The Forward-Deployed Divide

Julia Nimchinski:

Next up, please welcome Michele Buckley, AI GTM expert and former VP of Analyst at Gartner, and we have an all-star CXL panel. focus on the forward deployed divide. Welcome to the show, Michele. And I’m promoting everyone from the audience. One second… Everyone’s rejoining. Amazing. Welcome, Michele.

Michele Buckley:

I’m here, I’m lurking, I’m a bit dark. Sorry about that. But hello everyone, excited to be here, and we do have a star-studded panel. Thank you, Julia. I’m excited to meet everyone, and I’m going to introduce folks in, alphabetical order by their last name, but first I will introduce, Gal Aga, who is the CEO and co-founder of Aligned, based in Tel Aviv, doing some amazing things around digital sales, rooms. 2.0 AI-fueled type of, sales engines, which I’m excited to learn more about. I’m B as Michele Buckley. I am based in Boston, and I’m a former Gartner analyst that helps startup and grow CEOs build scalable sales teams.

Next, we have Tooba Durraze, who I’ve seen before. Good to see you again, Tooba. She is the founder and CEO of Amoeba AI, and she’s based in San Francisco, formerly an MIT cognitive research scientist and VP of Products at Qualified. And I should mention, actually, the Gal and I were both at Gartner for a while, so that’s interesting, because We also have some links between our other panelists. Jaclyn Rice Nelson, co-founder, I’m sorry, and CEO of Tribe AI. She’s based in New York, and formerly from Google. She’s described her company as talent first. and partner obsessed, which is so interesting for this topic, onboard deployed engineers.

Next, we have Brian Peterson, who was also formerly from Google, and is the co-founder and CTO of, Dialpad. And I don’t see him on right now, but no doubt he may.

Brian Peterson:

Am I here?

Michele Buckley:

Here? I see your picture. Oh, there you are. I’m sorry. That’s good.

Brian Peterson:

Good. Make sure it works.

Michele Buckley:

6 people at a time, I’m not sure. And then we also have Abhi Yadav, who is the co-founder and CEO of iCustomer. He is a three times founder that is focusing on the future of Agentic commerce. and audience data and intelligence requirements. So, Abhi is also a guest lecturer at MIT. Tooba was also a research scientist at MIT, Brian was at Google, Jackie was at Google, Gal was at Gartner, I was at Gartner, so… good times. We all come from different background. I think Brian’s the only one who’s actually coded, and is our resident Python.

Brian Peterson:

Okay. So…

Michele Buckley:

Let’s… let’s jump in. You know, today, business leaders aren’t really asking who has the best model. They’re asking who can combine this model with our context, our workflows, and our institutional knowledge. So. That’s the spirit of what for deployment is meant to do. I’d be interested in going around quickly to each panelist. And, we can go in the order that I introduced you for the first round. I’d be interested in hearing from you to get us grounded. What is a quick few sentences on what you see for deployed means in your world? And we’ll start with Gal, right?

I know you’re all muted, which is great, but hopefully we can… full session.

Gal Aga:

Absolutely. Yeah, so maybe for context, so Align basically is a workspace between the buyer and the seller that helps just drive better deal execution. It’s where you enable champions, build better buying groups, multi-thread, get insights about the deal when the buyer’s not in the room. So, really, and a big part of… of how we’ve built and how we’ve evolved the product over time. Of course, like, everyone is with AI, And really, at this point, it’s really become the space where you run the deal with AI.

So, for me, FDE really is about how do we… how do we adapt that AI and how we bring the context that the customer has into… into the learning, into the brain of our AI, And how do we, ensure that it learns over time and fully… and can get fully deployed within their organization to get all of the context on an ongoing basis, and to, over time, improve? So, when we look at it, we look at it as an AIFD, actually, which I know is a big part of the discussion today, and that’s something that we’re building, and I have a strong POV about.

I think that’s where the market is moving towards. But overall, that’s, that’s basically how I see it.

Abhi Yadav:

Oh, you’re on mute, Michele.

Michele Buckley:

karma, isn’t it? I told people not to go on mute. Okay, Tooba, you’re next up.

Tooba Durraze:

Yeah, so, for us, I guess forward deployed, Folks are called decision architects, and the way we look at it is, you know, their backgrounds are normally in some combination of data, organizational design, so understanding what is the system around a decision, and sometimes, the architecture of that involves things that are kind of not even related to, kind of, Amoeba as a product, right? It’s a lot of information around… education around AI, where AI is used, teaching folks how to take a strategy and break it out into work, like, how would you think about that? So we consider that a part of, like, raising the overall, kind of, knowledge and cognition of the team that we’re working with, as well as then architecting systems around, their decisions, and low to high fidelity.

And then within it, you can have folks that are kind of more data-oriented versus more strategy-oriented, but at the end of the day, it covers, like. the entirety of that loop. So, think, again, organizational design and psychology, mixed with data science, mixed with, kind of, like, a bit of, like, technical implementation and education trickled in there.

Michele Buckley:

Okay. Great. Next up, Jaclyn, or Jackie, if I may?

Jaclyn Rice Nelson:

Yeah, perfect. So we are similarly obsessed with the AI deployment gap, and so all we do is help large enterprises, so typically Fortune 1000 companies. go leverage AI, but do it to actually get value. So, either generally create net new revenue or cut costs. Those are sort of the two rough buckets of projects in order to make those things happen, we know they’re incredibly powerful models. The hardest part is then making it work in these organizations and picking the right problems for, us to point this technology at. And so that’s where we spend a lot of our time.

