How Suraj Kripalani and BonBillo are Using AI Agents to Accelerate Global Social Impact - Ep 27

:: 00:00

Suraj, welcome so much to the podcast today.

:: 00:02

I'm really excited to have a conversation with you, just looking over all the stuff that you're doing at Bombillo.

:: 00:07

Man, there is a lot of agents and a lot of impact that you're making across the world.

:: 00:11

So as I like to start all of these conversations, just open the floor to you and kind of tell us your origin story, how you came from BlackRock and managing income portfolios to kind of founding Bombillo and kind of the mission that you have there.

:: 00:24

Fantastic.

:: 00:24

Thanks for having me, Kyle.

:: 00:26

I was with BlackRock for the first eight years of my career.

:: 00:29

This was working across roles in risk management, portfolio management, focused on fixed income across New York and London.

:: 00:37

And I left BlackRock in 2012 from London, pursued a more entrepreneurial path.

:: 00:43

And I moved to Bombay.

:: 00:44

I had an edtech startup idea, gave it a go, and it took a year before me and my co-founder decided to shut it down.

:: 00:54

And I made all the rookie mistakes that a first-time founder does.

:: 00:58

And so to a certain extent, building Bonvilo for a younger myself as well.

:: 01:04

We kind of built product before we validated demand.

:: 01:08

What we learned in a year, we help our teams learn in four to six weeks now.

:: 01:13

And then really continued staying in edtech, worked with a non-profit organization that was looking to bring Khan Academy technology to low-income schools in India.

:: 01:24

That was a fantastic opportunity for me.

:: 01:27

That team has gone on to scale and serve schools across the country now.

:: 01:32

And I got it off the ground for them, transitioned to a team, and took the non-profit route to scale as well.

:: 01:40

I then spent three years working with a boutique consulting firm out of Bombay and working on transforming and growing companies across the country, very similar to a private equity model.

:: 01:51

We didn't have a fund

:: 01:54

ourselves.

:: 01:54

We often partnered with external funds to bring in capital.

:: 01:58

And I really enjoyed my time in consulting.

:: 02:00

The one thing I started to miss towards the end of the three years was the strong alignment of purpose with work.

:: 02:06

I was also ready to take a risk myself and build something new.

:: 02:10

And I had a couple of social impact startup ideas, but decided to dive deeper into the startup space before I started something next.

:: 02:19

with a little bit of research, found an incredible program out of MIT.

:: 02:23

This is MIT's flagship entrepreneurship development program.

:: 02:28

They work with 100 founders in their program, and it's a one-week-long program based on a repeatable process, which is called Disciplined Entrepreneurship.

:: 02:39

It's a book as well written by Bill Olett, and he and his team run a fantastic program, which was quite game-changing for me.

:: 02:46

And so I went to the program, and it basically

:: 02:49

led me to a certain extent to starting Bond Below.

:: 02:53

Doing a startup is hard.

:: 02:54

I've been down that journey more times than I care to admit myself.

:: 02:57

But you failed at it, but you didn't see it as a failure because you saw it as a learning opportunity, right?

:: 03:02

But you also saw it, I love the aspect of like, I learned this stuff, how do I teach this to other people, right?

:: 03:08

And then getting into the MIT part of the journey, like an education institution here, like, all right, how do we spread this wider?

:: 03:16

That's such a cool, like,

:: 03:18

reaction or response to that instead of just, I love to say we learn more when we fail than when we succeed, right?

:: 03:23

And it sounds like you're absolutely went through that process.

:: 03:27

Yeah, absolutely.

:: 03:28

So one of the things that attracted me to go to MIT was just hearing that they have a repeatable process to help startups launch and grow.

:: 03:37

And so for a consultant hearing that there's something that works, their own ecosystem has a remarkable success rate applying disciplined entrepreneurship.

:: 03:46

And so I go there and spend that week, and it was really transformational for me in terms of the content, the faculty, and the Boston startup ecosystem was exceedingly welcoming.

:: 03:58

And so I took a few weeks essentially to meet with founders, investors, as well as startup accelerators across Boston.

:: 04:06

And I basically had coffees with everyone in the startup community over the next four weeks.

:: 04:13

A couple of things came out from that time.

:: 04:15

One is, this is back in 2017, there was no leading startup accelerator focused on social impact.

:: 04:23

That was one.

:: 04:24

The second is that speaking with founders that had been through top accelerator programs, they all led with the key value that they got from the programs being access to capital and networks.

