The AI Daily Brief: Artificial Intelligence News and Analysis - Is AI Going to Eat SaaS?

Episode Date: September 12, 2024

Is generative AI about to disrupt the SaaS industry? Klarna’s recent decision to phase out Salesforce and Workday, citing AI solutions as a more efficient alternative, has stirred up discussions in the tech world. With AI engineers creating custom applications at a fraction of the cost, could this signal a broader trend in enterprise software? Explore the impact of AI on SaaS, the future of enterprise tools, and whether custom-built AI solutions could reshape the software landscape. Concerned about being spied on? Tired of censored responses? AI Daily Brief listeners receive a 20% discount on Venice Pro. Visit ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://venice.ai/nlw ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠and enter the discount code NLWDAILYBRIEF. Learn how to use AI with the world's biggest library of fun and useful tutorials: https://besuper.ai/ Use code 'podcast' for 50% off your first month. The AI Daily Brief helps you understand the most important news and discussions in AI. Subscribe to the podcast version of The AI Daily Brief wherever you listen: https://pod.link/1680633614 Subscribe to the newsletter: https://aidailybrief.beehiiv.com/ Join our Discord: https://bit.ly/aibreakdown

Today on the AI Daily Brief, Clarna is shutting down Salesforce and workday and replacing them with AI. Before that on the headlines, OpenAI's Strawberry appears to be coming within weeks. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. To join the conversation, follow the Discord link in our show notes. Welcome back to the AI Daily Brief Headlines edition, all the daily AI news you need in around five minutes. We kick off today with some interesting updates around OpenAI's much-rumored strawberry model. It seems that we are finally getting some real information, still in the form of reports, not things that are confirmed, but definitely a little bit better sourced, i.e., it's from the information, not from some random Twitter accounts. According to the information, OpenAI is planning to release Strawberry as a model option within chat GPT inside the next two weeks.

This report comes from two people who have tested the model. According to the people who've seen it, it's, quote, quite different from the regular service with some advantages and shortcomings. Now, the big idea behind Strawberry or the big hope is that it is a better reasoning model. The way that this works in practice is that Strawberry, quote-unquote, thinks before responding. That thinking stage usually lasts 10 to 20 seconds. We'll come back to what that suggests about how the model might work in just a minute. In terms of other differences from how ChatchapT operates right now, the initial version of Strawberry is not multimodal.

It's only a text-based model. It also appears at this stage like it will be priced differently. According to another information source, the most likely scenario is that strawberry has rate limits for pro users that restrict them to a maximum number of messages per hour, with the potential for a higher price tier that gets around some of those limits. In terms of performance, one of the big shifts is that strawberry is supposed to significantly simplify the prompting process. As the information puts it, currently customers have to type all kinds of additional words

into chat chippy T to get the answer they want, such as telling the chatbot to walk through its immediate reasoning steps to arrive at its final answer, otherwise known as chain of thought prompting. Strawberry's capabilities are supposed to help customers avoid doing that or other hacks to achieve smarter results. The end result of this should be that strawberry will be better at math problems, it will be better at coding, but it should also be better at subjective business tasks like brainstorming. Now, brainstorming is one of the most reported tasks that we see on super as things that people are using AI for, so that actually could be really valuable. But what do people who have tried this actually think of it? At least from this reporting, it's definitely not a

a slam dunk. For example, it sounds like sometimes when people ask it a simple question that strawberry should be able to skip that thinking step, the model doesn't always do that, meaning that people are sitting there waiting for that 10 to 20 seconds, even with a really simple query that the regular chat chip ET could answer almost instantaneously. People who have tried this also note that that 10 to 20 second wait time feels really, really long. And they're not sure that the answers are really worth that extra waiting. The users have characterized the responses as only slightly better. Strawberry is also supposed to have better memory. And, you're not supposed to have better memory, being able to incorporate previous chats, but it appears that it has sometimes struggled with

