transformation too. So it’s not just about the tools, but it’s about the process. So so let’s, let’s kind of move into that. I don’t know if you’d agree with that, but feel free to comment on that. Tell us a little bit if you want more about your AI journey, but then let’s go in and talk a little bit about contents.com.
Francesco (05:48): Yeah. Yes. So ma Milano, our founder, and executive officer, we, we were acquaintances from you know, 10 years ago. And when he see me as a feeling for openings on the market he ended up thinking I was the right person to bring some journalistic mindset to work and design the system with the engineers. So, he chose not to have a journalistic system designed by engineers, but by a journalist. So that was you know, my point of view. And I think it’s been really rewarding work working with my data science team, and my colleagues in data science. That’s very interesting. So then we added a computational linguist and I think, by the time we started merging and having a common framework in languages, things really improved a lot. So when I speak to clients in meetings with my sales colleagues, I find that I often speak the same language because I had the same problems and the same workflows that the journalists have. And so it’s very easy to bring something that creates value for them, with automation basically.
Speaker 3 (07:06): And Francesco, what perhaps we could get into the service of the software. What are the capabilities of the software? What does it do?
Francesco (07:14): Yeah. So Scott mentioned that you need workflows. You obviously need some technology and we work with open AI by the way, but you also need talent. So what you know, the defining characteristic of contents.com is that we blend human talent with technology. So we try to get the best of both words. That’s a very basic concept in AI-assisted or human in the loop. So basically the company had a copywriting marketplace from a Francesco few years ago and it kinda went out of fashion. You know, you cannot really speak about this sort of product with DC anymore because it’s, it’s very old, but it’s coming back as a strategic point because you can have a finishing layer and and and have the machine designed by a journalist and finished by a journalist.
Francesco (08:13): So I think this is gonna be a very interesting concept if you been reading Andre in order with the magazine, that’s this article that came out just couple days ago, and this is their point. So there’s gonna be a deluge in AI content, and we don’t really know what school is gonna do in a couple of years or next year. So having a sort of blended approach where it’s AI, but it’s human definitely makes some sense. So this is a key concept. And in terms of a service we provide to the client, the client can often choose to have just a human professional working on his content, just the machine, if he wants to make it cheap and, and quick or a blend. So this is the this is something that’s very different because it’s kind of easy nowadays to just build you know, a front end on open AI or any other language model of your choice and build a little pipeline to do some pre-processing, but really to have a complex system. That I think creates a different scenario
Scott (09:21): For that’s very interesting Francesco and I, I love the approach because as we know many times you can’t, well, we, you never can just tell a machine what to write and have it come out. Perfect. Right? So you need, you need some inputs and you need some massaging. Tell us, I think everyone can understand human writing and everyone can understand machine writing. How does the blend work?
Francesco (09:56): Yeah, so it’s blended by workflows basically. So what we, we have and we gonna improve and build and invest in is newsroom workflows, basically. So a full system where there are roles, there are teams, and there’s the process of commissioning a piece commenting on a piece, reviewing the draft, approving it, and publishing. So the last mile in the system is connectors to the CMS. So you see, we currently have a WordPress and a Shopify connector. We are building an Adobe connector for an enterprise large project with a big consulting firm. And so it’s, it’s really a lot of different pieces that mix for, for to have something that’s not easy to replicate. You know, that’s, that’s kind of like our mode, our defensive mode, it’s so complex, it’s kind of a hassle. You cannot replicate it faster.
Scott (10:59): Excellent. And so as you look out at the AI landscape, where do you see the strength of contents lying as opposed to other systems?
Francesco (11:15): Yeah. I’m sure there’s some room in Europe because we got all these languages. I don’t want to speak about Africa because they got thousands of languages, but in Europe we got, we have a few billion, so that’s a lot anyway. So in this disrespect, we building native data sets and one of the characteristics is many competitors. They just translate everything from English because of large language models, they are proficient, and they are mostly proficient in English. There are a few great projects in Europe at the moment, but it’s still something that that needs to evolve a little bit. So we got French, Spanish, and Italian datasets, and the language model is working natively in those languages. So this is something where I think we can be strong because we work with datasets in four different languages at the moment. And that’s also something that takes a little time. It takes a little love to have journalists working on this. We don’t really just scrape data from the internet. Everything is, you know, manually edited and proofread and worked on.
John (12:28): Let’s go through some quick, rapid-fire questions Francesco perhaps we can ask a couple of questions about how you support the clients you know content strategy team approach around some of the following areas. How about ideas and research?
Francesco (12:50): Yes. So the idea of the founder was to have a suite of services covering the whole customer journey. So basically a content strategy journey going from media monitoring. And we got a pocket solution I think, is N cause it’s got a very strong signal-to-noise ratio. You can it doesn’t require you to spend the money of one of the enterprise or mid-market competitors, which usually costs from one or 200 to a few thousand a month. And of course, it’s working across languages with MLP. So across industries, keyword entities, fairly basic stuff. Then we got a brainstorming service which is basically a text box where you can work with a language model and do a semantics search test, which is very new. And I think this is gonna be the future of search. So not just getting 10 links, but getting, you know, a proper original answer to your search.