:: 04:38

No one really led with the fact that they'd been through an amazing training program that made them more confident about building their own startup.

:: 04:46

That was always an afterthought.

:: 04:48

So it was clear that the top programs really focus a lot more in terms of access to both networks and capital, much less in terms of a structured training program that improves success rates.

:: 05:00

And so the thought was MIT has already figured this out.

:: 05:03

They have a repeatable process that works.

:: 05:06

Can we build on top of MIT's disciplined entrepreneurship framework to provide mission-driven founders with the support, capital, access to networks, as well as training that they need to be successful contributing to the UN Sustainable Development Goals?

:: 05:23

And so that's how Bunbelow really came about.

:: 05:26

We ran our first pilot accelerator program in the fall of 2017 out of Boston.

:: 05:32

That was a three-month accelerator.

:: 05:34

And then

:: 05:36

Based on the results of that, really launched our first micro BC fund back in 2018.

:: 05:42

And so that's how we got started.

:: 05:44

We now run programs across Boston, Cape Town, Bombay, Bangalore, Singapore, and Bali.

:: 05:50

And we're very excited about their next phase of growth to support mission-driven founders globally.

:: 05:58

I think it's important, if you wouldn't mind, like define, if you would, like social impact.

:: 06:03

Like what do you mean by that so that the audience, because it's one of those terms that it leans a little bit different things to everybody.

:: 06:09

So I want to make sure that we kind of have that clarity.

:: 06:11

That's a great question.

:: 06:12

The one thing that we wanted to do was to be more inclusive with what we call social impact.

:: 06:19

And the world at the time had largely coalesced around the UN Sustainable Development Goals.

:: 06:26

governments, nonprofits, as well as private players.

:: 06:29

And so the thought was, can we really screen for startups contributing a more sustainable world as a whole, contribute to the UN Sustainable Development Goals?

:: 06:39

And then we picked 5 themes that we want to focus on that we believed captured the spirit of the goals as a whole.

:: 06:46

So our top five themes, the first is education and jobs.

:: 06:51

Second is food, health and wellness.

:: 06:53

Third, sustainability.

:: 06:55

both financial inclusion and last gender equality and diversity.

:: 06:59

And I think everybody kind of realized like you need all of these things to have a functioning, profitable society, right?

:: 07:06

Like they all go together.

:: 07:08

And yes, like businesses come out to be profitable and we all know the capitalistic way, but like to be sustainable long-term, you need to think about all of these things and how they all kind of go together.

:: 07:19

Right.

:: 07:19

But like, so, you know, our audience,

:: 07:22

is definitely kind of like, all right, tell me about the AI stuff.

:: 07:25

So like, you've been doing this for seven, eight years, right?

:: 07:29

And in the last few years, AI's really come into play, specifically kind of the agentic piece.

:: 07:34

So talk about how, like, tie that together for everybody, right?

:: 07:38

How did you get into like, all right, we could start using AI to help in these startups, or we could use it to kind of scale the knowledge we've learned.

:: 07:46

Tell us how that transition kind of started.

:: 07:48

Yeah, super.

:: 07:48

So

:: 07:49

It was towards the end of last year, that's November, December, 2024, that I realized that the space is moving really quickly and that I need to come up to speed with what's happening within AI.

:: 08:02

And so I essentially took a couple of weeks off to focus on this towards the end of last year, gave myself time to play with tools such as Replit, Cursor, Lovable was fantastic, as well as

:: 08:17

as well as Agent.AI.

:: 08:19

The one thing that became very clear very quickly is that these tools had gone from being fairly sort of in the early stages, I would say basic, to now very powerful for a non-programmer, right?

:: 08:34

And so they become super easy to use.

:: 08:38

And essentially anyone that took the time to pick it up could build anything they wanted.

:: 08:43

And then within Agent.AI,

:: 08:45

That particularly just seemed like a very intuitive platform that was also very powerful.

:: 08:53

And so I took some time to play with the templates that were on the platform already.

:: 08:59

And I quickly saw that I could chat with a website, I could chat with a document, and just that was sort of value added in the very simple templates that you could then extend and build to your use case.

:: 09:12

And so the thought was, my God, this is like game changing for what can be done.

:: 09:18

And so that really led me to then say, okay, can I spend a little bit of time tinkering and build something that is valuable to me and my team for some use case that we have this week?

:: 09:30

And that led to the first set of agents really getting built to solve our problems that we had at the time.

