Bridging the Capacity Gap: Why Enterprises Are Investing in Agentic Digital Workers

Enterprise appetite for AI agents is far outpacing the ability to operate them. New Futurum research finds that nearly two-thirds of decision-makers say manual, repetitive work consumes more than 40% of employees' time, and two-thirds plan to invest in digital workers within 12 months. Only one in 10 organizations runs mostly autonomous AI today.

Analyst Keith Kirkpatrick, Vice President and Research Director at The Futurum Group, is joined by Somya Kapoor, CEO of IFS Loops, to dig into Futurum's newly published report, “Bridging the Capacity Gap: Why 66% of Enterprises Are Investing in Agentic Digital Workers.”

Kapoor explains what stands between an agent demo and a digital worker trusted to run in production, from monitoring and guardrails to escalation paths, access control, and audit trails. Agents change the economics of process change itself. Kapoor points out that design time and runtime now happen together, so a workflow that once needed a team of developers and a nine- to 12-month change cycle can be adjusted with natural-language instructions. That moves the bottleneck to the organization, since her customers estimate technology is 20% to 30% of the work and changing people and processes accounts for 60% to 70%. Because agents make their own decisions and can run around the clock, ERP-connected processes need deterministic behavior and a clear audit trail when something fails.

Key Takeaways:

🔹 Building an agent is no longer the hard part. Kapoor jokes that her 11-year-old daughter can build agents. Reaching enterprise production means answering who monitors them, who trains them, who approves their work, and how they meet security standards like SOC 2 and ISO.

🔹 Governance decides whether ERP agents can be trusted. Agents make their own decisions, so processes like purchase orders need deterministic behavior and an audit trail that explains every failure before an auditor comes asking.

🔹 Change management outweighs the technology. IFS Loops customers put technology at 20% to 30% of the effort and people and process change at 60% to 70%, which is why outcome-based digital workers that deliver early wins matter.

🔹 Autonomy is a journey, and not always the goal. Regulated processes may keep a human in the loop, and Kapoor compares a new digital worker to an intern that may be only 40% accurate on day one but grows into the job within two to three months.

🔹 Imperfect data is no reason to wait. One customer found that only 60% to 65% of its parts were correctly identified across systems, and used agents to build a parts data dictionary while running material replenishment at the same time.

🔹 Reclaimed capacity funds growth as much as savings. Two manufacturers using the same supplier order manager digital worker took different paths. One absorbed rapid growth with its existing 20-person inbox team, while the other looked at redeploying staff toward better service for vendors and customers.

Enterprises that build governance, escalation, and change management into digital worker deployments from day one will convert reclaimed capacity into growth while competitors still wait for perfect data.

Read the full Futurum report at https://go.ifs.com/bridging-the-capacity-gap.

Learn more about the 8 lessons to successfully deploy AI agents at https://www.ifs.com/en/insights/assets/8-lessons-to-successfully-deploy-ai-agents#gatedform.

Watch the full video at sixfivemedia.com, and subscribe to our YouTube channel so you never miss an episode.

Disclaimer: Six Five Media is for information and entertainment purposes only. Over the course of this video, we may discuss companies that are publicly traded, and we may reference their equity share prices. Nothing discussed during this webcast should be considered investment advice or a recommendation to buy or sell any security. We are not investment advisors, and you should not rely on this content as financial advice. Six Five Media collaborates with technology companies and industry leaders to produce research-driven interviews and multimedia programming for enterprise technology audiences.

‍

Somya Kapoor:

Where do I build the escalation process? Do I want to make sure my data is not getting exposed? The governance aspect of it. Who has access to it? Who can train the agent? Who will give the approvals? All those elements lead to the very fact of building and taking this to production readiness, right?

Keith Kirkpatrick:
Hello, and welcome to the Six Five Virtual Webcast. I'm Keith Kirkpatrick with the Futurum Group. Today, we're taking a deeper look at Futurum's newly published research report. Bridging the Capacity Gap. Why 66% of enterprises are investing in agentic digital workers. Today, I'm joined by Soumya Kapoor, CEO of IFS Loops. Welcome, Soumya.

Somya Kapoor: 

Hi, Keith. So nice to see you again.

Keith Kirkpatrick: 

Nice to see you. So today we're going to talk a little bit about what the research is telling us about capacity pressures that are facing enterprises and what it actually takes to move digital workers into real world operations. So, one of the findings that kind of jumped out at me after doing this research is the sheer scale of the capacity problem. We found that nearly two-thirds of the decision makers we surveyed said that manual, repetitive work consumes more than 40% of employees' time. So from your perspective, what does that tell us about why digital workers are actually becoming an enterprise priority right now?

