Episode 11: The Reality of AI and Making Bets with HubSpot's Former Chief People Officer and Investor
Hosts Rob May and Brooke Torres interview Jim O'Neill, HubSpot's former Chief People Person and investor, on the reality of AI today and making bets. Tune in to get the perspective of somebody who has made investments, has worked at a company doing AI related work, but does not come from an AI background.
|Rob May, CEO and Co-Founder,
|Jim O'Neill, Former Chief People Officer, HubSpot||
Brooke Torres, Director of Growth,
Rob May: Welcome to the latest edition of AI at Work I am Rob May and I'm here with Brooke Torres as my co-host today. We put this on for Talla which is a new kind of knowledge base for sales and support teams. We merge content, automation and machine learning together, so that when you're looking for information, you can also make that information actionable, you can also find it easier. If you like the podcast, we hope you'll check us out.
Today we have Jim O'Neill here. Jim is an ex-HubSpot executive. He's had a couple different roles and before we get started, Jim, why don't you introduce yourself and tell us about your background.
Jim O'Neill: Hey, everyone. I'm Jim O'Neill. I had the pleasure of working at HubSpot pretty much from day zero through the IPO and have a mutt of a background. I was originally the CTO. Then I evolved to the CIO, which anyone knows is career is over is the acronym for that. Then moved onto the Chief People Person, and really had the pleasure, honestly, of going through the IPO and trying to help figure out the talent and the culture side. Because HubSpot had such a strong culture.That was an amazing role. Then ultimately I ended up doing the traditional entrepreneur in residence as I really tried to figure out where the markets were going around both the HubSpot ecosystem then the larger Boston ecosystem.
Then ultimately transitioned to my own small, boutique hybrid investment operations firm helping early-stage startups with both operational help, meaning go in and give both advice and hands on work as well as evergreen capital, meaning long-term permanent capital without the alphabet soup of venture - which is an interesting view of the world.
RM: Part of what we tried to do today is to bring in somebody who was not an AI expert. We've had a lot of AI experts on this show. We thought it might be interesting to get the perspective of somebody who has made a couple of investments there, has worked at a company that has done some AI related stuff but doesn't come from an AI background per se. With that in mind, how do you learn about AI? How do you approach it when you're a CTO, CIO, whatever, but not an AI person? What are the resources you learned to get started?
JO: I'm going to give Rob a shameless plug for inside.com, because it actually is a great series of aggregation of articles. For the folks that don't read it, I highly recommend it. To be honest, I think I've got a dozen plus people now feverishly reading it every day. I think beyond that, you have to be thoughtful on your approach to research on the internet. I think everyone recommends a couple of the foundational books The Master Algorithm being one or Life 3.0 being another, which are fundamental books that a mere mortal can read and understand a majority of it.
Then honestly I think it's just narrowing in on a couple of thought leaders in this space. I tend personally to, clearly you read what the Googles and the IBMs of the world are doing, but I actually also try to follow a couple of the more edgy startups. I'm not going to name names but just really try to get a little bit deeper so that you have a broad perspective through something like inside.com, a macro perspective from some of the larger names of investing and then a couple of niche perspectives just to see what the possibilities of AI are. In the hype cycle we're probably still early in it, but it's massive. It's definitely a wheat in the chaff problem right now.
BT: What about in your time at HubSpot, because they're pretty forward leaning with a lot of AI stuff, what was going on there with AI when you were there?
JO: When I was there, I think there were a couple of angles from it. First and foremost, HubSpot really tries to build products that are enjoyable. I know that sounds trite or trivial. When you're selling into the SMB world we have to remember that we're selling to individual business owners, operators, marketers, not necessarily CTOs and CIOs.
You honestly have to approach the technology as, how does it benefit their day? At the end of the day, you're trying to save them time or money or improve their general outlook on their customers. How do you take AI and make the process better, make the information better, make the workflows better?
It can be a subtle hint of AI without being AI in your face. I think more specifically what HubSpot did pretty early on was look to find great AI talent. That's an oxymoron in this space, to be honest. It's so hard to find the emerging AI people that aren't already, I'll say, locked up frankly by the big AI big tech companies.