that as well. The question then kind of obviously becomes, is strawberry being rushed out because of competitive pressure? OpenAI spent a very long time in the catbird seat leading the AI industry, but is obviously under increased competition from all sides. Do they feel like they have to push strawberry out? And is there the potential that it actually does more harm than good for them if it fails to be noticeably better? Now, one of the things that some people are exploring is exactly how strawberry works. Rohan Paul tweets, Google's recent paper may be helping Open AI Strawberry. The relevant paper from Google DeepM was called scaling LLM test time compute optimally can be more effective than scaling model parameters. The paper basically says searching at inference will give you great final results

from the LLM. Now from rumors, Strawberry is using some form of inference time compute strategies, using search techniques over the response space to improve reasoning. Basically, this is not just the matter of a model being bigger, it's actually a slightly different approach to how it retrieves information. A lot of the chatter is around how subdued the messaging is around this. AI explained, writes, we just heard that the famed chat GPT upgrade strawberry is coming by September 24th, but something doesn't make sense. It was a threat to humanity, according to certain OpenAI X-Staff. It rises to human-level reasoning, according to a leak to Bloomberg. But according to early testers, its slightly better answers aren't worth the 10-to-20-second wait, and it often

thinks for that long, even if you ask it not to, and it will be pricey. Something doesn't add up. Swix had a similar response. As rumored Strawberry will be releasing in time for OpenAI Dev Day, it looks like. OpenAI seems to be downplaying this a lot, concerning or just sandbagging. Now, of course, it is important to note that OpenAI itself has not released this information. It's just from people who are in the know. But it does have a bit of a feel of a coordinated leak. And if so, it certainly feels like one meant to tamp down expectations rather than ramp them up.

Second today in the headlines edition, on Tuesday night, there was a debate in the U.S. presidential campaign between Kamala Harris and Donald Trump, and AI made a little cameo. VP Kamala Harris attacked Trump over his approach to chip exports during his term. She accused Trump of, quote, selling American chips to China to help them modernize their military, whereas Harris said that she was an advocate of, quote, investing in American-based technology so that we win the race on AI on quantum computing. Trump bit back, saying that Chinese tech companies, quote, want their chips from Taiwan, not the U.S., and that the U.S., quote, hardly makes chips anymore, blaming, of course, democratic policies.

Adam Kovacevich of the Chamber of Progress, writes Kamala Harris outhocking Trump on China, love to see it. And I think while it was just one line, it shows that this sort of technological superiority and the geopolitics of AI are front and center in this campaign. Still, maybe even the more notable invocation of AI came in a post after the debate from none other than Taylor Swift. In a post where she endorsed VP Kamala Harris for president, Swift wrote, recently I was made aware that AI of me falsely endorsing Donald Trump's presidential run was posted to his site. really conjured up my fears around AI and the dangers of spreading misinformation. Now, she went on to say

that it brought me to the conclusion that I need to be very transparent about my actual plans for this election as a voter, because, quote, the simplest way to combat misinformation is with the truth. Bloomberg AI writer Rachel Metz writes, the rising impact of generative AI cannot be overstated here. Taylor Swift cites a recent AI generated image of her that made it falsely appear as if she supported Donald Trump as the reason she decided to speak out regarding her decision to vote for Kamala Harris for president. And so, friends, AI getting a mention by the potential future president in the U.S. and current vice president? Fairly big. AI being called out as a point of discussion by Swift. Well, now it's in the cultural conversation. That's going to do it for today's AI Daily Brief

Headlines Edition. Next up, the main episode. Today's episode is brought to you by Venice. Venice is a private, uncensored generative AI app. It accesses open source models to enable text, image, and code generation without the fear of being spied on or having your data exploited. Discuss anything with Venice without concern about it being monitored, sold, or given to advertisers and governments. Venice is different because your conversations and creations are kept securely within the browser, never stored or accessible by Venice. Unlike other AI apps, Venice won't tell you what's okay to say or not. Venice won't patronize you. It simply provides direct access to machine intelligence, no topics are off limits, no ideas, or taboo. With Venice, you're in control of the AI as you should be.

