As agentic AI hits 'a difficult age', Salesforce's Forward Deployment Engineers are on hand to, well, hold enterprise hands. Jennifer Cramer explains how it works

We’re at an interesting stage in the enterprise roll out of agentic AI. The case has been made for the benefits of the technology and there are flagship pilots out there that have been able to report back with positive and promising results to show. 

But for a lot of organizations, the majority in fact, those pilots are as far as the agentic journey has progressed to date. Agentic AI is, you might argue, at that 'difficult age' of growing up. What matters now for vendors is how to encourage adoption to provide the necessary success stories to encourage other enterprises to come on board. 

Take Salesforce as a case in point. Since the launch of Agentforce at Dreamforce 2024, the company has pivoted around agentic AI. The results have been impressive on the face of it - more than 29,000 Agentforce deals have now closed, according to the latest fiscal update. But not all of those are paid-for deals, rather exploratory pilots, and there are over 150,000 customers for Salesforce worldwide, so there’s a lot of room for growth there yet...

Forward

The challenge for Salesforce then at this point is how to ensure that customers who are dipping their toes in the agentic water are then ready to go for the full swim. One initiative to enable this can be seen in the shape of the firm’s Forward Deployed Engineering (FDE) team.

Forward Deployed Engineers are not unique to Salesforce. Their origin as a job function can be traced back to Palantir in the early days of SaaS, where people in such roles were known as Domos. They were sent in to customers to work alongside their own IT teams to integrate the new breed of cloud applications with legacy estate. 

The same principle is being put into practice today at Salesforce, as Jennifer Cramer, SVP, Forward Deployed Engineering and Customer Success for AI Product, explains:

I've actually been on the journey with the Forward Deployed Engineering team since the very beginning, actually, even before we launched Agentforce almost two years ago now. The Forward Deployed Engineering team, they are most definitely our most technical and technologically-savvy Agentforce experts. They sit at the intersection of product and services.

Operating in ‘pods’, an FDE team is hybrid by nature in terms of its skills. That engineering expertise that Cramer speaks of is clearly crucial, but this is also a highly customer-facing role so possession of diplomatic and communications capabilities is also enormously important. Cramer says:

I would say they are a new blend of consultancy and deep software engineers -  the consultant has that diplomacy, if you will, and the engineer has the deep technical experts....The pods are in in teams of three, where we're merging the skill sets, so that you have more junior resources who are learning and expanding their their career, and you have more senior people who get [to have] the executive level conversation, but also can talk tech and build at the same time.

Customer engagement

There are a number of objectives behind the FDE push, explains Cramer, but it all begins, of course, with driving customer success to build up the ranks of positive use cases out there in the wild. Engagement comes at different stages of maturity in the Agentforce end user journey:

We work with customers across all industries, across all shapes and sizes, and at different phases of their journey. There are a lot of customers where pilots are incredibly important to their future success with Agentforce, and we want to make sure that, even in that pilot stage, they are doing the right things, learning the right skills, and setting them up for future success. So we will get involved with customers, both in the pilot stage as well as in the future deployment.

She goes on:

We want to make sure that our customers are deploying agents in the best way possible. Within that scope, we consider ourselves both builders and master fixers. Sometimes our customers are early in the journey, and we're helping them get their first agent up and running; in some cases, they've started on their own, and they're stuck in some capacity, so we'll get in and help unlock the value that they're trying to get out of Agentforce.

A lot of this is done alongside customer teams, often, but not always, on-site:

Customers will be doing it on their own [with] really strong development teams, and so we'll get in on site, shoulder-to-shoulder with them. We're not just there to build for them [in the traditional sense where] you build, you deploy, you implement, you're done with that, and then you leave. In some cases, they're building, we're next to them; in other cases, we're building, they're next to us, but at the end of the day, we're leaving that account, and they have the skills and expertise to take it forward.

Stuck?

When customers are stuck, it tends to be around engineering things, says Cramer, or through trying to ‘boil the ocean’ too early. She points to the use case of transport and hospitality specialist, Engine:

One of the reasons they were successful is because they started with a very unique, very specific use case. They had a condensed problem statement where they were getting way too many cancellations for the flights. They said, ‘We're just going to solve this one thing'. They tried not to ‘boil the ocean’ with their data strategy. A lot of our customers will get into an analysis paralysis situation because they think they have to clean up all of their data at one time, and everything has to be perfect.