Somya Kapoor: 

Absolutely. I think, Keith, automation is not new to any enterprise, let alone a customer in the industrial environment as well. We've been doing automation for the last 30 years. This was very manual. It was always when you wanted to do an automated process, you had to get a bunch of developers, they needed all the requirements, then they go build. And it took nine months to 12 months to do the change management cycle. With the agents, what has changed now is the very fact that this can be done, both design time and runtime can be done together. So if you want to change a process, based on simple instructions in natural language, you can go ahead and do that. It's relatively easier. So the engines that are coming to your environment to do the automation are now very outcome-driven. So you don't have to be bogged down in thinking, oh, you know what? I want to automate the process. What outcome am I driving? The outcome is very evident from these digital workers that we've designed out of the box.

Keith Kirkpatrick: 

So I'm curious that focus on outcomes, does that sort of also speed up the time it takes to get to the business results that people are looking for?

Somya Kapoor: 

It kind of helps people understand why I'm bringing technology. It's not bringing technology for the heck of bringing technology. It's the technology to drive an outcome, which leads to the change management that it needs to bring as well. Most of the customers that we speak to, they say technology is 20 to 30 percent of the work, change management is 60 to 70 percent of any organization. Changing people processes and how they're used to doing their jobs is very different than just bringing a technology. The technology that we have today very much can do that. That's why we've designed digital workers that are outcome-based, gives you sort of a roadmap in your environment, where do you want to start from, how quickly you can get wins to show management on the things that you're bringing, and how you can drive them, because each of these digital workers are additive in nature.

Keith Kirkpatrick: 

You know though, Soumya, one of the things that I also found in doing this research is that there is a significant maturity gap. We found that two-thirds of the organizations we surveyed, they plan to invest in digital workers within the next 12 months, but only one in 10 is actually running mostly autonomous AI today. What do you think needs to change within the enterprise to close that gap and turn investment intent into operational adoption?

Somya Kapoor: 

You know, actually a very good point, right? Keith, I think getting to autonomous is a journey. And in some cases, you probably don't want to get to autonomous too, because you are being bounded by regulations to have human in the loop, right? So I think it's the hindrance of the process of people. So we've learned eight lessons. We have another paper that we write. It's actually trying to understand the outcome that you want to drive of the automation of the technology that you want to bring in and then bring people process through the loop in kind of doing this stuff. Apart from the technology itself, you have to see this piece as bringing an intern within your environment. And can that intern eventually do a job of an employee or can do partial skills that an employee does? is the key difference of how that happens. In most of the time, what we realize is people don't have the right systems that they're connecting to a process that they would like to drive an outcome from. They're looking at demo wear technology and overly inundated with too many people coming and hitting you with agents. And then, you know, once they do narrow down a process, it's actually bringing people along. where they're seeding the process and being part of that transition and transformation that change management aspect of it as opposed to being hindrance and say, oh it doesn't work on day one itself. Yes, probably it does that it's doing the job only 40% correctly but will it get there in the next 2-3 months. it absolutely will. So that journey cycle is very different for every customer and being on the journey cycle of bringing people, process and technology is a convergence where it happens in getting it to autonomous.

Keith Kirkpatrick: 

You know, it's interesting. One other thing that really kind of stuck out to me that came through the research is the need for predictability when we're talking about AI. They need AI that can operate within an enterprise's core systems. with the proper level of governance, auditability, a clear model for escalating real exceptions, real genuine exceptions to real live humans. So what does that operating model need to look like for leaders to actually trust digital workers with end-to-end processes that hopefully will become fully autonomous at some point?

Somya Kapoor: 

You know, Keith, this is such an important question that I think every enterprise has to look at because every customer meeting or every partner meeting that I get into, they're like, oh, we can build it. And I'm like, yeah, sure. Go for it. It's one way to build a demo ware. software and show management, and it's totally different from taking a demo solution to production-ready scale, especially in enterprise-grade levels. It's not about building agents. I kind of tell everyone my 11-year-old daughter can build agents right now. But how do I monitor? Where do I find the guardrails? Where do I build the escalation process? Do I want to make sure my data is not getting exposed? The governance aspect of it. Who has access to it? Who can train the agent? Who will give the approvals? All that thing, including security. We haven't even gotten into security. You know, SOC to compliance ISO. All those elements lead to the very fact of building and taking this to production readiness, right? And all those aspects are something that we're doing in our digital worker platform from IFS Loops is where we're taking the plumbing, the thought process, the decision around these downstream aspects that you have to do. So you can only worry about the process and the outcome that you want to drive and bring your people along on the table, as opposed to thinking at every element of deploying an agent and then monitoring and triggering it as well. Because these agents, unlike automation historically, has their own decision-making power. So you are leaving them on for 24 seven, guess what they might do? We just don't know. So the governance on them is massively important, especially in highly regulated industries. And when you're exposing them to ERP processes, they have to be deterministic in nature as well. So when a PO doesn't get processed 99.9% of the time and doesn't have an audit trail or why it failed, guess who's going to come and audit that system?

