Joe Inzerillo is a fairly recent recruit to the world of enterprise software, but he's effectively the famous 'Customer Zero' made flesh at Salesforce.
Sometimes you'd love to claim, 'Wow, we really fully understood how this was going to work', but it seemed like a good idea.
It’s an interesting confession from Joe Inzerillo, Salesforce’s first ever Chief Digital Officer, especially when it’s being made on day one of the company’s annual Dreamforce jamboree where tens of thousands of customers have turned up in San Francisco to be told what the future is and how it’s going to work. Fortunately he quickly follows up with:
It really has been a good idea!
Phew!
What Inzerillo is talking about is - go on, have a guess! - agentic AI and specifically its application to empower and enable a Digital Labor future of work that sees humans and agents working in tandem within an enterprise. The theory is straightforward enough as Inzerillo explains:
Humans are very expensive, it's very hard to find great ones. I really like to focus on this concept of humans for impact, agents for scale. To do that, we have to re-evaluate every single system we have, how they talk to each other, how you layer the agendas on top of it, and how you empower the humans to really derive big impact on top of that.
Interestingly he’s the only Salesforce exec during Dreamforce week that I heard bring the cost of human labor into the discussion quite so openly. Most conversations have skirted around the topic rightly perhaps preferring to focus on productivity gains and the opportunities to power additional revenue growth via such collaborative workforces. But of course the cost arbitrage element is never going to be far from most C-Suite mindsets among the Salesforce Ohana in attendance at Dreamforce, so it’s good to get it out in the open a bit more.
Towards Zero
Inzerillo is also on the frontline of Salesforce’s conscious effort to be seen as 'Customer Zero' for its own agentic vision, hence the creation of the new role:
Previously, we had CIOs, but I took over the Office of the CIO, brought in a CIO, and added Mar-tech and a bunch of other things that are sort of adjacent to us [to my responsibilities], using our own technology. That's really the theme [coming] from customers - how do we [Salesforce] use our own technology better?
In terms of the corporate hierarchy, Inzerillo works for Steve Fisher, who’s Salesforce’s Chief Product Officer rather than a line of business executive, such as the CFO as has been the case with his CIO predecessors. This revised reporting structure has tangible benefits, he argues:
Not only are we developing products and doing things where we're sprinting ahead of the product to figure out how to solve these actual challenges that people are going to have in their implementation, paying down tech debt, etc, but we're actually drilling those capabilities back into the product. Between now and [last]February, there's several things that my team actually built and refined that are now being built into a mainline product, and our customers get the advantage of everything that we're doing.
That said, he is working very closely with Robin Washington, Salesforce’s recently appointed Chief Operating and Financial Officer to promote the ‘Agentic Enterprise in practice at Salesforce’ exemplar:
Living the experience of being the Agentic Enterprise, or, more importantly, becoming the Agentic Enterprise, is exactly the shared vision that Robin and I have for how re-build the underpinnings of the business processes and things like that, so that the technology becomes even more capable of shining on top of that.
While Inzerillo doesn’t come from an enterprise software background - his last role was in B2C at Sirius XM, while previously he worked for Disney where he launched the Disney+ streaming initiative - he clearly has a keen eye on developments across the sector outside of Salesforce, not least the fever pitch across AI stocks and the resulting fear of a market bubble bursting at some point, a sort of re-run of the dot-com-apocalypse. His take on this is rather different:
A lot of people are talking about the AI ‘bubble’. But at dinner last night, I was talking to the CEO of another, not a competitor, but a SaaS Type product, and we were talking about it. The thing that we came to agreement on is, if you take two versions, the two halves of AI - the core foundation models and all that stuff that's going on. That is the arms race, unlike no other. And then you have applied AI, [as in] 'What do you do with that? And how do you take it there?’.