Pro subscriptions are available for $49 a year or $8 per month. AI Daily Brief listeners receive a 20% discount on Venice Pro. Visit venice.a.I slash NLW and enter the discount code NLW Daily Brief. That's NLW Daily Brief, all one word. Today's episode is brought to you by Super Intelligent, which is of course our platform that helps you learn how to use AI tools and perhaps even more importantly, gives you ideas on the best use cases that are actually going to help you achieve, whatever it is you want to achieve.

To recognize the end of summer and back to school slash back to work, we are running our best promotion ever when you sign up for Super Intelligent between now and the end of August using code so back, your first month will be 100% free. The platform features over 600 fun, highly practical AI tutorials that get you using AI fast and with an eye to actually transforming how you get things done. We've just launched Super for Teams, so if you have a group of people at your company that want to figure out how to use AI together, I highly suggest you check it out. But for those of you who are using Super Intelligent as an individual, once again, if you sign up for Superintelligent

between now and the end of the month using code so back, you will get your first month 100% free. Go to B-Supert.a.I. And check it out today. Welcome back to the AI Daily Brief. Today, we are exploring a big question about the downstream implications of artificial intelligence on a particular slice of software, specifically the SaaS industry. The question we're going to to ask, simply put, is generative AI inevitably going to eat the SaaS space? Now, there has been more and more chatter about this question over the last three or four months. Back in June, HFS wrote a piece called Generative AI Eat SAS. The piece was prompted by a Salesforce earnings call that led to that company's stock tanking by about 20%.

They write, the current economic model SaaS firms used to generate revenue and shareholder value is being upended as enterprise signal a fundamental shift in how they'll buy software and the future. While many financial analysts claim this is due to economic headwinds leading up to an enterprise IT spending slowdown, HFS believes the deeper issue is enterprise's increasing focus on generative AI becoming the leading application and data orchestration vehicle in enterprise IT. So there are a couple things going on. First, there is a simple dollar and cents question. They write, we believe the advances in Gen. AI, typified by the recent OpenAI launch of Chatschipt 4O, are making it impossible for C-suite executives to ignore, and they need to find budget from

legacy SaaS investments to help fund GenAI initiatives. Basically, there is some amount of one-to-one switch from SaaS spending over here to AI spending over there. However, they think that there's something more going on as well. This they characterize as enterprises reaching a breaking point being held hostage by SaaS premiums. They write, the dollars that SaaS firms have been able to charge based on user's Cedar endpoint have become so significant that enterprises are eager to revisit what they can build, not using emerging AI technologies to refactor, re-architect, and make new composable applications. Additionally, we see competitors emerging that can quickly build modules or data-centric software that can rapidly be adapted or actioned into use at fractions

at fractions of the cost of third-party enterprise SaaS applications. In fact, ultimately, they liken the shift from SaaS to Gen AI to what SaaS did to on-premise software back in the early 2000s. They write, the trend mirrors the way Salesforce itself led a charge to replace monolithic on-premise enterprise applications and ushered in the cloud era. The rapid adoption of generative AI solutions to drive software and business processes may only take months. Therefore, many SaaS vendors can expect this shift to increasingly impact sales pipelines, revenues, and their customer-installed base. This then is the interesting context for news from Clarna

that the company plans on shutting down both SaaS and Workday. On a conference call, Clarna CEO Sebastian Semiat Kowski said, there are large ongoing internal initiatives that are a combination of AI, standardization, and simplification. As an example, we just shut down Salesforce. Within a few weeks, we will shut down WorkDakowsy. day. We are shutting down a lot of our SaaS providers as we are able to consolidate. Now, Klarna has been very on the front of the AI transformation. OpenAI has a customer story