Pretty much everyone at Tribe is considered forward deployed, or forward deployable. And, I really think about AI engineering, so core software engineering, deep expertise with the tooling, that’s, that’s really the core, as well as AI product development. design, because ultimately, consumers, people, employees have to use these products, and they have to be usable, and sort of things that can be adopted in order to have the value. But we have a far more technical view on forward deployed, generally, than I think the market does today. And that only works if you do still connect to business outcomes, business goals. and business and financial metrics.

And so we do kind of, like, bridge these two worlds, but we do it with the technologists themselves.

Michele Buckley:

Okay, thank you for that. We’ve got Brian up next, and then Abby.

Brian Peterson:

Yeah, so the… the thing with Ford… the Ford deploys and engineers are really new, as I think everyone realizes now, but the reason they exist is because when we’re talking about AI, we’re now talking about Agentic AI, and it’s not just using models anymore, it’s taking action with those models, and it’s doing super, super complex workflows. that is completely different for every, every company. And so, what’s happened is, this role’s come out of nowhere, because once you start to go Agentic with these companies, every company has a completely different workflow.

They have different systems they integrate with, they have different types of data they have in different formats. they have their own, just completely way they want to handle customers, and we do, you know, AI platform for customer experience, so we’re dealing on, like, the customer interaction with the business, but that can be anything. It can be, you know, I needed to do this, then this, then this, then that, and it needs to work reliably, it needs to work the exact same way the business wants it, and you can’t just throw these AI models at this stuff, because it’s not gonna guess it right.

Because it’s built on data that is the whole world wide web of data and randomness. Whereas the new thing that’s happening is. these AI models need to be trained on exactly what the business needs, so it’s getting verticalized, but it’s even going into the exact what this business needs. And the only way to do that is to get a very technical person right now embedded in what they need to solve, and working through on, like, okay, we need this, then we need this, then we need to integrate with this, and then this needs to work. and make sure it’s completely tailored towards the business.

That’s gonna change over time, it’s gonna evolve a little bit, but where we’re at right now is they’re super hands-on for these forward-deployed engineers, because you have to be. You have to customize things for that business. So it’s sort of a completely new way of doing things, because before, SaaS was just press a button, connect your login, and you’re pretty much good to go. It’s a website, use it, you’re good. But now it’s, like, super integrated workflows that need to work.

Abhi Yadav:

Awesome. I can go next. So, from my perspective, you know, just being at the intersection. across multiple companies, where earlier, there was a component in SaaS we used to call professional services, which is usually, like, okay, you gotta implement this product, and, you know, and then I… I coincidentally had worked personally as a decision scientist, and I had built out, like, COEs for Fortune 100 companies, you know, Decision Science COEs, and… data science COEs, so we’ve seen this movie a little bit earlier as well, but that time, the goal was, like, COEs was, like, more for, like, human agent and not AI agent, and then the goal was to kind of Captivate this captive intelligence, you know, engines, and still, like, human travel knowledge, all that, and, you know, they used to support, core business outcome guys.

Unlike now, what we are seeing in this company, iCustomer, and we evolved from you know, decision intelligence and brought to more, like, audience and decision intelligence. We’re, like, honing in on the most important aspect, which is what is the context on your customer, and then what is the context on your brand. And then, basically, the forward-deployed engineers, and we sometimes call them forward-deployed operators now, all of us are in our company. Our goal is to ensure that we work with the clients. to make sure that the harness which we have built for them on their customer data, audience data, kind of takes over.

So it’s almost like four deployed operators needs to get fired, you know, fire themselves. Within a matter of a few weeks, months, when the Harness, like, really takes over. And that’s a journey sometimes, because most of the companies are not ready with the kind of data governance, and they’re, like, still talking about you know, the symbolics, and kind of semantics, and things like that, so it’s largely focused on the kind of data and decision science kind of an area, but, like, we appreciate, like, business know-how, and that’s why we call them operator. All of our earnest works toward not for some fancy report, or workflow, or things like that, it’s very outcome-focused.

That’s what we have sort of honed in. Our… Power customers are some of the hyperscalers, and they kind of laser focus on, you know, here’s my… account-based marketing approach, and here’s my goal for this, here’s my net retention rate as a KPI, like, how do we… what is that intelligence, or what do we do, really, to get that moving, and then demonstrate that in an actual outcome than a fancy data, or report, or a workflow, or a dashboard, or something, you know, so…

Michele Buckley:

Oh, well, thanks for that, Abhi. I… I really… find it interesting what you all have said, and I have to say, in general, I probably agree with you, but… There is a new dynamic here. that I think we can’t dig into in a little more detail, because, like you said, Abhi. you know, I was… my first project was in the 90s, where Oracle was implementing a new system, and so they had a consultant on-site, permanent, you know, for, like, years, 2 or 3 years, to actually get the new Oracle system up and running. So. people from the outside, I think, when they look at first glance, they might hear things like, well, they’ve got to train… train the model.

Well. you know, that doesn’t sound like anything special. You have, you know, or we’re gonna help you with your workflow, that sounds like consulting, or… you know, training model, isn’t that just, like, data entry and the like? And… I think it’s… and we mentioned AI engineering, but I am curious, genuinely, I mean, I have some ideas, but… we obviously know that Palantir pioneer this kind of model with very complex software, but doesn’t everyone have complex software? I mean, what is the real difference here that we didn’t have before, that a consulting company can’t do, or that, you know, an internal software engineer can’t do.

I’m interested, if anyone wants to jump in.

Abhi Yadav:

I think I… I’ll just say one quick thing, and we… this is how I talk about… my background, my previous startup exit and all that was a CDP company, and we were like, okay, you know, when CDP came in, and, like, half of Martech is CDP, eventually later then. The whole goal was, the tool is not the solution, the data is the solution. So if you centralize the data platform, centralize the customer data platform, then you can do a lot. Now the expectation, and this is what we’re also evangelizing, you put… but there’s no brain on the database.

