Solution Providers Prepare For 2025’s AI Agent Era

'We’re learning how to capitalize on taking that human time, going down, and leverage agents to do as much as possible,' says Corey Kirkendoll, CEO of 5K Technical Services.

Retrieving company policies, persuading customers to upgrade their purchases and interviewing job candidates are some of the use cases for artificial intelligence agents solution providers are leveraging internally and exploring for customers, partners tell CRN.

Executives with AI leaders such as Microsoft, Google and Salesforce have started talking about this next part of the AI journey–which is already having an influence on enterprise technology by some measures. Salesforce reported in December that AI and agents influenced $60 billion in sales during the 2024 Cyber Week shopping season.

Deloitte said in a November report that a quarter of companies using GenAI will launch agentic AI pilots or proofs of concept (PoCs) in 2025, with the share reaching 50 percent in 2027.

Gartner in October said that 15 percent of day-to-day work decisions will be made autonomously through agentic AI by 2028, with 33 percent of enterprise software applications including agentic AI.

AI Agents In The Channel

Corey Kirkendoll, CEO of Plano, Texas-based 5K Technical Services–a member of CRN’s 2024 Managed Service Provider 500–told CRN in an interview that he has leveraged agents for employees seeking policies around time off and procedures plus speeding up the hiring process by pre-qualifying candidates for technology specialist jobs.

AI copilots are still an important part of the AI adoption journey, with customers needing to master the more general AI use cases before adding in autonomous AI agents to perform a more complicated task, he said.

In the future, Kirkendoll hopes to use agents to create marketing material, onboard new 5K employees and for some self-help customer service.

“We’re learning how to capitalize on taking that human time, going down, and leverage agents to do as much as possible,” he said.

Here’s what six solution providers told CRN about how agents mark a new evolution in the AI era.

Corey Kirkendoll

5K Technical Services

CEO

Plano, Texas

We have a partner that we've been working with (for) … getting our phone agent going.

You can train up a model and app and go to that process. Then now you can either talk to it via the chat, or you can actually dial in and call and actually walk through stuff with it as well, which we're finding pretty cool.

Because once you train it up and give it information, it is very good on giving some feedback on some answers.

The other agents we’ve been using in house is that we took our internal policies, procedures, handbooks and things like that and created an internal agent.

If anybody has a question on handbook, PTO, all that good stuff, and they get feedback from that perspective.

What we're also seeing from an agent perspective is leveraging it from a hiring perspective.

What we've been working with our HR person is that we put in a job description, what we do, how it does. And we had several hundred different interview questions that we put in.

It can actually pre-qualify and and actually call in and say, ‘Hey, thank you for calling 5k today. We're going to be interviewing for a tech specialist I. I’ve got a few questions for you I want to get done. And it's recording all this information.

They go through, they ask the questions, they can answer back. ‘That was kind of a generic answer, can you give a little bit more detail?’ You can go back and forth.

Then once it's done, it will send out a summary of the meeting and candidate and what was said and give us a score based on our requirements, based on how the resume looked. And say, ‘Hey, this was good’ (or) ‘this was in the middle’ or ‘no way.’

We're learning how to capitalize on taking that human time, and going down, and leveraging the agent to do as much as possible.

Definitely copilot before (agents for customer adoption). You’re training them up. And then next piece, you want to start bringing in some of the agent pieces. Saying, ‘Hey, now that you tried this and now it's in here and natural, now let's talk about how we can make this automated.’

Agents are definitely going to be where it is. What we're seeing is a lot more custom developments and support of the LLMs and helping them (customers) get through that process.

Adoption is just going to continue to rise over the next year.

Amas Tenumah

Slalom

Global Leader, Service Transformation, AI Innovation

Seattle

The agents parts of the equation–it's now (like) an intern. They can do stuff.

I can now have this bot not only converse with the human, whether that human be an internal marketer or a service person, but it can actually go execute the thing. It can issue the credit. It could make a decision. It can go in and ship the product out. And it could be as autonomous as you want.

And the building of it–not that we couldn't do some of this before, but you *would’ve had) to pay me lots of money, and it'll take lots of time for me to piece all of this data.

How long is it going to take to set this up? Minutes. Oh, by the way, who is going to set this up? There is now a shift from the person who is best equipped to actually build this bot in the very near future is actually the SME (subject matter expert) who knows nothing about technology, because the way you build the bot now is to explicitly in English, tell it, ‘You are a bot.’

‘Your job is, anytime someone comes in, you're an IT help desk (bot). You are friendly. This is what you do. These are the systems I want you to go connect with.’ And then it connects.

The game changer now is we now have these bots, purpose built. You don't want them to have too much power. You want to limit them to–’you are here to do these three things. But you can do those three things on your own, without any human (interaction). Or with very little human (interaction) if the client is conservative. But technically it can go do it all on its own, adjust, get smarter.

In terms of efficiencies, in terms of effectiveness, in terms of accuracy, the outcomes are quite dramatic.

It's so far ahead of the business implications that I think the other side is trying to keep up.

We've done at least 40 of these GenAI bots. … We are finding one of the obstacles to adoption is businesses like predictability. They like to know it's $200 per user, per agent, per month (for example). And now we're saying, ‘it depends’--on how many, volume.

And that's a transition that is in progress. It'll take probably a couple of years. That's one part of the conversation that is a little wonky. We are the implementer and the advisor. But it plays a part into how we deliver value to our clients.