:: 09:36

And I'd say the one big thing that it kind of led to is it got

:: 09:41

me and the team to reflect on our own product roadmap for 2025 and think about ways that we can integrate AI into it, which then led to being sort of an AI-first accelerator platform that actually leverages AI across the board to help our startups grow.

:: 09:59

And so it was really starting with this very hands-on tinkering that then opened up a set of insights that we could leverage from a strategic standpoint.

:: 10:08

And what I think is really interesting is

:: 10:11

Especially with startups, it's not necessarily that a lot of the things that startup founders need to do is hard.

:: 10:17

It's just, they don't know what they don't know.

:: 10:19

Ignorance is bliss, but ignorance is also like, I don't know these things because it's such a generalist palette of like, what is your value proposition?

:: 10:26

Social impact, building your team, sales and marketing, product design, like all these different things that go into building a successful company that people are subject matter experts by nature.

:: 10:38

So somebody might be a software engineer, but they don't know anything about like identifying personas or coming up with a sales process.

:: 10:44

Exactly.

:: 10:45

Or that's why they come in teams.

:: 10:48

But there are these kind of, you've got all of these tools here that go into the different pieces that make you think about all those different elements, which I think is super, super powerful and interesting.

:: 10:56

And you've got years of knowledge base and training and stuff to kind of really help kind of go, like know where to go.

:: 11:04

So expand upon that for me a little bit.

:: 11:06

Like how did you decide where to go first here and what is the reaction you're getting from people who are kind of coming in and using these things now?

:: 11:14

There's something that you hit upon there that I think is extremely powerful is that one thing we found building agents is that for subject matter experts learn how to build, that becomes a very powerful skill set because now they bring their deep expertise

:: 11:33

into what an agent should be able to do, how to do it, and what does amazing look like versus what does good look like, right?

:: 11:43

So you can use ChatGPT today and get, I would say, fairly good results for most things that you want to do in the world.

:: 11:51

It might hallucinate once in a while, but for the most part, you'll get fairly good results.

:: 11:56

The one big difference between an agent that has been trained for a particular use case is that

:: 12:02

You're then, hopefully, as a subject matter expert, you bring a point of view.

:: 12:06

You bring very specific things that you want this particular agent to do that leverages your experience as well as expertise.

:: 12:14

You set really strong guardrails around what it should not do so that you get reliable results, you don't get hallucination, and you are able to test and make sure that it consistently gives you excellent results better than you would get using just a generic LLM.

:: 12:31

So that I think is the powerful unlock that we as a team have found.

:: 12:37

Very proud of the whole team building agents across the board now, which is why we've reached the point where we have 30 plus and looking to continue shipping every week, really.

:: 12:47

And I'm going to do a shout out here to Julia, who we interviewed a number of months ago.

:: 12:52

Go watch that episode where we dive super, super deep into fundraising through grants and stuff.

:: 12:57

It's just one tool, but

:: 12:59

I'm curious, right, you've got all these agents and like you said, you have a bootcamp, you have a training program where you take somebody through something in four to six weeks.

:: 13:08

How do you weave these agents into that?

:: 13:10

Is this all part of the curriculum to like use certain agents in certain steps as they build up their mission statement and all of that of their business and their marketing plan and their pitch deck?

:: 13:22

Do you weave all these into that process?

:: 13:24

Yes, absolutely.

:: 13:26

So we've got a two-week accelerator program that we run, and we now also have a community model and a growth partner model.

:: 13:35

The sort of community model is meant to leverage the 30-plus agents alongside Bonbilo's mentorship as well as peer learning.

:: 13:44

And the growth partner model teams get the benefit of Bonbilo's hands-on support as well.

:: 13:52

And then

:: 13:53

The accelerator model is where our fund makes an investment into startups that we work with, and they leverage Bond Below's curriculum platform, and they pitch at the end of two weeks.

:: 14:04

So these are the three types of programs that we run right now, and all three leverage Bond Below's AI agents, which are essentially specific milestones within our curriculum.

:: 14:15

And so really starting with customer discovery,

:: 14:18

identifying who is your target market as well as customer, what are the key pain points, triggers, priorities, and then really building on top of that.

:: 14:30

And to a point where you get to crafting your pitch deck as well as a fundraising strategy.

:: 14:36

The sort of end-to-end, the 30 plus agents are all mini milestones.