You just stored a centralized database system. And in order to build a brain, you need to build the ontology, the semantics, the neural nets, and, you know, the reasoning, and all of that shouldn’t leave the customer from ice. You know, it shouldn’t… be fed into Frontier models, so our leads are leaking in into everyone else. Everyone knows, like, what’s going on. So I think it’s a little bit different, much different that way. We’re trying to build a brain. inside the company, and forward-deployed operators, engineers are kind of enabling that, and basically they’ll be gone after that.

That tribal knowledge leaves with the system and the people who’s gonna operate, run, get the outcome, is gonna leverage it. So, you’re technically giving them a piece of software which works for them, instead of, here’s the software, go figure it yourself.

Michele Buckley:

I think I’m… maybe I’m… I’m just really old school, because I think, like, well, that’s what solution providers are due, right? You’ve got these big, old-school software providers, like… like Microsoft, and like, they’ve got thousands and thousands of partners that do that type of thing. And again, I’m not disagreeing with you, but I’m wondering, like. There is something new here that… Is it… And is it just as simple as, well, this is a whole new wave of technology, and the market isn’t educated on it yet? Or is it that there’s not enough talent out there inside of these corporate customers, or is it… go ahead.

Tooba Durraze:

Yeah, I don’t think it’s as new as you think it is. I think the terminology is new. I think the skill set of, like, some of this technology, and implementing it to the point where, like, if you have a proficiency in things like ontology, etc, it’s, like, a lot faster for you. it’s… you’re incentivized as a vendor to make sure people are getting value out of your platform, right? That was always the case. So if it helps people get value by having folks who have an expertise in this area kind of embed within those organizations.

And, like, mind you, like, when you’re going into an organization, like, they have their day jobs, they still have to keep the business running, right? And a lot of what happened with the AI was, like. It just, like, added hours to people’s, like, lives, essentially, right now, as you’re tinkering. So this idea of, like, experts coming in and helping implement, we feel, to Abhi’s point, quite strongly about the IP then staying with the company. We’re teaching a behavior, they’re going to have to take the system and kind of maintain it. if it allows… if that kind of model allows them to kind of leapfrog in terms of value, then I feel like, you know, why wouldn’t they do that?

But the concept as a whole. it was… when they were split up into the ecosystem, like, your technology partners, service partners, etc, that concept as a whole has existed. This is just, like, a new kind of, like, umbrella term for it.

Jaclyn Rice Nelson:

I tend to agree. I think that in some ways, this time is different, and in others, we are making something sexy that was not ever sexy, frankly, right? And so, we get to now all of a sudden wake up to the reality that technology does not just work out of the box. That is shocking, shocking to everyone. Apparently. And, and it’s true. It doesn’t. This is a new technology. It is, you know, essentially still unproven how enterprises are going to unlock massive value, are actually going to be able to transform with this technology. And the existing players in the services ecosystem. are not native to this technology.

And so they are learning it, and they need to transform just as much as their customers. And so you have… to me, Michele, that is what is different, is that it is a new technology wave, and you need the experts in that technology. We only know a services ecosystem And they are not yet the experts in that technology, and so you have to actually create something net new to service this. market. The market is all of the sudden the entire economy. And these companies are growing at a rate that is unprecedented, so the need, the pull from customers is so substantial, but the value has not been realized, in part because the technology does not work out of the box, and in part because the services ecosystem was too… was frankly not prepared and didn’t have the expertise.

And so that is where the SCOR deployed model comes in. And so, to me, it is a rebranding, for sure, but it is really saying that this is a AI-native model. This is a model. with technology at the core, and where you are layering technology or software plus services, and that is the model that I very much believe in. It’s the model that Tribe AI is built on, and I think it’s the right answer. What that actually means in practice, what FDE means, has become so diluted that it has essentially become meaningless already, and I think that’s why all of these things have gotten So, so muddled.

Brian Peterson:

I’m gonna… so, I think the thing with… with a lot of this is. To simplify it, 4 deployed engineers only exist because of Agentic AI. I wouldn’t even say that AI is what kicked it off with Palantir and all them, it’s the Agentic part of it. if… when models came out, and everyone’s using ChatGPT and whatever, or you’re using Google Gemini, or whatever it is, you didn’t need a Ford deployed engineer. Like, you just… you just started using it and asking questions. The second it became something that’s going to automate something in your business. That AI, that AI model does not transform to automation.

Like, it doesn’t, because it hallucinates, it’s not deterministic, it can’t do this, then this, then this, then this, all the time, accurately, which is obviously super important for a business. It can’t just, like, magically change its answer one time from the next. What did happen is these AI models enabled Agentic to be a thing, to actually be something that can work now, but then everyone’s rushing towards Agentic, and they started building these really complicated, you know, the hot term is harness now, like AI harness, which just basically means an Agentic thing to make the AI do the thing it’s supposed to do, and not… and have guardrails, and make sure it doesn’t, you know, say something that you shouldn’t say as a business.

But that’s still really custom right now, and really, really hard to understand and deploy to what everyone else is saying, too. It’s super customizable, which, like, AI is the most customizable software in the history of time, which means it’s like… and it’s also the most complicated, and so when you get into a business. they all, to make this magical, Agentic workflows happen and do great stuff, it has to be so bespoke. It has to be… and you can’t just have, like, a services person do it, because it’s not a typical API integration anymore. The, like, you’re kind of like… You know, you’re designing this system that the product person… so, like, us, us companies, we know how to design the system better than anyone.

