Greg Shewmaker

This is not an IT initiative. It’s not just about IT introducing AI agents, or, the Head of Operations introducing robotics into the warehouses or to the manufacturing line. This is about, how do we balance this with our human employees? And it’s not so simple.

A highly pragmatic –  and as such, very welcome  -assessment to balance against the current wave of agentic AI hype that has built up over the past six months or so. It comes from Greg Shewmaker, former SVP of Global Operations and AI at global talent solutions provider Adecco Group. Adecco is one of pioneering flagship use cases for Salesforce’s Agentforce platform offering and Shewmaker has been at the forefront of how that has been rolled out across his organization.

But he’s now taken on the role of CEO of r.Potential, a new company spun-off from Adecco, set up in partnership with Salesforce to provide senior business leaders with assistance on navigating a brave new world of work, the world of Digital Labor. It’s a world that presents enormous potential for organizational transformation and productivity gains, but also one that may seem daunting to many C-Suite execs today.

For his part, Shewmaker is a fan of the Digital Labor concept, arguing that it forces us to think about AI with a much higher level of responsibility. He also buys into the thesis regularly propounded by Salesforce CEO Marc Benioff that today’s is the last generation of leaders to manage human-only workforces. But there is a combination of FOMO – fear of missing out – and nervousness about how make the necessary changes for a Digital Labor age that business leaders need to tackle:

Every leader is at the same point at the starting line, not really sure what to do in this world of agents. They know they need to do something, but they’re not sure what to do. That was the initial need that we saw. Then we decided to say, ‘We [as Adecco] have all these decades of labor data, and we’ve been working in workforce configuration for a really long time. Really all agentification is is just the digital version of that, so let’s use that as a starting point.

He adds:

I think more progressive leaders are already thinking, ‘Hey, this is a major change initiative within the company’….I don’t think it’s an intelligence or a technology issue; it’s a strategy issue, and it’s going to rest squarely on the shoulders of HR leaders. How do we introduce this into the organization? If you see AI simply as software, as a low cost alternative to humans, you’re going to get some short term gains. But if you see this as a way to amplify existing employees and allow them to do more of what they’re really good at and less of what they don’t enjoy doing, companies that are thinking that way are going to be the leaders of the future – and that starts in HR.

Pause for thought

What r.Potential sets out to do is to build what Shewmaker calls “units of potential” and, again, these are couched in a very ‘hype-resistant’ framing:

These are combinations of work that we believe humans should continue doing, or want to do, inside a company, coupled with the work that agents are actually capable of doing beyond the hype. That is relative to the situation of a company. So it might be different in France than it is in San Francisco because of regulations and legislation. It might be different for brands that are very customer-centric or let’s say a luxury brand versus a back-office customer support organization, for example. Just because the technology is feasible, doesn’t mean we should actually be pursuing it, is kind of the bottom line here.

Getting caught up in the agent hype without thinking about the strategic implications would be a major error, and one that Shewmaker fears it is inevitable some will make:

I think we’re going to see, over the next year, a lot of companies adopt AI agents very quickly, which is great because we need to do that and we need to learn by doing. But I think if you have that short-sightedness, I think what’s going to end up happening in a very short amount of time is, it’s probably going to end up breaking the system. I think what we’re going to find out is either you’re going to turn the agents loose and you’re going to give them such autonomy that they’re going to make so many mistakes that you’re going to have to back out of those things, and we’ve already seen that with a few companies as example. Or secondly, if you don’t do that, you’re going to have to keep so many humans in the loop that it’s not going to be financially viable, because now potentially you’re installing extensive agents and you’re still employing just as many humans.

I think we’re going to have to be really sensible about how we do this. Again, it’s about AI agency. We shouldn’t just be doing this for the sake of AI. We should be saying, ‘Hey, where does this make the most sense? Where are the biggest problems or opportunities? Where can we go and launch new products or make new inventions that we couldn’t do as humans individually?’, or ‘Where can we give people access to new products and services that they wouldn’t have access to before?’. Thinking like that will create use cases and potential Return on Investment that will make it worthwhile. If it’s just about, ‘How do I reduce costs in a particular department?’, I think those efforts are going to be dropped pretty quickly.

Stop and ask where agents are most appropriate for your organization or industry, he argues:

In the world of AI, we always talk about speed as being most important, but I also think we have to now be more careful about where do we invest and where do we not invest? Everyone’s going to have to go through that journey, so I do think that there’ll be a bit of a pause to say, ‘Let’s think through this more clearly’.

