Newsletter #32: Building software vs AI apps and role of Domain Experts in accelerating Enterprise AI capabilities.

What if Clem Delangue 🤗, co-founder and CEO of Hugging Face is right about building software vs AI applications and the implications for accelerating Enterprise AI capabilities?

Clem is the CEO of Hugging Face, one of the most important Generative AI native companies valued at $4.5B with 5M+ AI builders that sees its members contribute a new model / dataset / app every 10 seconds.

Given the front row seat to observe how AI Builders are building AI Apps that is about as good a seat as anyone else has, his POV on where things are headed in Enterprise adoption and ultimately AI capability development is one to really pay attention to.

Building on my micro essay and the three key takeaways, I wanted to start this newsletter by connecting the rockstar operator’s perspective to a Boston Consulting Group (BCG) study - “Where’s the Value in AI?” - and bring the two together as it gets to the essence of where the puck is headed.

Let’s start with 4%, 22% and 74% reference in my micro essay...

“Our latest research, involving more than 1,000 companies worldwide, shows that only 4% have developed cutting-edge AI capabilities across functions and are using them to consistently generate substantial value. (See Exhibit 1.) Another 22% have an AI strategy and advanced capabilities and are starting to generate value. We call these companies leaders. The remaining 74% have yet to show tangible value from their use of AI.”

It’s safe to say the 4% and likely aspects of the 22% are embracing the difference between building software vs AI Applications and would likely agree with Clem’s thoughts including the following interview excerpts:

“So AI builders are kind of like the new software engineers in a way, right? In the previous paradigm of technology, the way you would build technology was by writing code."

"You would write, like, a million lines of code, and create a product like a Meta Facebook, like a Google, or all the products that we use in our day to day life."

"Now today, the way that you create technology is by training models, using data sets and building AI apps.”

“Because the beauty of AI is that it's a bit less constrained than software in the way that people can contribute to it."

"In a way, to be a software builder, you have to learn a programming language and write lines of code, which is a pretty high barrier to entry, versus for AI you can be considered an AI builder if you contribute expertise, if you contribute data to a model that improved the model."

"Maybe we're going to have 10 times more AI builders than software builders, which would be also good for the world…”

“We make the experience so much more seamless, so much less complex."

"Which is the name of the game for AI, right? AI is still complex for most companies…”

AI Build teams removing the constraint of having to learn the language of computers aka becoming a software engineer that knows how to code while simultaneously empowering subject matter experts aka operators to have a big seat at the table is a VERY big deal…

My man Harrison Chase and the rockstars at LangChain wrote a phenomenal blog post - “Communication is all you need” (micro essay) - that builds on this notion of what AI Builders look like and the importance of cross-functional leadership as a defining characteristic of these teams:

“As smart as the underlying LLMs may become, they will still need the proper context to function reliably, and that context needs to be communicated properly.”

“Communication is not just expression in natural language, but it can also involve code to communicate more precisely. The best people to communicate something are the ones who understand it best, and so building these agents will become cross-functional.”

“One key area is prompt engineering, where domain experts often write the best natural language instructions for prompts, since they know how the LLM should behave (more so than the engineers).”

“Yet, the value of domain experts goes beyond prompting."

"They can review the entire 'cognitive architecture' of the agent, to make sure all logic (whether expressed in language or in code) is correct.”

Revisiting the BCG study, what’s especially ironic is the way the research paints a picture that indicates Algorithms are only 10% and Technology 20% of the challenges faced by Enterprises trying to improve their AI capabilities…while People and Processes are 70%.

Not to oversell the argument but one more data point from the annual CIO and Technology Executive survey by Gartner and their “2025 CIO Agenda” (micro essay):

“All CIOs will continue to deploy emerging technologies such as AI in 2025, but Digital Vanguard CIOs help CxOs identify innovation opportunities, spot AI use cases and enable business-led innovation.”

“Nearly 65% of Digital Vanguard CIOs help business areas identify their own technology skill requirements compared to 22% of other CIOs.”

“In fact, 80% of Digital Vanguard CIOs prioritize changing IT processes and ways of working to enable business areas to work in agile ways, versus 30% of other CIOs.”