:: 14:41

that then unlock, I would say, a very solid growth roadmap as well as plan and a fundraising strategy for startups to be successful.

:: 14:50

We all know everybody learns differently, right?

:: 14:52

Like, and it's different than reading about this thing and then like, okay, that's great, but how do I apply that to what I do?

:: 14:59

And through these LLMs and these agents, like, okay, let us go ahead and give you a starting point.

:: 15:05

It's not going to be perfect, but if it gets you 80% of the way, it sparks that imagination, I imagine, for a lot of these companies.

:: 15:12

Like, oh, I get it now.

:: 15:13

I didn't just read it.

:: 15:14

Like, I get what you're looking for, the right sort of output and the right sort of things.

:: 15:19

No, this isn't quite right.

:: 15:20

Let me tweak that.

:: 15:21

I imagine you see a lot of that stuff, right?

:: 15:23

Absolutely.

:: 15:23

And to share an example there, for instance, we have a financial projections agent.

:: 15:29

that helps the startup craft, I would say, relatively high quality projections for the next five years, which just as a milestone to build yourself, takes a considerable amount of time.

:: 15:41

Say maybe like for someone that knows what they're doing still takes a considerable amount of time.

:: 15:46

And if a founder is doing this for the first time, there's a lot of learning that happens as well to get to a high quality output.

:: 15:52

And so to be able to start with

:: 15:55

something, as you said, 80% plus, we're trying to get to a point where it's not of that.

:: 16:01

And then for the founder, along with our team, to be able to build on top of that, to come up with something that works for the founder, right?

:: 16:09

And to be able to go from like a few hours of doing that to maybe half a day to a couple minutes to a starting point is just super powerful.

:: 16:17

And so that's what we found that

:: 16:20

starting from something that is already high quality as opposed to a blank canvas actually unlocks creativity because then you have these set of ideas that you can build on top of.

:: 16:29

And we've found that our founders are quite excited about what we built.

:: 16:33

Yeah, it just makes so much sense too, right?

:: 16:36

Like it's like getting in there and getting you part of it.

:: 16:40

But once again, you still always, and I feel like a broken record on these shows saying this, but it's still human in the loop, right?

:: 16:46

Like you still have to have the human that's reviewing and say, does this

:: 16:50

makes sense?

:: 16:50

Is this right?

:: 16:51

Does this really know our business?

:: 16:53

Let me clean it up.

:: 16:54

But to get the heavy lifting of like putting a framework around it with some mostly right data or answers or number projections in place, it's just super helpful for anybody.

:: 17:04

Because y'all have built so many of these things and many of them have been like agent of the weeks, won awards through the Agent AI platform.

:: 17:13

How has all of this changed the way that you approach?

:: 17:16

Do you take larger cohorts of

:: 17:20

of startups through the program now, it sounds like 2 weeks now for a bootcamp where it was longer.

:: 17:24

So you're able to like compress that timeline now.

:: 17:27

How is this changing everything that you're doing in the last year compared to what it was, you know, three, four, five years ago?

:: 17:33

Yeah, so what we're most excited about is to be able to serve more startups and to help them with their growth as well as fundraising.

:: 17:42

And the whole agent platform enables us now to serve a lot more teams than we could.

:: 17:49

without it, obviously.

:: 17:51

And we're thrilled to launch our community model as well as growth partner model.

:: 17:57

We think that with the community model, really, we can serve founders around the world that are contributing to the UN SDGs and to give them a platform to grow way faster than they could without it having to figure out everything for themselves from scratch.

:: 18:15

This is built on proven frameworks, really, tested.

:: 18:19

gets the benefit of our time and mentorship supporting them.

:: 18:24

So definitely, I think the human in the loop is a key choice that we made as a team, that we don't want these things to be guiding founders without us.

:: 18:33

What we found is that at the end of the day, the founders know their company the best, or I would say that they know the

:: 18:40

problem, the customer, what they're looking to solve better than anyone else, really.

:: 18:45

They have a vision and we're there to support with that.

:: 18:48

And that perspective is, I think, irreplaceable.

:: 18:52

But then giving them the toolkit of proven frameworks that allows them to solve that problem best and build a company with a high growth venture-backed model around that, is something that I think

:: 19:05

we can really enable at scale with the AI-driven platform.

:: 19:10

Let me ask you this.

:: 19:11

Going through this program now, especially using these agents to do it, this is selfishly curious.

:: 19:18

How many of these founders now say, oh my gosh, I get it.