So we would never want to hand it over to a services, you know, or consultant, because there’s no way they would possibly understand the best way to take advantage of our system. That will change over time. Like, as these things get more mature and are more self-learning, yes, they’ll start to be able to be outsourced more, but, like, right now, this is just… We kind of… everyone had to hack the system, and if they’re saying, oh, no, this is… They’re all hacking the system. They put out four deployment engineers because this thing is so complicated that there’s no way they could possibly deploy to an enterprise without having some person, you know, tweaking every little thing to make it work. this is purely an Agentic thing.

This is not, like, we made something up. It’s really, like, you only need it the second Agentic came out. So, but yeah, it is complicated. It’ll change over time, it’ll get easier, there’ll be more automation. for different companies, you’ll still need some form of deployment for a deployed engineer, but it’ll definitely evolve over time, and it’ll get easier for both sides. But this is all because we’re moving so fast in a space that, like, doesn’t fully work yet.

Michele Buckley:

Yeah, that… I… that’s making sense to me, Brian, and what you said, too, Jackie, obviously, like, yeah, it is the new buzzword, it is the new black, and, you know, the number of FDE jobs that I’ve posted, like, doubled, if not tripled, year on year, but… I’m hearing in your comments, Brian, that the it’s right, it’s not a typical API, it’s not automating something that you already had, like RPA, it’s an opportunity to transform a workflow as well at the same time, and you don’t want to automate things that aren’t working, so it’s… Perhaps that transformative nature, like you said, of transforming a workflow from, the… transforming a workflow that was there before, but you’re redesigning it, and you’re not going to deploy software unless it’s designed appropriately.

You need experts on your product to understand, like, how it was designed. Right? Like, you might say, well, you don’t need that part anymore, because we designed it this way to skip that traditional approach, and I’d… I’d be interested to hear from you, Gal on what If you’re finding that as well, because you’re in the area of sales where.

Gal Aga:

Yep.

Michele Buckley:

Some might say there’s been traditionally a lot of process and procedure around sales, and Are you transforming that workflow? Or, you know, what… do you have FTEs doing these type of things?

Gal Aga:

Yeah, so actually, so our, our product, maybe I’m the only one in this conversation who does not have FDs, so our, our approach, is, the way that the product is built and the way that the Agentic processes work, do not require an FD. And we’re… we’re trying to really skip that part and move Straight into a, Building the brain internally and building it in a way that does not need that layer in the middle. But of course, we’re not naive. We know that it’s not gonna come in a day. Basically, you know, an FD does extract a lot of the processes that people have undocumented in their head.

And there are, you know, there are successful workflows that should turn into a process, and and and the AI can detect automatically. You know, the AI doesn’t know how to understand the decision rights, or the political influence, or, you know, the hidden stakeholders that maybe are part of the organization in certain processes. And definitely, you know, maybe kind of exceptions, things that are… that happen at the organization that, only few people know about. Ayy. So… Yeah, it really depends on the use case, and I definitely see the need, and it really depends on what you sell as a company.

And for our specific use case, we’re trying to build, the learnings, On the end-user basis, on the company basis, and then cross-company and try to apply and build the brain that really learns over time all of these processes that are undocumented, learn from the things that are documented. And then build that knowledge in-house so that the AI can operate even over time. In a better way than, Than what the company knows, really, about how it should operate.

Michele Buckley:

Hmm. So you… you mentioned building the brain, and Brian mentioned that too, and now you mentioned over-deployed operators, although I’m not… Sure, anyone wants to be over-deployed, but, I mean, what specifically… what specifically skills do you see in your travels? And I’ll invite anyone to comment. like… What is the actual… what is required to build this brain. Beyond just, obviously, like. coding and… I mean, learn… is this a human that’s actually building the brain? Is this a human creating agents that are building the brain? Is this humans creating a workflow with agents and humans in the company building the brain?

I’m just curious of, like. what you see as… how… how… what form it actually takes, and then I’m also interested in, like, how… you’re selling this, because at the end of the day, you’re talking to a business leader who’s like, I want this outcome. Are you selling to the outcome, or are you saying, you know. don’t worry about it, we have a solution scope, or you give us the outcome, and we’ll charge you once we deliver it and the like, so I’m interested. In learning more.

Brian Peterson:

The, the, yeah, a lot of this, sorry, I’m sorry, yeah.

Gal Aga:

No, go ahead, yeah.

Brian Peterson:

This is… the most self-service you’ve ever needed to be in technology history, pretty much, on… like, the lack of self-service that you’ve ever been able to have. It’s not a typical, here’s a website, here’s some settings. These are things that are… the craziest, complicated systems you possibly can have, because we’re finally getting to the point where we can automate really complicated workflows. It’s not just a simple thing, it’s like anything you can imagine in your business that is digital, or even connects to a human, eventually, can be now automated with this advancement. But it does come with needing to understand every single layer of that business, and every integration, and the data impacts it too, because the data helps train the agents and point them in the right direction, know which answers to give.

The only way you can make this all work is if you are so embedded, and in our example. You know, when we go to talk to a company. they have, like, 30 different tools, on average, probably, that they have to integrate with to automate stuff. And a lot of those tools aren’t necessarily even normal, popular tools. They’re maybe a private… CRM? Well, they still need to be able to have something that, like, looks up the customer, makes a decision based on their own private CRM data, then… then takes an action in their own system through a private, you know, endpoint.

These are things that they could never do or know to do on their own. That’s getting better over time. A lot of this stuff is going to start being self-learning over time, where we’re using, and in our example, we use human conversations to train the AI, and AI conversations to train the humans, even. With live coaching, that’s where it’s eventually gonna go, and it’s gonna get smarter over time, but just to kick it off. You really need to have someone who understands exactly what they need to automate. And there’s no way you can guess that.

