2024
AI in marketing 2024
An opinionated guide to the current AI landscape
Please take everything here with a grain of salt, not really meant to be taken seriously and is not advice in any way. Someone knows the answers - but it's not me. Well, at least you know this wasn't an AI generated article.
I try to keep up with the constant change of technology, which is focused on AI heavily at the moment, and what startups are getting funding and if any of them stand out or are interesting to me. For example:
https://a16z.com/ai-marketer-how-gen-ai-based-software-is-advancing-marketing-and-sales/
I even plagiarized (literally) their graphic - see above. I wanted to check out the companies listed and try out gpt-4o multimodal image capability, so I uploaded the image and asked to create links. It got a couple wrong, but was overall useful. Here are the categories and links.
I try to keep up with the constant change of technology, which is focused on AI heavily at the moment, and what startups are getting funding and if any of them stand out or are interesting to me. For example:
https://a16z.com/ai-marketer-how-gen-ai-based-software-is-advancing-marketing-and-sales/
I even plagiarized (literally) their graphic - see above. I wanted to check out the companies listed and try out gpt-4o multimodal image capability, so I uploaded the image and asked to create links. It got a couple wrong, but was overall useful. Here are the categories and links.
Copy
1. Jasper - https://www.jasper.ai/
2. Writer (now known as Writer.com) - https://writer.com/
3. Copy.ai - https://www.copy.ai/
4. Hypotenuse AI - https://www.hypotenuse.ai/
5. Tofu - https://www.tofuhq.com/
6. Coframe - https://coframe.ai/ https://www.coframe.com/
SEO
1. Writesonic - https://writesonic.com/
2. Positional - https://positional.com/
3. Letterdrop - https://www.letterdrop.com
4. Daydream - https://daydream.ing/
3. Ripple - https://getripple.ai/
4. Typeface - https://www.typeface.ai/ (This company appears under multiple categories)
5. Regie.ai - https://www.regie.ai/
6. Lavender - https://www.lavender.ai/
7. SellScale - https://sellscale.com
8. Clay - https://www.clay.run/
9. Madmen AI - https://usemadmen.ai/ (Competitor ad research)
User Research
9. Madmen AI - https://usemadmen.ai/ (Competitor ad research)
User Research
1. Outset.ai - https://outset.ai/
2. Voicepanel - Unknown / AI hallucinated / I couldn't find the website
3. Validated - Unknown / AI hallucinated / I couldn't find the website
1. Creatify - https://creatify.ai/
4. Waymark - https://waymark.com/
5. Steve AI - https://www.steve.ai/
6. Captions - https://www.captions.ai/
7. Tavus - https://www.tavus.io/
Images
1. Midjourney - https://www.midjourney.com/
2. Ideogram - https://ideogram.ai/
3. Flair AI - https://www.flair.ai/
4. Pebblely - https://www.pebblely.com/
5. Treat - https://www.trytreat.ai/
6. Typeface - https://www.typeface.ai/ (Also listed under other categories)
Website Generation
1. Durable - https://durable.co/
2. Topline Pro - https://www.toplinepro.com
3. Framer - https://www.framer.com/
4. Relume - https://www.relume.io/
5. Coframe - https://coframe.ai/ (Also listed under Copy and User Research categories)
Summary
Overall, I am not yet a fan of spamming more AI-slop out there, so anything copy or SEO related I view as a tool for a human in the loop to use for ideas and inspiration. Some of these tools had no way for me to test them, but in general I need a convincing reason to try it as opposed to using GPT or Claude with a specific profile or prompt.
The benefit of using AI LLMs is mostly gained by speeding up processes. Not in the sense that it can be automated and then we "trust" the automation to complete some kind of task, but to inform human decision making.
I am optimistic on AI generated imagery as an exciting tool, but for specific use cases. It is unfortunate that it has a tendency toward misrepresentation, or showing something that did not actually happen, which is fundamentally conceptually similar to a lie. For this reason I think the capability of it being misused makes me nervous about the technology and this is not uncommon.
That being said, I think there are use cases for creating placeholders, quick or at scale generations, or showing things you either couldn't show in real life (such as bringing things from the past to life) or are not budgeted. It just takes a bit of going above and beyond to be transparent in your processes and being sure to present things in a way that creates and solidifies trust.
The benefit of using AI LLMs is mostly gained by speeding up processes. Not in the sense that it can be automated and then we "trust" the automation to complete some kind of task, but to inform human decision making.
I am optimistic on AI generated imagery as an exciting tool, but for specific use cases. It is unfortunate that it has a tendency toward misrepresentation, or showing something that did not actually happen, which is fundamentally conceptually similar to a lie. For this reason I think the capability of it being misused makes me nervous about the technology and this is not uncommon.
That being said, I think there are use cases for creating placeholders, quick or at scale generations, or showing things you either couldn't show in real life (such as bringing things from the past to life) or are not budgeted. It just takes a bit of going above and beyond to be transparent in your processes and being sure to present things in a way that creates and solidifies trust.
Images
Image generation is something I find fascinating.
I am simultaneously a huge fan of anti-AI art social media communities.
Midjourney I think is the leader in this category.
I have recently been playing with https://blackforestlabs.ai/ flux, which is a sort of, somewhat open source model.
I also like Stable Diffusion, but if an app is going to use it to build on top of, they probably need to contribute something else substantial for it to be interesting.
Ideogram was state of the art when it came out in that it could incorporate real (as in readable) text in image generation. While this was an improvement, Flux does this while creating high quality images, and I believe an update on Midjourney allows you to do this as well.
I was not able to actually do a proper review for all the startups listed, so feel free to explore them further!
I am simultaneously a huge fan of anti-AI art social media communities.