:: 19:21

I see.

:: 19:22

where AI is going and how it can be useful.

:: 19:24

And how does that impact what they build and how they roll out their programs, their software, their companies, trying to leverage more of these AI solutions to scale it up or go faster or optimize?

:: 19:36

Like, are you seeing more of that now because you've planted a seed with them?

:: 19:40

Yeah, so I'd say there's definitely been a wow moment or an aha moment when they see the first set of agent results, right?

:: 19:49

And so then just getting them to

:: 19:51

to see that and appreciate that and be like, my God, you saved me so much time and so much pain getting to the same output.

:: 19:58

And now we have something like pretty amazing that we can discuss as a team with you and see how we can refine this, take this up a notch.

:: 20:07

But there's definitely the strong aha moment and a sense of like, wow, this is what we need to do, right?

:: 20:13

And so I would say we've been working with a few teams that are part of Banbilo's growth partner model that get the benefit of our sort of execution support as well in just a few hours a month to help them grow.

:: 20:26

And with these one-on-one calls, we try to achieve or refine a milestone each call.

:: 20:35

And I would say

:: 20:36

being able to leverage these very specific agents has been game-changing with what we can do and how fast we can move and add value.

:: 20:46

And so that is definitely something that we're seeing that our founders at least are, I would say, really appreciative of.

:: 20:53

and are certainly taking that into account as they think about how to build their own startups.

:: 20:58

Sure.

:: 20:59

And how can they sort of leverage AI across the board within their own startups?

:: 21:04

What about, we talked a little bit about this before we hit recording, but I want to dive into it because I think it's a really interesting angle too.

:: 21:11

Knowing that you support teams internationally all over the world, right?

:: 21:15

There's definitely cultural, regional context that's important.

:: 21:20

How did you feel like that impacts

:: 21:22

these agents.

:: 21:23

Do you tweak them to be more appropriate for Southeast Asia versus Latin America versus North America?

:: 21:30

Or do you find like that happens after the fact that you work with these startups?

:: 21:35

But universally, that 80, 90% of kind of where the agents get you is pretty much universal across the globe.

:: 21:42

Yeah, so I'd say largely the types of agents that we've built so far, they've built on proven frameworks to

:: 21:50

to launch and grow startups.

:: 21:52

And those stay the same regardless of which part of the world you built in.

:: 21:56

And so what we found is that they give remarkably high quality, consistent results, regardless of geography.

:: 22:05

And the one thing though that I would say is that the sort of geographical nuances certainly come into play when you're thinking about fundraising, valuations,

:: 22:18

what is reasonable in a particular emerging market region versus the US.

:: 22:23

So as we build more in terms of sort of valuations and fundraising, that is something that I think comes into play more at that point.

:: 22:31

But in terms of really sort of growth, product, customer discovery, those things in terms of best practices largely stay the same

:: 22:42

across regions.

:: 22:43

That's interesting.

:: 22:44

And you also got me thinking bigger picture too, right?

:: 22:46

All of these agents you've built fit into that, what we call go-to-market stack, right?

:: 22:51

Like it's startups are trying to figure out go-to-market for the first time.

:: 22:56

So any go-to-market leader, any company, I'm sure could find some of these agents would be useful to what they're trying to solve or figure out now.

:: 23:04

So that's a really interesting, like you said, if it works globally, then I'm sure it works in different companies that already exist and maybe already

:: 23:11

trying to figure out their new product they're trying to go to market with and can kind of take advantage of some of these things.

:: 23:16

Yeah, so I'd say what was remarkable for us to see, to your point, was we've got a couple of very specific agents that gives you your top 50 customers as well as top 50 partners.

:: 23:30

And what we found is that, quite interestingly, that worked very well for our US-based startups, of course.

:: 23:38

But we tested with startups based out of Indonesia, for instance, and for a B2B startup based there, it actually gave a lot of their sort of initial customers, and then it gave them their next list of customers to go after.

:: 23:50

That's cool.

:: 23:51

Yes.

:: 23:52

So I'd say that in terms of the quality works across the board, and from a go-to-market standpoint, I think everyone that works in go-to-market now, this just becomes

:: 24:03

a pretty powerful skill to be able to leverage, to move faster to think about ideas maybe that you haven't thought about.

:: 24:11

The one thing that we have learned with sort of building is that one is it unlocks creatively to build on top of ideas that you get.