Carlos Marques

Mainline Information Systems

Director, Technology

Tallahassee, Fla.

We're hearing more from some of the major software companies that are developing these things. It's still a little bit early from what I can see.

Today, you have an assistant where you can ask it basic things, like copilot, like–if your company allows you to use ChatGPT–or something, where you can ask it basic things or have it do basic things for you.

The idea of the agents is to be able to give it instructions and have it act more autonomously on your behalf. We're hearing a lot from our software providers, like the IBMs of the world, about things that they're going to be building into their products to be able to make some more mundane and repeatable tasks–instead of me having to write automation to say, hey, every night for this database, go and do these four things, you just explain it the way I just did.

You list it, and it just goes and does it for you. It reports back ‘success’ or ‘failure,’ whatever. But in terms of day to day (tasks), it's still early.

We've been seeing more and more of our major OEMs provide AI assistance, natural language type of support. And things where you're able to train something on documentation, the knowledge bases. And then you're able to take your customer knowledge base, maybe an integration with your IT service management system, like a ServiceNow

People are really kicking the tires on that. We're definitely starting to see more and more use cases where that's going to drive hardware sales because you need GPUs to even do inferencing and things like that for some of these products. We're starting to see those things begin to ship.

The code assistant, that's an easier one (use case). But the system programmer assistant is a little bit more involved because you're doing more risky things as a system programmer on a mainframe than you would be just converting code.

But we're looking at stuff like that as a managed services provider, for example, to lower our cost of delivery. So you can, you can do more (projects) with less people. Therefore, your overall productivity will be better. And then your cost to deliver the managed service will be better and more accurate. And the customer will have a better experience. We are evaluating those types of things

But they’re still working on the adaptability of it, the quick adoption of it.

Mike Strohl

E360

CEO

Concord, Calif.

By creating use cases, we then are beginning to build AI agents to address those use cases.

We tell them (customers) to pick one thing, and once they see one thing start to work–whether it's a single AI or something that's using multiple agents.

Most of the ones that we're developing for customers are around more internal use cases. HR searches, things along those lines. But as they get better at it, we see them looking at opportunities to get more client-facing stuff in order to either drive new experiences or optimize, get ROI, things along those lines from real, real dollars that will come with this for us.

We have clients that are just at the very beginning of all this, and others that are in full force that we're in collaboration with. And it's a fun thing to be a part of the journey. It's also a great consulting business to be in right now.

The agent is helpful because agents get more specific to certain data types. You want to maximize what each one can do, but then you want them talking to a master agent.

Making them smart agents that share information through a single interface is absolutely critical.

I told somebody one day, it’s like those pet robot dogs you can get for your kids that you have to train. They come out of the box and they don't know what they're doing. And eventually they learn different things. This is just much more sophisticated than that.

Grant Davies

Perficient

Principal, Digital Strategy

St. Louis

When I do a demo with multiple agents talking to each other, they (customers)usually go, ‘Hang on a minute. They're talking to each other?’

There's a lot of value in it. I think it's very immature still.

I see a lot of customers trying to use GPT in a way that you'd use an agentic framework. And they're not very impressed.

And then I show them what you can do with the agentic framework, and they're like, ‘Oh, wow, that's a lot better.’

Somebody would upload a document to Claude or a custom GPT, tell it to do something, and then what came out wasn't great. And with the agentic frameworks, because they go back and forth and converse about the topic, the end product that they produce is a lot better.

We're doing a lot of PoCs internally using frameworks with typical customer problems.

The agentic frameworks are fairly immature. Some are really easy to get something up and running, but you turn around and say, ‘I couldn't productionize this.’ Hallucination is still a big challenge across the agents, especially when they start using their training instead of the documentation you gave them. But like anything, it'll get better.

And then a few of the newer frameworks that we're using are more prescriptive in how you define the agents. So for example, a lot of the older ones, you gave them a system message, and you put all your do's and don'ts in there as well as how you wanted them to do something.

But some of the newer frameworks we're using separate it out where it's like, here's what I'm going to give you, here's what I want out the other side. And that's one section. Then a completely different section is, here are the things you should and shouldn't do. So it's getting more mature and better. But it's still early.

Eric Walk

Perficient

Principal, Enterprise Data Strategy

St. Louis

My favorite example (of AI agent use by clients) is the ‘where's my order agent.’ And the reason that's going down an agentic path is because for the biggest of our clients, it's a lot more complicated than calling an API.

These agents that know how to interpret the actual context of the request, they can go figure out which API to call and why to call that one, and how to put together information from multiple API calls.

The next one is using AI agents to make our work faster, to reduce the cost of what we're doing. So if that's maybe a set of agents that are looking at user stories and adding the structured test case–or at least writing the first draft of it–that are collaborating with our consultants to make their work faster and more efficient. We start talking about pair programming.

And then the last area is making our own operations more efficient. So can we build a set of AI agents that can help us write a first draft of an RFP response? That can help us automate the process of producing a case study at the end of a project? Those are the kinds of things that we're working on.

As these new tools come out, they're great. They're helpful. They're driving new use cases.

We were talking about a social media team (of agents), the ability to automate part of the process of generating content for social media, for marketing.

That's something that was not fully automatable, that required, still, a lot of human touches in the 12 months ago range. And now we can build something that's pretty good with a lot fewer human touches because of these agentic frameworks. And that's really what we're talking about–is reducing the human touch using these agentic frameworks.