Keith Kirkpatrick: 

And that opens up the organization to significant risk. You know, it's interesting. We conducted customer interviews as part of this project, and it really surfaced something really interesting. If we think about the conventional wisdom in AI, there's this thought that, well, you need to make sure that you have all of your data properly organized. But these interviews surfaced a really interesting lessons. found that organizations don't necessarily need to wait until every data quality issue is resolved before getting started with digital workers. And we found that digital workers actually can surface problems hidden inside existing processes. So from your perspective, how should enterprises really think about data readiness as they move from pilots to production?

Somya Kapoor: 

Absolutely, Kira. Kira, I've been in the enterprise, I'm going to date myself for 25 years, and I feel like every decade kind of comes with a new title and a new thing to clean data up, right? Was it Chief Data Officers, Chief Digital Officers, and I don't know what we're now going to call Chief Agent Officers or something. And it all starts with data cleansing. And, you know, been now in this industry for three decades, your enterprise data is never going to be perfect. So are you going to wait for the holy grail to happen? The good news about this technology, the agentic technology, is that it does give you the gaps in process, not just data, in your process. So getting deployed fast enough tells you, for example, one of our customers uses it to do a part search and then do material replenishment on those parts itself, right? What they've realized is during, you know, their parts in the system were known only 60%, 60 to 65%. And there were about 30% to 35% where parts, even association of what that part was called to the part ID number was mismatched in different systems. So they didn't wait to get all the parts perfect, really aligned and tagged appropriately. What they did was they used a system to build a data dictionary around the parts and keep them evolution, avoid deduplication of the parts while downstream, working on the material replenishment, process as well, right? So bringing the agentic technology will help you understand not only the data gaps that you have in the environment, in this process they also kind of helped understand where the approval process should be going, in who should be adding the net new approvals for a part coming or a mismatch of parts happening, who are the experts in the environment, so it'll streamline to your process in terms of how to use the evolution of the process so the system gets running automatically and seeing the output that it needs to drive. There's endless examples that we are hearing, even in the procurement side of the house, where some POs, we just didn't know how work associated with the vendors. And then when the exception handling happens, when the agent finds an exception and brings a human in the loop, that interaction is actually helping us evolve the process and learn And the agent is learning at the every step of the way from an interaction standpoint to keep improving that process and evolving and getting much, much more autonomous.

Keith Kirkpatrick: 

So finally, if organizations are actually able to reclaim meaningful capacity through agents, This research has shown that leaders really want to redirect that capacity towards reducing costs, growth, faster execution, and strategic work that has traditionally been relegated to the sidelines simply because there was no capacity. Looking ahead, what would successful adoption of digital workers change about how enterprises actually deploy human talent and how would they measure the value of AI?

Somya Kapoor: 

I think what it's giving, so I'll give you an example of two manufacturing companies that we've worked with, not going to give their names out. One is on a very fast growth trajectory path. And the other just has to keep its lights on, on, you know, midsize growth that they have to achieve. Both of them are using the same digital worker, which is a supplier order manager, and seeing closer returns. One organization has about 20 people manning an inbox that are now seeing the benefits that they're on a super growth path and they can manage a lot more keeping those people intact. Whereas the other one is looking at the very fact that, oh, now I have three people manning an inbox and I have five inboxes. Can I repurpose some because I have capacity so they can take on more processes in terms of creating a better survey for the vendors or in terms of kind of serving their vendors and customers better? So at the end of the day, this efficiency gain can be refactored into not just reducing the capacity of headcount, but can I better my experience with my customers? Can I understand my customers a little bit more better in terms of interaction? Can I enable my employees with better business operations to run the system much, much more smoothly? Because that is what is going to be your competitive advantage. At the end of the day, every business is looking for growth. at a higher capacity, you know, a higher efficiency rate. I think that's been the standing, and this technology is kind of giving them the capability of doing that.

Keith Kirkpatrick: 

Well, it certainly sounds like there is a really sort of, you know, unlimited ceiling in terms of the benefits it can provide, that these digital workers can provide to customers. So we're looking to see how IFS loops in the sector progresses over time. Well, thank you very much today, Soumya, for joining us today.

Somya Kapoor: 

Thank you so much, Keith, for helping us and running the report and giving us feedback from our customers. It's so good to see all the feedback and also unsolicited remarks for them as well. And it looks like they're having fun with the technology that has been given.

Keith Kirkpatrick: 

And thanks to all of you out there for tuning into this Six Five Virtual Webcast. Be sure to check out the full Futurum research paper for a deeper look at the findings and enterprise deployment lessons that we discussed today. And don't forget to hit subscribe, follow us on the socials, and check out all of our coverage at sixfivemedia.com. Thank you, and we'll see you next time.

‍