With Engine, they took a unique set of data, a unique problem set for these cancellations, and they deployed that in two weeks and knocked it out of the park. And then once that was done, they started building more capabilities on top of that and expanding to new agents.

It was the same sort of thing in practice at retailer Williams Sonoma, Cramer notes, one of the flagship Agentforce deployments:

They have been on the bleeding edge of this new technology since the very beginning. We really got in with them early days. They've been incrementally rolling out...They started with [the] Pottery Barn Kids [brand], and then they did a rinse-and-repeat against eight of their brands. Those are the themes we're seeing - have a roadmap, know what your vision is going to be and what your path is, but start small and grow from there, because you'll learn a lot along the way. Even on data, of course, data is important, data is foundational  and data needs to be in the right place with the right context - but you don't need to fix it all before you just get started.

Feedback

A second role the FDE team has is to take the real-world experiences they encounter through customer engagements and feed that back to the product development group inside Salesforce. This in turn empowers functional enhancement and extensions when the team comes across customers who are looking to do things beyond what’s on offer in the core Agentforce offerings to date. Cramer says:

We'll build that not just for that customer, but for many customers, and then give it back to the product team - obviously without the customer's proprietary data and code and whatnot - but we'll give our core build back to the product team to re=integrate into the product. So that feedback loop isn't just about giving feedback and giving it back to the customer, but also about helping to ship and extend the product in a in a faster, more efficient way.

A third aspect of the FDE work is around what Cramer calls "up-leveling" the Salesforce eco-system, which includes both customers and third-party industry partners:

Our customers and our partners alike are loving this model, because we are raising all ships. We are there because of the speed at which the technology is evolving, the speed at which we're releasing new features and functions,. It is really hard to keep up. We're really working side-by-side with our partners, with our customers, and they're learning as as we're deploying, then we're leaving, and they are better off. They're able to take that learning to their whole ecosystem of customers. So ultimately, it's a raising all the ships kind of motion. We have seen massive success with our customers, but our partners love it when we're in there too, because we are extending their capabilities in a more natural way, directly with the customers.

The future

When it comes to the future, the role of FDE is one that will clearly evolve. At present, engagement with the team comes ‘invitation-only’ via account managers who make the introduction between the FDE division and the customer. There has been talk of it becoming a paid-for service that organizations can request and access, but there’s no timescale on this, and Cramer has no additional information to add here. The team is expanding however - there have been estimates of a goal of growing to a headcount of 1,000 - but again, exact numbers aren't up for disclosure at present.

What Cramer will say is that the FDE team has already seen evolution in what customers are looking for, even in the relatively short time that the agentic push has been front-and-center:

I would say early on we were seeing more stuck customers, but a lot of the times they actually weren't stuck technologically. It was early technology and customers were stuck in that they didn't know where to start. They knew that they had to do something - every company is under this imperative to use AI to transform their business - but they didn't know where to start. It was kind of, 'Help me build a roadmap, help me understand my use cases, help me prioritize my use cases'.

Our customers are now maturing to a place where they're coming to us for innovation. They're saying, 'We've got this part down, now we want you to help us deploy voice', or, 'We want you to help us go into this entirely new space that we've been eyeing for a while'. Some customers just come to us for innovation. They weren't stuck in any way, shape or form, but they were ready to explode their businesses, and that's where our team actually engaged with those customers to help them with very niche requests that they had in order to get out of the out of the gate quickly.

She adds:

Where we are no longer involved is in what we call ‘out-of-the-box implementations’, because we feel they can now do that. They have the skills to do that. Where we're re-investing our capacity and our expertise is in newer features and functions, like voice. What we like to say is we are co-innovating with our customers.

My take 

An interesting insight into what looks like a highly-productive customer-facing in-house initiative in service to the wider agenda of ‘normalizing’ agentic tech across the enterprise. What happens next will be interesting to observe. There’s clearly potential to monetize this service on a grand scale, although I am intrigued as to whether there comes a point in the future when Salesforce steps on the toes of its third-party implementation partners here. But that’s all to come. I’m going to aim to catch up with Cramer and her team at this year’s Dreamforce in September to get an update on progress and continued evolution.