What does it look like when domain experts lean into their role as AI Builders on a cross-functional AI team?

A straightforward, high impact example within Financial Services would be the way a Financial Advisor as a Domain Expert and part of an AI Build team would encourage the team to enable the FA to optimize for spending more time with their clients.

It’s not just more time with their clients but the value of those interactions, the insights, ideas etc they bring to those conversations…and the good things that will happen as a result.

“The cyborg wealth advisor” research from CB Insights including the Morgan Stanley results being driven in part by their strategic AI investments, are definitely worth a look (link)…

If you want to go a bit deeper, check out “The Librarian,” an AI Operating System that Union Square Ventures built to effectively create an LLM enabled brain that provides immediate recall, interpretation etc to “around 15,000 articles from the USV blog…a GPT that analyzes investments made by the firm…a tool that makes it possible for team members to interact with meeting notes” (link).

Taking things all the way out to the edge of what is now possible when Generative AI native tools are harnessed by AI Build teams…

John Rush - “Today, I run 11 bootstrapped SaaS projects with just a handful of people. These AI agents & tools 10x busy founders in marketing, coding, legal, operations, design, sales, anything…” (link).

Josh Mohrer - Given his background including launching / scaling Uber in NYC, some think he’s the leading contender to be the first solopreneur to build a unicorn…”I’m taking down OtterAI which has raised $70M in venture. I’ve done all the engineering - I’m a 40+ business operator first time coder - and I’m at $5M ARR. This is happening” (link).

If you’re putting together a Hollywood AI Build team, you most definitely want to consider drafting Ben Affleck as your Domain Expert as his 4 minute AI POV was absolutely lights out…

“AI can write you excellent imitative verse but it can not write you Shakespeare…the function of having two, three or four actors in a room and the taste to discern and construct, that is something that currently entirely eludes AI’s capability and I think will for a meaningful period of time.”

“What AI is going to do is disintermediate the more laborious, less creative and you know, more costly aspects of filmmaking.”

“Craftsman is knowing how to work. Art is knowing when to stop. I think knowing when to stop is going to be a very difficult thing for AI to learn because it’s taste.”

“My hope for AI is that it’s an additional revenue stream that can replace DVDs which took 15-20% out of the economy of filmmaking" (link).

“AI eats the world” by legendary Benedict Evans.

“Every year, I produce a big presentation exploring macro and strategic trends in the tech industry. For 2025, ‘AI eats the world.’” - Benedict Evans (link).

You can’t go wrong reading anything Benedict Evans publishes and I trust you’ll click through the presentation as it’s full of signal, super clean slides that don’t waste any words etc. but below are my 7 favorites with an important question I believe each one answers:

“How should I think about today’s AI capabilities at work?”

“So we clearly need more than just a ChatBot to generate business results?”

“You know what they say about how important it is to define the right problem upfront?”

“Who knows how to answer all of these questions?”

“Who do you think generates more revenue from Generative AI, Accenture or OpenAI?”

“No one and I mean no one knows what’s going to happen next but you have to be on the field to understand the game, you know what I mean?”

“Think of AI as intelligent systems that do what they do in the background so we, humans, can do what we do so much better, enabling our superpowers enabled by their superpowers, you know what I mean?”

Speaking of superpowers, what if Domain Experts could have 10x the impact via AI Applications, tools, intelligent systems?

No doubt, OpenAI CPO Kevin Weil is absolutely right about “the technology floor is not fixed” and “every month or two computers can do something that computers have never been able to do in the history of the world” (micro essay) but that’s even more of a reason to believe that these tools, harnessed by Domain Experts have 10x potential written all over them.

How much more impact could an Enterprise GTM leader, enabled by NotebookLM, generate? (AI with Alec fireside chat with Raiza Martin, NotebookLM PM Leader)

How much more revenue could an Enterprise CRO, enabled by 11x, generate?

How many more insights and ideas could an LOB leader generate, build and validate, enabled by Replit, without being constrained by a technical team without available bandwidth (link)?

The opportunity for AI enabled Domain Experts and Enterprise AI build teams is extraordinary.

Hope you enjoy the rest of your Sunday and talk soon!

Alec