Edge AI Has Solution Providers Eyeing New Frontiers, Even Outer Space

‘It’s insanely exciting, but it’s also one of those [things] where there’s no road map,’ Mark III Vice President of Sales Chris Bogan said of AI on the edge.

One of Mark III Systems’ latest forays in the world of edge computing and artificial intelligence might take the solution provider to outer space, a frontier its customer calls “the ultimate edge case,” Mark III Vice President of Sales Chris Bogan told CRN.

The project, for a private aerospace company, calls for a system that can withstand the operational conditions and requirements of space. And although the project is still in the design phase, it showcases the expanding use cases and roles for solution providers with a technology still being defined.

“The good news and the bad news is this has never been done before,” Bogan said. “So it’s insanely exciting, but it’s also one of those [things] where there’s no road map.”

[RELATED: The 2023 Edge Computing 100]

This year, AI is on the minds of solution providers, their customers and their vendors thanks to generative AI offerings such as ChatGPT, Microsoft Copilot and Google Bard.

And while edge AI predates the current GenAI hype, a combination of reinvigorated customer interest in data hygiene, the endurance of remote work and the complexity of edge AI and other factors cement the need for solution providers in bringing this in-demand technology to customers.

Growth In Edge AI

An August report from Gartner shows the growth in edge AI. The research firm predicts that by 2025, more than 55 percent of all deep neural network data analysis will happen at the point of capture in an edge system, up from less than 10 percent in 2021.

“Demand for edge AI is growing to enable the processing of data at the point of creation at the edge, helping organizations to gain real-time insights, detect new patterns and meet stringent data privacy requirements,” according to the report. “Edge AI also helps organizations improve the development, orchestration, integration and deployment of AI.”

A story that illustrates the importance of solution providers in edge AI comes from Phillip Walker’s work helping a clothing manufacturer get usable results from a nearly $500,000 investment in an AI and ERP project.

Walker, CEO of Manhattan Beach, Calif.-based Network Solutions Provider—a member of CRN’s 2023 MSP 500 whose partners include Microsoft, Google, Apple and SAP—told CRN that the AI received a different type of intelligence than originally planned. The customer had issues with misplaced, unlabeled and fragmented data, preventing a unified, integrated platform.

“The business process for the company was extremely different from how they worked in real time,” Walker said. “The lack of well-documented manual processes and department-specific workflows created issues and hindered the AI project. The loss of knowledge and expertise due to undocumented workflows and inconsistency in how AI models are developed and used can affect the overall quality of the AI solution.”

It’s up to vendors and solution providers to come together to fill the gap of job-seeking data scientists, Walker said.

“Vendors need to empower the channel to be the solution,” Walker said.

When it comes to talking to customers about data hygiene, Walker said that hype around GenAI has helped his practice.

“It’s more directive,” he said. “It’s also helping people understand because people want the result. They don’t want [to do] the labor.”

SADA Navigates Video At The Edge

Miles Ward, CTO at Los Angeles-based SADA—a Google partner and No. 108 on CRN’s 2023 Solution Provider 500—told CRN that most of his edge and edge AI projects have been with customers with thousands of locations and include work in video processing.

Use cases with video span use cases such as grocers making sure shelves are stocked to schools seeking gunshot detection to protect students.

Solution providers like SADA have plenty to navigate to stand up an edge AI project, no matter the vendor, Ward said. For one, the projects are complex. For shelf stocking, for instance, cameras need to incorporate object recognition, location data and time series data and then reconcile all that information, Ward said.

Bandwidth and connectivity are needed for consistent deployment of software at the edge—that’s AI, or “stupid software,” as Ward calls it.

“A lot of that has to operate locally even if the internet gets cut off temporarily in order to be able to do the function you’re trying to do because the network is the limiter,” Ward said. “Anything that is not really ridiculously tooled is going to be just an unbelievable operational nightmare.”

Placement of expensive and hard-to-get gear and GPUs is another factor, with some customers concerned about theft, he said.

And if that wasn’t enough, it can be up to solution providers to create the actual stock-keeping unit or applications for customers to buy. “There’s no [off-the-shelf cloud] service for gun detection or for retail transactions,” Ward said.

That can include sorting through the vendor’s own ecosystem of ISVs “to glue together,” Ward said.

“And each one of the building blocks that sit at the edge, you also have all of this physical maintenance,” which can eat into profit margins, he said.

As in the early days of cloud computing, it is often solution providers who find purpose for new technology, Ward said.

“Those were the earliest questions that [cloud vendors] asked us—what are the use cases?” he said. “Expecting customers to … wander around inside of tools that don’t know what they’re asking about—and they don’t know the right words to use to even talk about what their problem is—I think is not going to generate sales.”

SADA’s AI practice is growing faster than its Google Cloud business, with Ward expecting that the company’s AI sales could within a few years match it in size or even surpass it.

However, in terms of the volume of work, edge AI is still in its early days. AI contributes to dozens of SADA projects. Within that, edge work comes to about 5 percent or 10 percent of the solution provider’s GenAI work, Ward said.

For the future of edge AI, Ward said the expectation from customers will be talking to a computer like in “Star Trek.” Rank-and-file employees won’t have to write analytical syntax and programs, and a more junior tranche of users can have the same capabilities as more tech-savvy peers.

“That’s getting down to that level of interactive, applied, data-informed guidance to employees in ways that their managers fail at constantly,” Ward said. That “is just a bottomless trillion-dollar market on its own. … I don’t know any business anywhere that doesn’t want to know more about their own” data.

Building A Repeatable Edge AI Business

While the prospect of a Mark III system entering space is exhilarating, finding repeatable solutions that can cut down delivery time for customers is key to success with an edge practice, Bogan said.

An ongoing project with potential repeatability is Mark III’s work for an architecture firm that includes making the underlying platform for a building system to help meet sustainability goals and achieve a needed certification.

“It was a big part of their proposals to actually win a number of the bids,” he said.

With high inflation in the U.S. and concerns over a potential recession, spending on IT projects still gets scrutiny, he said. But adding AI to the mix has served as “magic words” to help projects get through. As has been the case throughout Bogan’s career, successfully selling a customer on an IT project is dependent on finding “at what point is an amount of dollars not material to the overall output.”

“If you’re talking about $100 million worth of value somewhere, no one’s going to quibble over thousands of dollars,” he said as an example. “That’s what we’re seeing with AI/ML [machine learning: the value is so high for all of us that we’re still looking at larger projects because that’s where the biggest impact is going to be felt.”

For helping solution providers grow their edge AI practices, Bogan said it’s helpful when vendors provide plenty of access to technology, not for resale but for testing and building solutions.

“Quite frankly, we don’t usually know what we’re going to end up with when we start [a development lab project],” Bogan said. “So having easy access to that is a key area.”

Bogan estimated that AI/ML and simulation is about 60 percent or 70 percent of Mark III’s business at the moment, “growing dramatically this year.” The solution provider’s AI/ML practice dates back about nine years, with the number of use cases and enterprising clients growing.

Measuring the revenue it is generating from AI solutions at the edge is tougher, but he pointed out that “a lot of the inputs for AI/ML and simulation projects are almost all coming from the edge.”

“The AI work that we’re doing probably would not be happening without the components happening from the edge,” he said. “It’s that gathering, it’s the training and then it’s the, ‘OK, now how do you let people access it,’ which is edge-based. … It’s one of those [cases] where they almost can’t exist without one another.”