Midjourney I think is the leader in this category.
I have recently been playing with https://blackforestlabs.ai/ flux, which is a sort of, somewhat open source model.
I also like Stable Diffusion, but if an app is going to use it to build on top of, they probably need to contribute something else substantial for it to be interesting.
Ideogram was state of the art when it came out in that it could incorporate real (as in readable) text in image generation. While this was an improvement, Flux does this while creating high quality images, and I believe an update on Midjourney allows you to do this as well.
I was not able to actually do a proper review for all the startups listed, so feel free to explore them further!
Video generation
Avatar generation
This is a subcategory that I am just -not- a fan of. It reminds me a bit of shady SEO practices in the beginning days of search optimization. If someone doesn't know it is AI, but later finds out, I feel like there is a breach of trust issue. If they know it is AI generated, I personally have a negative reaction - why didn't you just record it yourself? What does it get you to create the illusion of someone else? Sure it saves you time, and possibly production costs, but now I feel like you just don't care about me. I'm sure there are proper use cases and this is a somewhat unfair assessment, but I haven't seen it executed properly yet.
I do see the value of zero production cost production with a few different messaging strategies being tested, and then using that as a guide to find the best messaging. I'm just not sure this is worth the trust issue tradeoff issue mentioned above. It's a common practice to advertise, even sell, theoretical products to see if the interest is there, and if it is then deliver high quality production. So along this logic, it might be cost effective to create an AI avatar of a real spokesperson with various messages, and then record the actual final message as a real production. Not really sure how this might work, though. I might come around to appreciating this technology, but I'm not there yet. Just because we can, does that mean we should?
This is a subcategory that I am just -not- a fan of. It reminds me a bit of shady SEO practices in the beginning days of search optimization. If someone doesn't know it is AI, but later finds out, I feel like there is a breach of trust issue. If they know it is AI generated, I personally have a negative reaction - why didn't you just record it yourself? What does it get you to create the illusion of someone else? Sure it saves you time, and possibly production costs, but now I feel like you just don't care about me. I'm sure there are proper use cases and this is a somewhat unfair assessment, but I haven't seen it executed properly yet.
I do see the value of zero production cost production with a few different messaging strategies being tested, and then using that as a guide to find the best messaging. I'm just not sure this is worth the trust issue tradeoff issue mentioned above. It's a common practice to advertise, even sell, theoretical products to see if the interest is there, and if it is then deliver high quality production. So along this logic, it might be cost effective to create an AI avatar of a real spokesperson with various messages, and then record the actual final message as a real production. Not really sure how this might work, though. I might come around to appreciating this technology, but I'm not there yet. Just because we can, does that mean we should?
Video creation
Creatify stands out to me before I have had a chance to dive deep into the alternatives. I like the idea of breaking apart videos based on topic and using AI do extract topics and cut it automatically seems pretty attractive, although I'm not sure I trust the AI to go ahead and do this without a middle step where I approve where it is going to make the breaks and correct any mistakes. I might be impressed but at the end of the day it uses the same LLMs to categorize a text representation of the audio portion of the video, and I know I don't always make the same call it does - although in some cases this is because I have made a quirky call, not the LLM! Cutting out a specific topic to make a teaser or shortened version is solid as well, but sometimes editing for this type of quick introduction needs to be very succinct and not just a full broken apart topic but perhaps a soundbite of the most interesting part of the video, and the AI probably (maybe?... yet) can't do this the way a human can make that type of call. But better to have this than no summarized portions at all.
Waymark - this is really cool, I had never heard of this before. It seems very professional, a fully automated AI system. There's definitely some magic behind the scenes going on that I don't understand. It gets you a usable video version of your existing web assets (aimed at smaller businesses), so your video quality is going to be very correlated to the quality of your existing materials. Of course a business with any budget at all can probably get higher quality out of produced video from a freelancer, but it leverages existing assets well.
Creatify stands out to me before I have had a chance to dive deep into the alternatives. I like the idea of breaking apart videos based on topic and using AI do extract topics and cut it automatically seems pretty attractive, although I'm not sure I trust the AI to go ahead and do this without a middle step where I approve where it is going to make the breaks and correct any mistakes. I might be impressed but at the end of the day it uses the same LLMs to categorize a text representation of the audio portion of the video, and I know I don't always make the same call it does - although in some cases this is because I have made a quirky call, not the LLM! Cutting out a specific topic to make a teaser or shortened version is solid as well, but sometimes editing for this type of quick introduction needs to be very succinct and not just a full broken apart topic but perhaps a soundbite of the most interesting part of the video, and the AI probably (maybe?... yet) can't do this the way a human can make that type of call. But better to have this than no summarized portions at all.
Waymark - this is really cool, I had never heard of this before. It seems very professional, a fully automated AI system. There's definitely some magic behind the scenes going on that I don't understand. It gets you a usable video version of your existing web assets (aimed at smaller businesses), so your video quality is going to be very correlated to the quality of your existing materials. Of course a business with any budget at all can probably get higher quality out of produced video from a freelancer, but it leverages existing assets well.
Website generation
Relume has some really interesting resources, and I definitely recommend checking them out.
Framer is definitely a go to builder for heavy animations. In fact Motion Storyline uses the same motion library (but not the React version) for the animation transitions used in HTML. Highly recommended.
As for the others listed, time did not permit for a proper review. Again feel free to follow the links to check them out for yourself!
Framer is definitely a go to builder for heavy animations. In fact Motion Storyline uses the same motion library (but not the React version) for the animation transitions used in HTML. Highly recommended.
As for the others listed, time did not permit for a proper review. Again feel free to follow the links to check them out for yourself!