It’s not just like, oh, I want to automate sales. It’s like, no, no, no, like, how do you specifically need to automate it based on all the systems you use? And so that’s where this whole Ford deployed thing came about, is there’s no way you could have just released this and had the companies magically figure it out. They had to have someone as an intermediary who, like, will connect those dots. With them. So that’s how we ended up in this space, of, like, making that, I guess, that brain, is that it’s so custom. It’s the most custom… Software in the history of time.

For good and bad. Yeah.

Abhi Yadav:

And the output is actually, unlike that, is actually the working software which is useful, then a document, or… You know, whatever, or, like, a chatbot, like, how people are thinking, But, like, I just want to add to what Brian just said, and I agree with what Jaclyn said. There are… There are some uncomfortable things about the hard thing, and we’re trying to make the hard things more sexy. in the process, but going back to your question, I think from our perspective, and again, to what Gal said, like, the use case varies a lot, right? All these conversation varies by use cases.

So, in our situation, we’re working towards outcomes. We’re not enabling a workflow. In fact, we’re… most of our power users are actually AI-native tool users. Who have been… like at this cusp for the last one year, where they’re doing a whole bunch of data enrichment, a whole bunch of, you know, data automation, and sending one place to the other, and then there’s just execution-oriented. You know, they have just built workflows all this while. But that has not got them any outcome. And actually, that has intimidated their customers, you know, just by the overblown proportion of activities.

So, we’re taking a very, different route. We’re like, okay, if this is the KPI, the system, or this, we call it, it’s a recent launch we did, we call it Growth Brain. So, if the Growth Brain needs to optimize towards your audience experience, your customer experience, but also, like, what are you offering to your customer, and at what offer, and who, and when, and where. And it’s nothing to do with campaigns, or more creatives, or more content, any of those things. You know, we’re kind… so most of our skill set is more in understanding how does a certain KPIs work.

Like, most of our team members had spent, like, a lot of time fixing Fortune 100,000, like, very advanced companies, like, how to optimize CAIC to LTV ratio. You know, and then, okay, to do that, we need this sort of data, and this kind of thing. Oh, you may not be collecting that kind of data in the format agents want it. So, I’m just kind of dumbing that down, like, that’s how this whole, like… Go, like, reverse engineering works for our kids. So, basically, the skill set requirement is kind of anyone who understands the domain and the business and the KPI is kind of table-stake work within.

That’s why I kind of said forward deployed. operators, because engineers can actually code and create anything you want, and like Brian said, in the human history, this is the most flexible time, where you tell us today, and like, you can wipe code anything. You don’t have to be a coder, you can wipe code anything. That’s not the point. The point is, will this work? Will this get the outcome? You know, we’re looking for, and that’s really hard, like, okay, define outcome. Let’s… let’s start the conversation right there.

Michele Buckley:

It is… it’s really hard, and especially, you know, having been at Gartner, we’re heading into the trough of disillusionment around AI and AI agents, where everyone’s starting to realize, wow, okay, wait, I could, you know, vibe code a few things, but now it’s really hard, it needs to be… has governance, the data and the like, it needs to be… A lot more complicated than I thought. At the same time, now we’re finding that technology purchases are being dominated by business-led stakeholders. They’re not just buying software. because they want better data quality, right? They’re buying it because, like you said, they want a business outcome, and they want to get their sales faster, increase their average sales price, or increase the customer engagement, or resolve issues more quickly, and the like.

So, it’s… It sounds like we’re coming to a point where we’re saying, well, look, this forward-deployed engineer needs to be able to actually code, but also needs to understand the business outcomes needed, and I heard a former FDE from Palantir say that it’s almost like the FDE has to be a consultant and a product manager and an engineer combined, because you’ve… once you get the client talking, there’s going to be a lot of issues that they have, and a lot of outcomes that they want from your software and the like, so… I’m… Connecting.

Tooba Durraze:

Can I offer an opinion here? I think it’s, like, to Brian’s earlier point about, kind of, buzzwords and, like, harnesses being one of them, I think FDE as a term is defined differently in different areas and in different industries as well. There are versions of FDEs that are, like, very, very, like, technical, which is… which is where the gap was and where this started. So, Palantir’s a great example of that, because ontologies, Governance, creation of ontologies is a very, very hard thing, and what’s on top basically will fail in terms of agents and orchestration if you don’t really have the right understanding.

So, like, it’s like, the brain… to the brain piece, we thought if you can aggregate all the data and then put agents on top of it, they should be able to do XYZ. Okay, no, but they need more context. Like, the brain is basically a contract, I would say, between the person who understands inside the actual organization, the organization really well, and maybe paired with a subject matter expert on, like, business process, as an example, and a very technical FDE on if this is a problem statement, what does the construct need to look like at the end of the day?

So it’s like, it’s, I think we’ve, like, for lack of a better term, like, kind of bastardized, like, all these different terms, essentially. I, I have… Because FTEs, like, in my head also, like, jumped to… the need was, like, technically, this is very hard to understand and do, why don’t we leapfrog by embedding someone in it? And we, like, added kind of this consultancy on top of it, so then it became, like, kind of this umbrella term. I think for the audience’s purposes, like, it’s good to be aware that ask your vendors to describe what they mean by FDE, if they’re bringing one on, and, like, what exactly they would offer as well, because it varies.

It’s not the same, by any means. And the success of the project also varies because of, then, what is included in whatever that FDE model is, as well.