:: 24:19

But the second thing is that quite often you have a direction and

:: 24:24

where you want to go with a certain milestone.

:: 24:28

But what an AI tool gets you to do is it gets you to make sure that you're not missing any points that you haven't already thought of.

:: 24:38

So I think it's quite comprehensive.

:: 24:41

And so it gets you to just think about like, have I missed anything with this particular milestone that I should be thinking about that I haven't yet?

:: 24:49

so I'd say it works both ways, both from a discipline standpoint of making sure that, okay, that I haven't missed anything basic that I should think about.

:: 24:57

But then really, how can I be creative to build on top of this as well?

:: 25:01

That's such a good point, right?

:: 25:02

And I even said it earlier, right?

:: 25:04

Like, ignorance is bliss.

:: 25:06

And there's all these things that we don't know what we don't know.

:: 25:09

And if you're so focused on going on a specific path with these best practices and you read some playbook or something,

:: 25:15

just getting a second opinion, right?

:: 25:17

Like we've got a problem.

:: 25:19

Like we're sick or something and you go to the doctor and he tells you this is wrong.

:: 25:22

Like if it's a major thing, you're going to go get a second opinion.

:: 25:26

Like this is no different than that.

:: 25:27

And it's funny too, hearing you talk about pulling this list of customers, like that is account-based marketing is what you're talking about.

:: 25:33

This whole ABM term that we see so many, you know, go-to-market marketers talk about now, but it still starts the same way, right?

:: 25:40

And here's a list, here's a tool, here's an agent that can kind of

:: 25:44

think out-of-the-box in a different perspective for you.

:: 25:46

And maybe it finds some target accounts that you didn't think about that you can add to your list.

:: 25:51

There's no reason not to do something like that because the barrier to trying it is so low now, right?

:: 25:56

It's 5 minutes.

:: 25:58

Yeah, absolutely.

:: 25:59

Absolutely agree.

:: 26:00

Let's talk about like, where do you see this stuff going, right?

:: 26:03

Like clearly in the last year, you know, you just started playing these agents in the last year and it's fundamentally changed.

:: 26:09

how you're doing on your training, your coaching, your teaching, you're working with entrepreneurs, you're advising them.

:: 26:15

Where do you see this going in the next year, two years, five years?

:: 26:20

Can we even see five years out?

:: 26:22

Clearly, this is going to become a more and more of a picture as more and more people embrace it.

:: 26:25

But if they're already starting to see the light bulb now, where does that go?

:: 26:29

So I think in terms of with sort of leveraging AI to grow, one of the things that I think becomes

:: 26:36

more and more important from a differentiation standpoint is how do you maintain memory and context to maintain personalization and build on top of where you are already, right?

:: 26:48

To really become a founder's partner at the end of the day.

:: 26:52

And so that's something that we're excited about, Agent.AI, as well as HubSpot's roadmap.

:: 26:59

And so I want to get there sooner rather than later.

:: 27:03

I would say that a lot of the

:: 27:06

The agents that we've built already, they're well-trained to solve a particular milestone.

:: 27:12

Where possible, they leverage sort of knowledge bases as well as context wherever possible.

:: 27:20

But if we can extend that to have sort of selective memory, that would just be exceedingly powerful.

:: 27:26

And I think that is the sort of direction this is going in.

:: 27:32

which sort of really leads to a level of personalization that is quite incredible, I think.

:: 27:38

And so excited about that from a product standpoint.

:: 27:42

What you're saying is very important, but I want to make sure I'm understanding and the audience is too, is it starts with you need good data, right?

:: 27:50

Like garbage in, garbage out.

:: 27:51

And if you're not getting good data, you can't do these good things.

:: 27:54

But then, and I haven't ever heard or thought about it this way, but now you're talking about like permissions around that data.

:: 28:01

right?

:: 28:01

Depending on maybe the user or the role that they have, that, all right, you've got this database or LLM or however it's structured, unstructured data, but you can like permission off certain things for certain queries or certain users and things like that.

:: 28:16

So it's kind of like you're in your CRM and you give salespeople only access to the contacts that they own.

:: 28:22

But is that kind of where you're going with this kind of permissioning thing?

:: 28:25

Kind of.

:: 28:25

So it's similar.

:: 28:26

So I'll share an example maybe.

:: 28:29

Let's say we have a mentor call with a startup that is coming up with their five-year projections.

:: 28:35

And we kind of have a call where we sort of align that the five-year projections that we came up with, we like, and that we want to work towards achieving, right?