What we did, I'll say the royal we, was we did a few strategic acquisitions of companies largely for the talent, sometimes for part of the IP or the intellectual property. They had to be complementary to what our customers did. We didn't find an AI company that was building machine vision and things like that. We found companies that were looking at sales data, or operational data or buying decision data, and how they were using machine learning or other algorithms to improve that. I think that was the first was strategic “aqua-hires” would be the word that people would use with some IP.
Two, was HubSpot did actually a really good job from my point of view finding some great talent and building small centers of excellence that, again, they don't talk about it a lot. But, for instance, in Dublin one of the lead Google AI engineers HubSpot was magically able to convince to work at HubSpot. He's built out an amazing group of AI people around him.
What I like about that is HubSpot doesn't shout from the rafters that it's an AI based product. What they're really focused on is how do we take the talent we've grown, the talent we've acquired, and then the IP that's becoming more available in this space and wrapping those three up. There's really two views on that. One, is how does it make the product better? How does it make it better for their customers to use the technology? Then two, how does it make the company better? How do you internally use artificial intelligence to make the employee's experience better?
I think they do a pretty good job with that. If you look at the press, you never see much talk about it. Because they think in their minds it's a bit more hype than it is reality. They're trying to make it as real as possible.
RM: Tell us a little bit about what you've learned about rolling out AI? I don't remember if you were involved in this, but two and a half years ago, one of our very first beta customers was HubSpot. We had this recruiting workflow AI tool. We went over and I think it was Becky that we worked with. After about six weeks I called her and I said, “Hey, what's going on?” She said, “You know, it actually makes our workflow harder. The AIs make it harder not easier.” These things are tough to roll out. We learned a lot from that and it's influenced how we built the Talla product today. Have you guys had other similar experiences?
JO: Yeah, and I’m smiling, people on the radio and the internet can't see that. The reason I'm smiling is one of my personal mental conflicts when I was at HubSpot was, I was starting to become an active angel investor. I had always made a very bright line that if I was looking at a company to invest I had to recuse myself, not to use a political joke, recuse myself of actually participating in the implementation. Talla was in my sights at the time. So, Becky clearly took the lead on that.
To answer the question, I don't think it's unique to Talla. I think there is a buying mismatch between what AI software and the current state of the art requirements and what people expect when they're trying a product out, specifically from a tech startup. There's this almost competing time investment, perceived or real, that it's more enterprise like than traditional startup software, which is more consumer like. I don't think those lines have crossed yet.
When you think of the investment, I think there's still a mystery around the time investment or the complexity. What I don't think the AI world has done from the layman's point of view is educated people enough on the initial time investment versus the long-term reward. You almost think of it or I think of it like a reverse bell curve, where there's this really big time investment up front. You actually don't get really great result until you come through the other end of that bell curve. Then the results can be phenomenal.
I just don't think companies, because their people, and not HubSpot specifically, I think most tech companies are people are the most valuable asset they have. So, any time investment that they make they can probably think of a one to three month payback. I think it's really hard to think of a six, 12, 18 month payback. Alternatively, they'd invest those people into their own tooling and their own system.
I think there's just this competition, frankly, for resources. Now, that may be changing as AI becomes more mature, the datasets become more established. My take on AI is initial use takes the initial feeding of information to make it personalized enough to make it meaningful enough. I just think those things are at odds today. I think that was HubSpot's experience from what I recall.
RM: We had that experience with a couple of early customers. What it led us to do was really redesign the products such that it could be deployed and show value. A lot of what you do is, we do a similar thing, you put a human in the loop.
You start with the human in the loop and you say, “I'm going to improve this workflow and I'm slowly going to eat into what the human has to do.” That tends to be a way that the startups can get around that.
JO: If I can add one thing. I think that's fascinating. It's not a great example with the Mechanical Turk of AI. I don't think enough companies are willing to do that, frankly. I think they rely too much on algorithms. A lot of the businesses I see, the more successful ones have a small army of people actually curating information and they have to do it in a timely fashion. I think the best example would be Fin. If you guys all use Fin for personal assistant, that's probably the more well-known one.
I've seen companies in real estate, tech, in lending teck, and a lot of those kind of technologies that are using AI. But truly they have a services component just making sure that the time to enjoyment is reasonable and keeps people interested.