Michele Buckley:

I might direct your, sort of, question or comment to Jackie, because your business has been… Around for a while, almost 10 years, and… You’ve obviously seen a lot of dynamics with customer demand over that time. You know, what do you see that… They… they are asking for… Today, and how has that changed? Are they asking for people? Are they asking for projects? Are they asking for outcomes? Are they… You know, what would be your response to some of Tooba’s comments?

Jaclyn Rice Nelson:

Yeah, I mean, it depends who’s asking, or doing the asking. If you’re asking procurement, they’re still, you know, they’re still talking about inputs. And, you know, that’s when we have to explain, well, of course, that doesn’t make sense anymore, because the inputs, then, are the agents, not just the people. And so, you know, and you want that operating on your business. You want that extreme intelligence. If you only wanted the people, you’ll get a far worse result. And so I do think there’s an element here where, we’re still in an education moment, and so there’s sort of, like, what people are asking for and then what they want.

What they actually want are the outcomes. What they are asking for, you know, may not tie to that, because they still don’t really know how to think about it. And to me, that is where I think the market, FDE models, services providers need to be getting more prescriptive around how to get to outcomes, and what is required to do that, and also what the monetization models look like for that. Because I do think that if you ask people what they want, what they need, what their problem is, they do not know. And I think that, like, that is why we are all spinning our wheels.

And so, like, again, we need the model providers, we need the technology providers, we need the services providers to come in with an opinion. on what works, where the value can be accrued, and then help these companies to actually get that value. And I think, Michele, this goes to your question on, you know, the frothiness in the market, and that trough of disillusionment. Like, I see my job as preventing the trough of disillusionment. I know what is possible from this technology. I know that this is not a bubble. I know that this outsized value creation and transformation as possible, and my job is to go make it real.

And so, what are the things we need, to, like, remove the bottlenecks? To, like, create the space to go make those outcomes happen, to show what’s possible, because the market needs these stories of value to better understand what their problems are and what they should be asking for. Because we’re just still so early in this cycle.

Michele Buckley:

Hmm.

Brian Peterson:

That’s the problem with AI, is that the joke is, we have is, like, you have your AI bot, and it says, what would you like to do today? And you’re like, I don’t know, like. how about you tell me? Like, aren’t you supposed to know? And the thing, we run a.

Jaclyn Rice Nelson:

Aren’t you intelligent? Aren’t you super intelligent?

Brian Peterson:

yeah, like, why do I… why do I… why aren’t you more proactive? And that’s where the FDEs come in, that’s where a lot of this new technology, like our businesses come in, is that we’re going to customers, and the first problem they have, and we’re obviously in customer experience, so a lot of, like, sales and support conversations. they have no clue where to start on automation. No clue. Like, if you ask them, what are your conversations about on your call center, or whatever, on your text line every month, they’re like, well, we roughly know rough, like, sort of topics, but that’s it.

So, like, how do you know what you can even automate, what you should automate? Because it’s now going into, again, this is all Agentic. FDE is just Agentic. That’s all it is. It’s because Agentic is super complicated, but also super powerful. you have to go in and, like, analyze all this stuff and be that product manager, be the engineer who can connect all the systems. It is what everyone describes, it’s kind of an all-in-one, make this deployment successful, and you can’t make it successful without the FD anymore, because it knows how to build against your Agentic harness. perfectly for that business.

And, like, one of the examples, one of our favorite things we did is we do something called skills mining, where it analyzes all your human conversations, and then tells you, oh, here are all the repetitive things you do. So, like, we can… let’s automate that. You probably shouldn’t have a human doing a password reset phone conversation a thousand times a day. So let’s… let’s automate that for you. And they’re like, oh my gosh, now I know. This is where our software comes in, and, like, everyone here, and the four deployed engineers. They basically unlock what are those workflows in your business. and they customize it down to your business, because now it’s a human interacting with these agents in a lot of places.

That’s your brand, that’s your personality. It’s not just like a web form anymore. So that has to be so, again, custom to what you want as a business, because this is now going to be your front door to, like, your… the first impression that your customers have with your business, it should sound like your business. It should behave like your business. It’s just more complicated than anything in the history of time now. So it’s just… it’s a… it’s wild right now. I would…

Michele Buckley:

Oh, this is… this is why we’re here, and that wasn’t a bad rant at all. You should see mine, okay? So, we have a couple of questions for the audience, and, they’re, like. five-part question, so I’ll just try and condense them into… into one. The, and one of them is directed towards you, Brian, about… You know, you mentioned earlier that you walk in, they have 30 different tools. Right? You’re… tool… are you tool number 31? Like, how do you assess to… What custom systems you’re going to… work with and what’s not, or is this much more targeted of just saying, I’m just going to focus on the business problem, I can’t handle your, like, legacy debt of architecture that’s incredibly complex?

Brian Peterson:

They just, they just start with their, like everyone’s saying on this call, you start with the business problem. You work your way backwards, rather than just saying, I want to apply. Agentic, or, like, AI agents. You start with a small problem, even.

Michele Buckley:

Right.

Brian Peterson:

something that you know is repetitive, and you do approve of concept, but that’s just in general, if you’re doing any kind of AI transformation at your company. But it… that’s where we come in, and where all this… these new Agentic, you know, players come in, is we come in and help you. That’s why we have to have a forward-deployed engineer, or product manager, or operator, whatever you call it, because we need to understand how your business works, so we can tell you what you can automate. And the best thing about this Agentic world now, you can literally automate anything that a computer can do.

But it’s not, like, a plug and play. That’s the biggest difference.

Michele Buckley:

No, sorry.