:: 28:47

And so now we have a sort of complete milestone that hopefully the next time we chat, we build on top of, and our AI tool remembers exactly that.

:: 28:58

And this tool maybe doesn't need to remember all the things that we rejected.

:: 29:02

Maybe it does, that's our choice.

:: 29:04

But at least it should be able to remember what we wanted to selectively remember.

:: 29:10

And because then you can continue the conversation exactly where you left off and build on top of that with much greater context than if you were starting from scratch or if you didn't really have that sort of memory capability.

:: 29:25

and you would have to be reminded or sort of that context would have to be re-shared with you to a certain extent.

:: 29:31

But I love the concept of sort of selective memory as a whole, because then you can really leverage AI with the memory that you want it to have rather than everything that you ever discussed, which also becomes really expensive over time.

:: 29:44

So that's kind of where from a product standpoint, we're excited to be heading there and hopefully very, very soon.

:: 29:50

Just usually one of the recent models that's come out, right?

:: 29:53

Like remember when Deepsea came out, like, oh my gosh,

:: 29:55

this thing's revolutionary.

:: 29:56

But really all it did is it was optimization, right?

:: 29:59

Like nobody in the US who was building these things ever thought about, well, maybe we need 5 decimal places instead of seven to like save these things.

:: 30:07

Or how do we purge things at Octabit?

:: 30:09

But yeah, like how do we keep memory clean?

:: 30:12

How do we keep the things that we want to know?

:: 30:13

And not everything that someone says or even every little bit of this conversation will be useful to everybody, but selectively choose the things that they are that they want to keep and store.

:: 30:24

And keep that optimized.

:: 30:25

It's defragging, right?

:: 30:27

There you go.

:: 30:27

That's the right model.

:: 30:29

So how could people help you?

:: 30:31

Like where, what are you looking for?

:: 30:33

Are you looking for more people to use agents or how can the community support what you guys are doing?

:: 30:38

Are you looking for more startups?

:: 30:40

You know, are you looking to partner with more people?

:: 30:42

Are you looking for more ideas, for more agents to build for startups?

:: 30:45

How could we help you and kind of the Agent AI community?

:: 30:48

So I'd say firstly, big shout out to Agent AI's

:: 30:53

support team as well as the community.

:: 30:55

The reason that we've got as far as we have, I would say, is largely thanks to them.

:: 31:01

The first set of agents that we had, we stumbled our way to sort of building.

:: 31:06

And I would say it was the support from the community as well as the support team that got us to build our first set of successful AI agents.

:: 31:14

And we still are very active with the community.

:: 31:18

And so I think you'll have done an

:: 31:20

Unbelievable job there in terms of the help and support that we need.

:: 31:24

So I'm just going to go back to your earlier question as well, because I did share about the product vision we're excited working towards.

:: 31:33

But there's very much the sort of the community aspect that I think is essential to helping people with growth.

:: 31:41

I think the level of learning that comes through seeing a strong peer network build alongside you,

:: 31:48

learn with you and learning from founders that are one or two stages ahead of you that have just completed the seed fundraise or a series A fundraise that you're heading towards and to actually be able to hear about their journey, challenges they had and how they overcame them.

:: 32:05

That I think is very powerful when you get that from founders that are one or two stages ahead.

:: 32:12

And so the goal of the community is really learn from the Bandilo team, mentor network,

:: 32:18

as well as investors, but very importantly, learn from each other because that really becomes very powerful.

:: 32:25

And then you start to win together and the startup ecosystem starts to win more as a whole.

:: 32:31

So I think that is really, I would say, the most important aspect of what we're looking to go after here.

:: 32:38

And AI is an accelerating towards that, but it's really one piece of that.

:: 32:43

I would say the set of the human community piece is exceedingly, exceedingly important.

:: 32:48

So just want to stress in it.

:: 32:50

And our five-year goal really would be, can we support 100,000 founders and help them grow towards contributing towards the SDGs?

:: 32:59

We think if we can accelerate innovation towards the UN Sustainable Development Goals, then we would have achieved what we set out to go after.

:: 33:07

And in terms of support that we need from here on, I would say from the HubSpot Agent.AI team, what I'd be most excited about is more integration of the Agent.AI platform into HubSpot.

:: 33:20

So we spoke about the memory piece, but really what makes this unbelievably powerful is when your agents reside in the same place as your customer data, as your customer conversations, as your workflows.