RM: The problem with that model then is that while it's better for your customers, which is probably why you should do it, you're venture capitalists go, well, there's no AI here.
JO: Capitally intensive.
RM: You get it on that side so pick your poison when you're starting a company. Anyway.
BT: Jim, do you look at a lot of AI? Because you do a good amount of angel investing, do you look at a lot of the startups yourself for those investments?
JO: I tend to shy away from them now, to be honest. I think, in full disclosure, I invested in Talla early on. I think my learning for that process, and clearly Rob had a great reputation, I knew him a little bit, was that it was early and he needed to place some bets. I think that was a great bet. I hope to see that come true. As you look at other startups, I actually shy away on multiple fronts. If they lead with, for an AI based, machine learning, buzzword compliant, blah, blah, blah, blah, I don't want anything to do with it. I guarantee you they don't have a product. I just run away.
Now, if the pitch is much more around the problems they're solving and how they solve it - by the way, we do have machine learning, AI or some variance of that, that's of interest. Because that shows consumer improvement, which to me it's around what AI can benefit is how how's that consumer experience? What's the benefits? What's the reduce in friction? So, that's the two axes I look at.
Third, which is really related to number one, and I stopped investing in this space specifically because I was unclear on who was writing algorithms or who was using algorithms. I was trying to use that as a decision point for my own investing. I think I actually made a big mistake where if you didn't write the algorithm, I discounted you because I figured you were just using algorithms if you didn't really understand AI. What I've come to learn is that that may be too binary. The fact that most of the people writing algorithms are building general purpose algorithms, where many people can use.
I thought it was a little too complex, as Rob opened up, I'm not an AI expert. I don't have a CS degree in AI. I know enough to be dangerous. For me it was simpler to say, if it's a supporting capability, I'm very interested. If it's a leading capability, I'm uninterested. As an individual investor I'm not going to get leverage out of that, and I'm not going to understand it.
BT: I think it's interesting. Even on the marketing side we've seen this where whether you're talking to investors or you're talking to buyers, being too AI focused is a flag that you're not solving a problem. You're either bringing in the wrong leads or whatever, so the focus on the problem is a big thing across the industry.
RM: As you're out doing this new thing, making investments and helping operationally with these companies, what are the applications or use cases of AI that excite you most either that you want to invest in them or you wish you had them to apply at some of your companies or your personal life? What kinds of things are you thinking about?
JO: Improved customer or employee, employee being a customer experience, probably first and foremost, reducing friction in any process. I think, personally, people will enjoy talking to bots if the bots are actually solving a problem, and improving their experience with the human frankly. I know that might be a little totalitarianism. We've all had bad call center experiences. It still makes us all crazy. We're in the 21st century and we're still dealing with it. Yet a bot, nine out of 10 times can get the right answer.
RM: We haven't had the Drift guys on here yet. But when we do, I'm going to push on Elisa David, because they're really big on conversational marketing, conversational marketing. People want conversations, whatever. I think people want answers. People want results. And if it's a conversation that's fine. If it's not, it's fine. I'm not sure they want conversation for its own sake. We'll see what their take is on that.
JO: I think that's the first one is just making these experiences better. I think on the second, it's a broad range of scale. Just doing things that humans may not be good at doing or don't enjoy doing anymore, and the labor market is showing that. Where AI can truly change the output of those efforts. At some point, we're just incapable or unwilling as a species to want to do certain things. You can just continuously throw the CPUs at it.
The things that I get fascinated by is going back and looking at all the images from space and re-analyzing for artifacts and finding things that human hours just couldn't do. It was a function of man and woman power. That stuff really gets me excited. We're learning patterns and information from data that we already have. We would never have put the resources against it.
As far as startups go, when startups are doing that, like geospatial analysts, all those kind of things, I get very excited. I think it can actually yield better results from stuff that we already have without, honestly, wasting more resources.
RM: One of my angel investments is a company called Smartvid. They analyze construction images or construction sites. They look for safety violations. This was a thing that, again, humans have to just look at them. You miss some and whatever. Now they can basically tell the human, look at these 5%. It's changing, I think, a lot of industries, it's early on. The folks that are adopting it, embracing it are seeing great results.