Dog food

As an early adopter of Agentforce, Adecco has ‘eaten its own dog food’ here and picked up some valuable learnings in the process. Adecco had a dual approach to AI adoption and getting employees comfortable with AI. First was what was called Everyday AI, which was about providing safe access to the likes of ChatGPT, to get people comfortable with using AI tools and engaging with the general technology.

On a more revolutionary scale, there was also a Strategic AI strand, which includes using Agentforce to build a digital workforce, with agents integrated into what the Group calls “the talent supply chain” with its  largest enterprise customers.

Agents are being deployed to do work that previously was impossible for Adecco, states Shewmaker. On average, the Adecco Group receives about 300 million applications every year, but traditionally, has only been able to get back to about 10% of those. It also receives millions of job vacancies to fill from corporate clients. Normally less than half of those are filled. It was never practical for Adecco to hire tens of thousands more recruiters to try to improve those stats dramatically. But with Agentforce, that traditional recruitment work is divided up between human and digital workers, with agents able to engage with 100% of applicants within seconds of them applying for a position. Agents are also able to go through hundreds of thousands of resumes or CVs using all-new workflows, and empower more effective re-deployment of people. Shewmaker explains:

Re-deployment is the absolute Holy Grail of the Human Capital business, and it’s been so difficult for us to do that in the past. We just did something with two different agents in Slack which took just 30 seconds to identify what we would call the full potential of people. One is an agent that goes and talks to the hiring manager to say, ‘OK, I got your job description, but what is it you’re really looking for? When? How do people succeed in this role? What does success look like? Where are people who have maybe not succeeded in this role?’. Then on the other end of it, an agent is going out to individuals, the employees or the candidates, and saying, ‘Hey, tell me more about yourself. I got your resume, but I really want to know who you are’. If you can make a much deeper match between those two things, I think that’s a game changer for this industry, and that’s something that we’re really looking to do.

For all that, Shewmaker recalls that it wasn’t actually the potential of agentic tech that was the focus not that long ago:

The thing that we started talking to Salesforce about a year ago was actually Data Cloud, because one of our biggest challenges is that we have 45 different instances of Salesforce around the world. Every country essentially has their own. As we serve global customers, they expect standard service regardless of the country, and that was impossible for us to do with using Salesforce as the front office. Salesforce has been the front office for recruitment long before the AI era, but it was completely disconnected. So initially it was Data Cloud. Then Agentforce was launched last August, and that became sort of a natural extension of that. But initially it was just that we needed to give customers information and visibility that we couldn’t do very easily.

The close relationship Adecco has with Salesforce has given it considerable clout it seems, clout that carries on with r.Potental. Shewmaker explains:

We are absolutely pushing the boundaries of all the Salesforce products right now. When it comes to AI, we are that edge use case, and I think that gives us an advantage that you wouldn’t normally have as a start-up, in that we get access to product and engineering teams that you wouldn’t otherwise have. But we are frustrating them, and we are coming back to them and saying, ‘It doesn’t work yet. We’re not there yet. Come on. Come on’. You know, we’re on the edge of this thing. We’re trying to get out there fast enough so we can sort of look backwards and help enterprises with that journey, and help them map their way forward. So we are out, let’s say, several months ahead of where Salesforce is today. That’s good for them because they’re learning, and good for us because we’re learning what’s possible, what we have to go build ourselves, and what we could potentially leverage from them.

My take

There will be a lot more to be said about r.Potential in the months to come as it gains traction, building up to “a huge launch” at the Dreamforce conference in October.  

For now, I’m immensely drawn to the grounded nature of the early messaging from the firm, with Shewmaker insisting:

We are trying to help companies enter this era in a responsible way. We are not naive in the sense that we’re not saying, ‘Hey, we’re 100% team human. We’re anti agents’. We’re just saying that whatever you do, whether you plan to identify 100% of your workforce or zero percent, you just need to do it responsibly and thoughtfully, and we want to provide the information for companies to do that.

What will be interesting to see as time rolls on will be the reaction of the major Systems Integrators (SI), all of whom have their own Salesforce practices these days and all of whom are getting stuck into agentic AI consulting. Shewmaker states firmly that what r.Potential does will not compete directly with them – “I don’t want a single service employee or service unit within the company!” – and that he wants to work with the SIs.

For now, this all looks heavily Salesforce-centric, but Shewmaker makes the point that while Adecco is a big Salesforce shop, it’s very multi-vendor in its tech stack, citing Microsoft, OpenAI and Anthropic as other suppliers in use. So long term what other vendor customer bases might r.Potential reach out to?

Definitely one to keep a close eye on. 

Source link