:: 33:34

Because then it really, like, then your entire company operating system is in one place with one comprehensive platform that talks to each other completely.

:: 33:43

And so that is something that we would, that would be our sort of, that would be on our wish list.

:: 33:51

It's definitely coming.

:: 33:52

It's started and it's just getting easier and easier and easier because yes, engineers can do it now, but right now the ability for people, you know, that aren't engineers to be able to do that through these layers,

:: 34:03

of communication and APIs, and I love it.

:: 34:06

I think it all makes so much sense.

:: 34:08

Raj, what about not asking you?

:: 34:09

What do you want to share with people about kind of where you see things going, building for AI, any advice or lessons learned that, you know, scars from your back that you want to make sure people don't stumble through, or just advice in general that you want to ask people as we start to wrap it up here?

:: 34:25

Yes, so I'd say one of the things is with new technology, the easiest way to learn is by doing, I think.

:: 34:33

And so I would never have imagined having this conversation with you close to a year after I started to tinkle with my first set of AI tools, right?

:: 34:42

I would have never dreamt, like if you asked me towards the end of last year, are you going to build an AI platform?

:: 34:47

It would have just not been something that I would have expected.

:: 34:51

And so sometimes it's just the big things start with very, very small things, right?

:: 34:58

And so just giving yourself permission to

:: 35:02

try, experiment, fail, and learn, I think is the most important thing you can do to get your hands dirty and start building.

:: 35:11

Because what that does is it really then exposes you to where technology is today and what you can do with it.

:: 35:18

But it also really unlocks insights that could affect you at a company level and a strategic level as to where you want to take what you're building.

:: 35:28

And so I'd say really being open to

:: 35:32

So experimenting.

:: 35:34

The one other thing I'd say is that very quickly when I was tinkering with these tools, I found that, oh my God, there's like 7 amazing tools that I can use, right?

:: 35:44

And it's overwhelming.

:: 35:45

It's A lot.

:: 35:46

And every day there's a new AI tool that is incredible that you can use.

:: 35:50

And it can be highly intimidating as well as to where do I spend my time?

:: 35:54

What do I do?

:: 35:55

And I'd say the one thing I'm very happy about is that we took a little bit of time upfront to play with tools and then kind of realize, okay, Agent.AI works for non-builders.

:: 36:06

It's highly intuitive, it's powerful, and it can really leverage HubSpot long term, which is where most of our startups' data is.

:: 36:14

Most of them use HubSpot.

:: 36:15

We're HubSpot for startups customer as well as partner.

:: 36:19

And so the thought was there's this platform that actually makes sense to build on, and we don't need to be

:: 36:25

an expert across platforms, we need to just start building on one where we can get some value and make progress.

:: 36:34

It doesn't have to be perfect.

:: 36:35

We just have to add some value with some progress with the first set of things that we build for ourselves.

:: 36:41

And I think you said the most important word of all in there, to me, that I've heard a lot of builders and people in the community say, Play, start playing.

:: 36:50

Don't be afraid to play.

:: 36:52

Raj, what's the best way for people to connect with you and to find Bonbilo or any upcoming events, projects that you want to kind of shout out as we kind of close out here?

:: 37:02

So I would love for folks to sign up to Bonbilo's newsletter, which is on our website, Bonbilo.com.

:: 37:08

If you can both follow us on Bonbilo as well as myself on LinkedIn, that would be much appreciated.

:: 37:16

That's where we share sort of upcoming events, news, new AI agent showcases,

:: 37:22

helping founders use them, demoing them as well.

:: 37:25

Suraj, I really appreciate you taking the time to have this conversation with me.

:: 37:28

And I'm really excited to see the next wave of companies that you help and all the stuff that y'all are doing.

:: 37:33

And as you continue to pay it forward and to see these hundreds of thousands of entrepreneurs build on top of what you do and accelerate how they're doing it even faster.

:: 37:43

So thank you so much for joining me here.

:: 37:45

Kyle, thanks so much for having me.

:: 37:47

Enjoyed this very much.

:: 37:48

and look forward to chatting again soon.

:: 37:50

And to everybody out there, keep playing, keep getting in there.

:: 37:53

And if you haven't already and you like what you hear, we're on all your audio platforms as well as YouTube.

:: 37:59

So go out, give us a five-star review, thumbs up, subscribe, do all the great things.

:: 38:04

And until next time, keep playing, everybody.

:: 38:06

Take care.