BT: What are some of the lessons that you've learned working with tech startups in this space or even outside of the space that are deploying AI?
JO: As far as tech startups in and around AI, there's the dangers of the platform approach. I think any super high growth tech company wants to be a platform. We all have aspirations. Any entrepreneur has aspirations to build a platform. I think in the AI space, again, unless you're taking a really broad approach at the Google scale or the deep learning scale, it's really, really hard and expensive to compete in that space.
I mentioned earlier, I really shy away from anything that's this just broad, ambiguous platform. Rob and I were talking a few minutes earlier of there was lots of platforms even three years ago that were trying to build the AI conversational bot space. Everyone could just build bots on top of it. Because it's so broad and vague, frankly, it's really hard to build specific use cases.
I think those companies have had a hard time breaking out, to be honest. That's one. I think two, and it's evolving and the AI stuff I do get excited about is data population of the datasets for AI. That has been one of the barriers, I think, in AI adoption in general is you can write the best algorithms but the learning information to feed those algorithms, no one has in my mind, and again I'm not an expert, has really tackled the authentic data generation or data collection. They're finding datasets and scrubbing them and then putting them in. Then kind of looking at the results. I think there could be a whole market in doing nothing but generating data for algorithms. I think there's probably people already doing that, other venture community's funding it. They just haven't made it to the headlines yet. So I think I shy away from companies that don't have great datasets in the absence of somebody building a company to generate these datasets, which I think would be interesting. Then I think third is I believe startups are learning to not overhype their use of AI. That's something you heard me say earlier. I do get very vocal with any investments I've done is let it be something they ask you about. Don't lead with it. I think it creates for uncertainty in both your product, your valuation, your ability to find talent. I think it's easy to get people excited about it. I think it's hard to get people to really do it. I still think we're on the knowledge curve very, very early on in being able to deploy it.
RM: One of the things that we learned early on was we actually did some tests on Google AdWords around using AI driven language for people just doing general searches for a related product, and then using non-AI related language. The non-AI related language clearly won. AI scared people away. It's changing a little bit I think now. I think what you want to do if you're an AI company right now is you want to catch people by the way you would always catch them, searching for their problem, not mentioning AI, whatever. But then when they hit your website, you want to say, “By the way, we're new and different than everything else, that you're going to see, because we have this AI. Here's how it works in AI. Here's how the AI solves your problem in this way.” That's the strategy that we've seen work the best.
JO: That makes sense.
RM: Good. Well, we're running out of time here. We should end with the question that I always end with, which is, people who are in the AI field tend to have strong opinions about whether or not we will ever get to generalized AI. And so TBD if that happens. But there's this big debate, Elon Musk versus Mark Zuckerberg where Elon Musk says, this is the biggest threat to humanity. Mark Zuckerberg says, quit whining. It's not that big a deal. Where do you fall in that spectrum?
JO: Oh boy. That's a great question. I'm going to be a little bit influenced. I was watching a somewhat old conversation between Peter Thiel and Marc Andreessen from a few years ago. It was actually five years old, but the topics were as relevant as ever. It was this exact question where I think Peter Thiel did an interesting job saying that we have to stop villainizing technology at large. Every Hollywood movie it shows doomsday and technology is the end of the human race. He's like, where's the positive side of that as far as the advancements we've made? And so I think that's going to bias my response a little bit. I don't think there's anything wrong with that. I think it's ours to screw up, which humans are good at screwing things up by nature.
I think in general the quality of life has improved greatly over the past generations, be it, I know people say earnings are flat. There's a whole bunch of smart people figuring all those things out. I generally look around, at least in the small worlds that I see, and people are generally happy. The quality of life seems to be a very good. We're not dying in our 40s. We're not worked to death. There's probably pros and cons at a much higher level.
With that, sure. There could be some awful things that we create. But there can be some amazing things that we create. I'll take the amazing road any day. And so more power to the people that figure it out. Hopefully it makes humanity better.
BT: Awesome. Well, Jim, thank you so much for joining us. We will be back next week with another episode of AI at Work. You can find us online at Talla.com