Brian Peterson:

Everything we’re talking about now, it used to be a plug-and-play. It’s like, just connect these two APIs, you’re good to go, here’s your CRM, here’s your, whatever, sales funnel software. But now it’s like, no, to have this magic happen. It has to be able to do all these connections to these different, like, 30-plus systems. You don’t know that as… and the biggest thing we see is people don’t know this. It’s really hard for them to understand what they can automate. They know they need to get involved with AI, but they don’t know where to start.

Because 90% of companies are not tech companies. They don’t have, you know, engineers and all these people who can spend time building custom software. That’s where we come in. It’s this magical platform that can just kind of do all this for you, but we need someone to guide it based on what you need as a business.

Michele Buckley:

And you’re finding that kind of demand in, like, Fortune 500 companies?

Brian Peterson:

Every single company in the world.

Michele Buckley:

Yeah.

Brian Peterson:

The bigger ones are obviously more complicated, but you can do it down to a two… because they all have really complicated things they do. That now, though, that’s the best part about all this, now they actually can be automated, and you can have better ROI and better customer experience, because the AI is going to be faster and better, potentially, than the humans will be. They’re not gonna replace all the humans, but they’ll be… they’ll be able to make your business better, and faster, and optimized.

Michele Buckley:

Thank you for that, Brian. So, I see we are having a great conversation and chugging along with time, so I… I like to… to use the Remaining minutes we have to… Take a forward look. So, I think we’ve… all got a lot of experience. We see it’s the birth of something fantastic, incredible power we haven’t had before. that we’re still not even sure what to ask. Or, or where to have the most impact. But I’d be interested in hearing from each of you. And again, we can, Go into alphabetical order again. That means, Gal’s up first.

What… Do you see… This area looking like? two years from now. So this is fun. So this is… this is not a ramp, this is now we get into the prediction world of looking ahead. And we’ve covered off how things are right today, but I’m just curious, in 2 years from now. where do you see this whole discipline ending up? Is it still around? Is it morphing? Is it changing? Are clients becoming more educated or not? Is it becoming more complicated or not? Is there gonna be a massive… platform that will suddenly automate all of this, and we don’t need FDD to do it at all.

Let’s start with you, Gal.

Gal Aga:

Yeah, I think… I think the… the two-year range is, is actually harder to predict than, than the longer. I think it’s…

Michele Buckley:

You can go 3 years.

Gal Aga:

I don’t know what that, you know, what that longer is, but I think at some point. Yeah, it will, I think… I think at some point, there will be an AI fully deployed engineer and forward deployed engineer, and I think that… The gaps are basically… You know, it theoretically could, and I’m pretty sure we’ll get there, I just don’t know to say what the time is, and in two years, like, we’re gonna see, I think, breakthroughs On the way to get there, but basically, you know, the gaps are, one, you know, it’s context. it’s still, like, con… like, really a lot of the context is just not there, it’s not documented.

A lot of it just sits in a lot of different tools. It sits in people’s heads, it sits in a lot of different areas. But really, over time, I think that gap will disappear as… you know, more and more things will get connected, and more and more processes will just enforce that context, and the AI will be able to just, learn over time from what people do, and then, on an average, it will just be better than the average employee, so… Like, I’ll take our world, so, you know, if not all… for example, in sales, like, if not, I don’t have all of the contacts from WhatsApps or texts, right, that people are sending, or just a sudden call that someone makes to someone else, a buyer to a seller, vice versa.

But… and that’s critical, maybe, for a specific deal forecast. But then, if on average, the AI just knows so much about how buyers behave in that world, and how sellers behave in that environment, in that company, how each seller behaves, and it even looks cross-company. Then the accuracy of it, for example, for forecasting, the accuracy for it to be able to forecast. Would overall just be greater, so we will just accept That, even if we have some context windows, some context issues, Then, it’s just better. So I think that’s one thing, that’s context. I think that the second thing is just judgment, just a lot… a lot of it is just human judgment.

And… I think that a lot of, like, humans are just still better in a lot of situations, just reading the situations and making decisions about the workflow. And we all see it, like, it’s… it’s all of the small nuances, it’s the creativity, it’s, you know, AI just still misses. I think it’s… and I think this is really just an LLM gap. I think that, you know, we’ve all seen how it can do magic, but then… a lot of times, you just, what the hell, and you correct it, and then, oh yeah, I agree with you, and then… and then it does the magic.

So, you know, it just doesn’t have that level of judgment yet, but I think that’s just, You know, that’s just gonna be breached over time with the technology. I think the last piece is really human trust. And it’s just a big leap, right, to let something like that happen, autonomously. But there’s no way around it, because theoretically, it can happen, and the more the models get better, and the more we kind of fix the previous issues, I think it will happen. And when you think… I’ll give you an example from our business, like Claude Code… At some point, like at the beginning, it was asking you to approve the workflow before Right?

You ask it to write something, and then it kind of… it iterated it back to you, it shared the plan, right, that it’s going to execute, and then you approve the plan, and then it went to execute. Right now, you know, it jumps over that. Right, we got… we got to a point where we trust it. And, I think it’s gonna happen in the same way, it’s just gonna overcome judgment and context, and it’s gonna be so integrated, and then people will just trust the system more.

Michele Buckley:

Hmm. Tooba, what are your thoughts?

Tooba Durraze:

I’m gonna go quick. I have some thoughts, some predictions, and some hopes. Hope, people stop using the term AI, or just generalizing a lot of these terms, or to mean just large language models. There’s a lot more in terms of Agentic AI that can include other things. Two, I think for all of the vendors who are operating with the forward-deployed model, who will realize when the forward-deployed engineers start hitting the bottom line, I would see a lot more advancements in the next two years, and those vendors then, knowing the space really well, automating what FTEs will look like, so just, like.