about them on their website, discussing how after ChatGBTGBT launched in November 2022, Klarna became the, quote, first European company and the first fintech firm globally to launch a ChatGPT plugin. Back in February, the company reported that its new AI assistant in its first month had handled two-thirds of customer service chats. That meant it was doing the equivalent work of 700, full-time agents. Clarna also said that the customer service chats handled by AI were on par with human agents in regard to customer satisfaction score, more accurate in terms of error in resolution, leading to a 25% drop in repeat inquiries, and much, much faster. The average time in which

customers resolve their issues dropped from 11 minutes to less than two minutes. And that is, of course, to say nothing of the fact that it is 24-7 and can communicate in more than 35 languages. Perhaps unsurprisingly, then, Klarna is looking to reduce its workforce. The company announced that it would reduce its workforce over time by almost 50%. They said that its use of AI will enable it to reduce its staff from 3,800 to 2000. Now, the approach that they're trying to take, they call natural attrition. Basically, they're saying they're not going to lay off people, but instead, as people move on to other jobs or seek new opportunities, the company will instead just opt not to replace

them. The CEO did say that the company would not be employing common strategies like halting promotions, freezing pay raises, or increasing performance improvement plans in order to encourage people to leave. Instead, they'll just let the natural attrition do its work. So back to this idea of Klarna shutting down Salesforce and Workday. For many, this was confirmation of a bigger trend. Bidu Ready writes, as AI engineers become prolific, you can create custom applications that are 10x cheaper to run than these SaaS applications. Investor Gokul Rajaram wrote, Enterprise SaaS Stickiness. What stickiness? This news from Klarna should have every enterprise SaaS company shaking in their

boots. If an internal team using AI can replicate 20 plus years of work and customization from Salesforce and Workday, to the extent the company doesn't feel the need to pay for these tools anymore, everything we know about stickiness and durability of enterprise software needs to be rethought in the light of AI. In fact, their comments indicate that they were able to use AI to rethink the products from first principles and make them simpler and easier to use. I wouldn't be surprised if the mandate of the head of IT at large enterprises gradually expands to not just negotiating supporting enterprise software licenses, but replacing them with custom-built products from the ground up, especially for the largest software products that cost 7 to 8 digits per year.

There were also, however, some that were a little bit skeptical. Investor Rex Salisbury wrote, Klarna is ripping out their SaaS providers including Salesforce and Workday. This is after reducing customer service headcount 50%, all because of AI. They are either, one, getting amazing leverage from AI, or two, giving a master class and how to put a positive spin-on layoffs. Reality is a bit of both. Probably not getting as much leverage as it seems, but they are 100% capturing a lot of media

attention. Great way to drive attention in advance of an IPO. The other skepticism is that there's more to software than just building the software in the first place. A16Z's Martin Casato writes, someone is going to learn that state consistency and integrations are hard. Zach Cantor quoted that and said, has been easy for 10 plus years now to build bespoke replacements for SaaS products. AI makes that even easier. The hard part isn't building it. It's operating and maintaining it when the best engineers want to solve problems for customers, not upgrade your crappy Salesforce clone. Bucco Capital had a similar thought. Regarding Klarna ripping out Salesforce and Workday, even if it's true, is it actually the best use of capital to rebuild in-house? Feels like a

massive distraction, especially when your business has no path to selling the in-house solution. I'm deeply skeptical the math works. Another investor Stevensonovsky responded, The number of internal tools that turned into successful side hustles is effectively zero. The number of times this has been tried is absurdly high. Still, it's hard not to ignore the treadlines. Separately last week, Rohan Paul had tweeted, video of someone creating a full-stack SaaS app with just cursor and Anthropic. And increasingly, this is being done by people who have never coded before. Now, I do think it's true that there's a lot more that goes into software than just the code to release it in the first place, but it feels fairly

undeniable that there is going to be a shift in how we think about these things. If you want to read more about this, I would recommend A16Z's essay, death of a Salesforce, why AI will transform the next generation of sales tech. For now though, that is going to do it for today's AI Daily Brief. Appreciate you listening or watching as always. And it's a lot. Until next time, peace.