If foundation models got no better than they are - let's just pretend everybody gave up and said, 'Yep, this series of models, that's the best we've got' - we could still have 20 years of applied work to extract all the value of that to get to the enterprise. So I think sometimes when people talk about 'the bubble', they conflate these two things. We're still in the early days of figuring out how to adopt this and how to actually bring impact, but we literally could take 20 years off and still not exhaust all the potential of those foundation models.
Apocalypse not now, then...
Changes
That said, innovation and iteration are two things that have characterized the various rapid releases of Agentforce since it went on general release around a year ago. This is not a product that has stood still even in the short space of time it’s been released into the enterprise wild. Inzerillo comments:
If you look at where we were with Agentforce a year ago, we were making multiple trips through the LLMS (Large Language Models), just iterating and cycling and cycling and cycling to get an outcome. When you're testing you're trying to figure it out. You start with the most generous amount of compute to make sure the problem gets solved. And then you look at where we are now and we are far more efficient with what we're doing.
As we start to see these patterns, we're developing our own technologies, planning technologies that are built into Agentforce. We're doing the optimization as we go. I think in the long run we're going to get into a situation where there's an inflection point of downward pressure, but that's because of inefficiency. I think the efficiency curve is going to way offset that.
It’s important to think about agentic tech in a different way to traditional enterprise software, he suggests:
One of the things that I think about a lot with agents is it's not about launching them. It's a different way. It's a different type of software. It's not just like I have some code, I ship it, that's the version, and we're done. Consumers are interacting with this. Based on my background, I'm used to the notion that consumers are driving the actual consumption of these things, and their habits change.
He reaches back into his consumer past to provide a for instance:
The questions that people asked ChatGPT two years ago are very different than the way that they ask those questions today. They used to look like Google queries. They now look like conversations, and so the models are drifting for that reason, or the agents are drifting for that reason. And then obviously our data changes. We have a bunch of product launches, we re-name a couple of things.
With all that change ongoing, keeping ahead of the game is clearly going to be competitively critical. Inzerillo is aware of this. That’s where being 'Customer Zero' really matters:
It's not just about launching these things, it's about continuing to improve or making them better over time. Our product team certainly knew that we were going to need something, but the clarity of what we needed is what we developed as 'Customer Zero', and then we were able to immediately transition that into the product. It's really the difference between being a software provider and being a practitioner of using software.
In practical terms, Inzerillo’s team and Fisher’s team meet weekly to discuss what’s working in Salesforce’s offerings and what problems need to be addressed or opportunities pursued. The idea is that Inzerillo's people will run ahead of the curve on what’s needed, focusing on 'the next big thing', while Fisher’s lot focus on ensuring the current iteration works:
I act a little bit like an advanced team in that respect as well. It's not obvious, in some cases, what the right answer to some of these problems is, but if I go and do four different variations of it, I can come back with some very valuable recon, and we have the scale to really spin these wheels.
Impact
Going back to the idea of the combined human/agent workforces to come, Inezerillo adds:
One of the things I really emphasize is that the great thing about our platform is you can use it to also instrument humans. So if you have some job or a portion of a job that was done by a human that is now done with an agent, you need to know where the humans were as far as quality goes,. Those critical KPIs that drive your business, you need to make sure the agents are achieving and exceeding those.
Another thing that we have developed is a rubric that allows people to plug in a couple of business members all the technical calculations about the testing and variation of those, as opposed to humans. As the agents, sort of experiment themselves, we can quantify what that actual impact is on the company. In a lot of cases, it comes right down the bottom line, like our lead generation agent - what about the pipeline. Is it creating one? Those become real core things.
All of this will vary on a case-by-case, industry-by-industry basis of course and be very dependent on context:
Each agentic use case is going to have slightly different metrics, and then there'll be these sort of more general. Is it efficacious? Is it is it working with things that it needs to work with? That's what we're trying to do.
My take
Practice what you preach. Eat your own dog food. Show, don’t tell. Take your pick on cliché of choice, but what the 'Customer Zero' helping to shape future product direction in a customer-responsive manner. Inzerillo has a big job on his hands, but one that’s worth the effort.
October 16, 2025