Agentic FDEs, if you want to call it that. the re… The disillusionment that will hit when FDEs step out, and a whole year has gone by, and the maintenance piece emerges will give rise to probably, to Jaclyn’s point, another term that will come about at that point, essentially, is what we’ll see. But I think overall, my prediction is, like. it’ll be… it’s slightly more partnership-esque if the vendor-client relationship, and that will continue to build, where it is more of, like, a, like, I’m buying a piece of software, and then I’m just gonna go use it.

I think there’s a lot more information exchange on both sides that’s helpful to both parties, so we’ll see.

Jaclyn Rice Nelson:

I’ll, I’ll add, I think we will move away from talking about incremental gains, efficiency, cost cutting, I think we will talk about growth. I think we will have real examples of companies who played to win, took bigger swings, actually were able to really change, and, it paid off in meaningful ways. And I think that, we are still, of course, early, but I do think there’s a lot of conversation, and still the approach is so, even if it feels like a big investment, so incremental.

And I think that’s gonna change massively over the next two years, and then the divide between companies is going to widen, because a lot of the value that we are talking about creating, the brains, the knowledge bases, the context graphs, like, all of these things compound in value if you are able to establish them early and build the capabilities to build Agentic systems on top of them. And so, I just think that not only are we in an exponential moment of technology innovation, but that is then going to translate into a dramatic widening of the haves and have-nots in this world, and that hasn’t happened yet.

Michele Buckley:

The changing of the guard. Yeah, go ahead, Brian.

Brian Peterson:

I have a hot take, I don’t think we’re ever gonna automate. Ever… everything. We’re not gonna automate everything. And I think a lot of people jump to the, like, doomsday scenario of everything’s automated and… you’re just gonna magically take over. There’s nothing that we can tell from our team of experts and the technology out there that there’s anything that’s gonna automate everything. I think a lot of people think, oh, we gotta automate all this stuff. It’s like… and oh, like, where do we get started? It’s like, don’t… Don’t try to automate your whole business, you don’t need to.

There’s only 1% of, probably, automation going on in the world right now. We don’t need to go from 1% to 100% overnight. It won’t happen overnight. But it should be… it should be… I’d say the best we can possibly get in this world, and we specifically talk, obviously, in customer experience world, is, like, 80% AI, 20% humans. But the 20% human is gonna be super important conversations. That have just as much impact as the 80% of automation, but a lot of people think it’s gonna go to 100, or some magical thing, and that’s just not true, and if they’re saying that, they’re probably lying, so don’t trust them, or they don’t know any better.

Just honestly, because we were in this space for a very long time. And then the second part of that is where it’s going. Everyone threw out Agentic, but Agentic requires a lot of maintenance. then they’re like, great, but it’s magical that it works once you set it up, but it’s also very sensitive, and very, flaky and all that, even once you get it running. It’s still great, and it’s awesome, and everyone should use it, but the next wave, in the next two to three years, is gonna be what fixes it. Which is reinforcement learning. So, all these Agentic systems are great, but they don’t self-improve.

So what you’re gonna see in the next, like, 2 to 3 years, and a lot of, obviously, I think this group is working on that, I know Dialpad is working on this, already, is it’s going to start listening to your business once it’s deployed, and start automatically adjusting things. So if all of a sudden your business changes, or your workflows change, it’s gonna detect that, it’s gonna know customer likes to hear this now better than this last thing, it’s gonna instruct the AI agent to say it this way now, without you doing a single thing.

And then I tell companies. your data is your moat. So, the data becomes what the actual value is at this point. It’s not… because everything’s going to become kind of a commodity, it’s going to be the data is what you have that makes you unique, and it can the AI automatically learn off that data.

Abhi Yadav:

Yeah. No, in the interest of time, I agree with a lot of folks, what you said, but I think from my side, I’m optimistic. I feel that they’ll be really intelligence, which most of the chaos, if you look at it in the companies today, is not actually tool-driven. It’s actually operations-driven. And forward-deployed operators or engineers or, you know, these… and that’s why they’re forward-deployed, is actually to help build some of those kind of chaos in place and things like that. So people can actually harness the intelligence which they have been promised for many, many years, like, we’ve been… you know, we have way more data than people are consuming.

Not even AI is consuming in the way it should be from an output we’re expecting it to be. I mean, I’m sure the how public models are hungry for data you know, companies could actually harness on their own data and can get the intelligence what they need, you know, and not worry about tribal knowledge so much. I mean, tribal knowledge and humans are super important, and that would, like, completely dominate the real creativity, or that taste, and whatever buzzword people say. But, I think the future is very, like, human and AI, kind of hand-in-hand. And I think the only difference of companies winning over the other company will be this power of intelligence, the right time, right place, and kind of making this all AI data in the back burner, and then just back to basics, you know?

Awesome.

Michele Buckley:

Yeah, well, a lot of first mover advantage available, and still yet to be proven, and… it’s not… you’re not sure who… right, what does success look like? It’s not exactly clear yet. That’s what I suppose the FDDs are coming in to do, is… is to coach everyone and be those pioneers.

Abhi Yadav:

Capitalist economy, right? So…

Michele Buckley:

Y’all know what. In most places. Well, thank you, everyone. I’m expecting Julia to appear at any moment, because we are at the end of time for this part of the session. I know we could keep talking, and Julia, we are prepared to do that, but you’ll need to let us know if that’s what we want to do. It’s such an exciting time in the market, and we seem to… Have a lot of consensus here on this panel, so…

Jaclyn Rice Nelson:

Jay Hack coming on, we gotta hear from him.