Francesco (14:00): So this is also where we can peek inside the mind of clients and get a little fresh data to see what’s going on. What are the needs and the pain points and the jobs to be done? Then of course the main, the main service is content creation where we can do copywriting in. As I mentioned before in assisted mixed hybrid, purely human or purely machine, we got a couple of different input styles. So one is the instruct, the open I G PT three instruct model. So it’s natural language and clients like this, a lot, the possibility of input having natural language input. I want an article like this and that they enjoy this, but the other one is purely a CEO routine. So basically you input a keyword and we do some sort of data science that zoom gap analysis with competitors and building an outline for the, for the article that’s you know, geared to, to generate organic traffic on Google.
Francesco (15:01): Then we got translations. Again, you can do this with machines or humans, and we are moving into multimodel. Of course, this is the big, the big research race at the moment between the large companies. So we’re gonna start at the beginning of September. So I don’t know when this interview is gonna be polished, but starting from September, we’re gonna sell text-to-image services. And we are in the process of refreshing the audio services. So basically we do a basic speech-to-text that’s a utility but there’s a lot going on in text-to-speech. So you can go to the very high-end Hollywood-level voiceover, or you can do many different types of services in with synthetic voices. And so I’m, I’m very interested in in this and some of the clients we began speaking about this topic in July, they very interested because that’s very glamorous that’s hype, but they can speak to clients about something that’s cool, and nobody has at the moment. So basically that’s the suite of services at
Scott (16:19): The moment. That’s very broad. I’m really impressed. And I noticed that you’re hiring for someone that’s gonna be using Dolly too very soon too.
Francesco (16:27): Yeah. We hiring for Dolly too. Yeah. <laugh> and, you know, that’s very interesting because the ideal candidate is, is somebody who knows about photography, illustration, graphic design. So it’s, it’s, it’s really, you know, it’s, it’s very, it’s, it’s very generalized across visual disciplines. And we got a, I think we got an incredible debt for Italy at the moment it’s going out next couple of weeks to our sales team. So this is very, very, very promising.
Scott (17:01): That’s very exciting. I’ve been playing with Dolly too. It’s a lot of fun. <Laugh> yeah.
Francesco (17:05): Yeah.
Scott (17:07): Excellent. John, do you have any more rapid-fire questions that you wanna follow up on, or is
John (17:12): Well, you, you covered a lot of it, Francesco. Yes. I did wonder about expansion and metrics expansion would be you know perhaps doing the general AI content generation, but then maybe for social media or something like that, you know, splitting content up. Yeah. Do you have any aspects of that?
Francesco (17:32): Yeah. You know, when you mention metrics, I think this is very interesting for the gigs of the NP gigs and the energy gigs because actually, this is evolving so fast. There’s no, there’s often no coded KPIs for energy. If you go into research, the pre-print papers, you know, there there’s, there’s no real standard at the moment. So I think this is very interesting and we working a lot on KPIs for the quality of machine translation of NLG. This goes from, you know, very basic like syntax grammar the, the flare journalistic flare, which is very difficult to, to, you know, to digitize, to compile, certainly because it’s you know, it’s, it is the most difficult to plagiarise to that there’s a lot of to similarity of translations, the quality of translation. So I think this is a very interesting area of benchmarking and having common KPIs for energy because often you see as I said before, it’s very easy to build a system, but the large language model will just give you average input.
Francesco (18:43): That’s not enough for clients if you if they’re demanding. So I remember a computational linguist. We were working this winter, relaunching the product, and this was scheduled for spring. So one day we were testing different pipelines, different data sets, and different models with opening, I had just an incredible acceleration around the end and the beginning of the year. And she, one day came into Francesco. We jumped from I don’t know about school grades in America, sorry, be patient with me. But we went from, you know, Italian of 14 years old to Italian of 20 years old. Huh. This is great Francesco. We did it. <Laugh>. Yeah. So this is really, this is happening. And I think it’s very interesting to go behind the scenes and know what the data scientists are doing, what the, what the competition linguists are working on, what these then pleasure and satisfaction.
Speaker 3 (19:40): Well, that’s great. That’s great. Thank you so much for, covering those different aspects. I really appreciate it.
Scott (19:48): So then Francesco, I have one kind of wrap-up question, and then if you want to make any final comments at the end feel free to do so, but I’d like to know you know, this is such a fast-evolving field right now, and everyone’s got all kinds of thoughts about the AI and content generation I’d like to know from you. One thing that most people believe is true about AI content generation, but you take a very contrarian view about it. And I think it’s not true at all.
Francesco (20:28): Yeah. So you see, I think a system that just helps you create an article is just some very basic task, and it doesn’t really give you a proper idea of what you can do with these models. They’re so powerful. You can do incredible things. And I think for large businesses, I mean, million URL websites a hundred million companies in, in the Italian market. So everything is different, in the United States on a global scale, but they got a different set of problems. They don’t want you to create one article for them. They want to manage a million hotel project listings. So I think that’s so much that we’ll come to the market in the next few years, because, you know, when I came to the company, I told the founder, I don’t really wanna do a high copywriting assistant, a Milano. This is so basic who cares about this? Let’s go after, you know, affiliation, let’s go after hotel businesses, let’s go after eCommerce. And so I think that lot is gonna come to the market in terms of specialized pipelines and companies.
John (21:36): Well, thank you so much for joining us, Francesco. We really appreciated the podcast. So it’s a great interview today. Yeah. Thank you very much.
Francesco (21:44): Thank you for us to me. It’s great to be here with you
John (21:48): And thank you to our audience for joining us today. We’ll see you next time.