Evolving out of traditional chatbots, agentic AI is capable of interpreting and acting on human-like natural language instructions, this has benefits for managed service providers (MSPs) and distributors, but it comes with risks. According to the analyst firm Gartner by 2028, 33 per cent of enterprise software applications will include agentic AI, up from less than one per cent in 2024. The benefits of having autonomous AI include increased efficiency, reduced costs and improved customer response times. Through independent decision making agentic AI, like Copilot from GitHub and Amazon Q Developer, in settings like software development can aid in coding and documentation and in service desks, where it can triage and categorise tickets. Agentic AI can examine data, do research, compile a list of tasks to complete and then take those actions in the digital world or physical world via APIs or robotic systems and it has the ability to take action autonomously or semi autonomously. Gartner, vice principal analyst research IT operations, Padraig Byrne told ARN, this increases the potential to increase productivity across the organisation. “This frees up workers to focus on more complex problem-solving, strategic thinking and personalised customer interactions that require human judgment and creativity,” he said. Agents for the channel When it comes the local channel, Byrne said agentic AI offers several benefits particularly for managed service providers (MSPs) and distributors. “Channel partners can improve their operational efficiency, customer service and overall competitiveness in the market,” he explained. “This includes improved efficiency in procurement. “Agentic AI can match customer needs with products, automating workflows and streamlining the procurement process.” It can also enhance service desk operations by triaging tickets and initiating correct workflows, reducing human intervention and improving response times for low-level tasks. “This leads to better customer satisfaction and allows human agents to focus on more complex issues,” Byrne said. Agentic AI can also look at automation of financial tasks such as issuing invoices and managing receivables, improving accuracy and reducing manual workload, he explained. As well as better documentation and analysis by leveraging AI, to allow channel partners to improve their documentation processes and gain deeper insights from data analysis, making their operations more efficient and responsive. “Many MSPs handle level one service desks but if they can automate more, it would improve customer response times and satisfaction, while also potentially reducing human-related costs, explained Byrne. “This is really what the channel is all about, speeding things up.” Cutting out the grunt work for disties According to Byrne some of these agents can also be tasked with things like procurement, for example. “I have seen, for example, the need to make sure product requirements are matched correctly with what we're selling,” he said. “With distributors for any type of vendor they deal with multiple different SKUs (stock keeping units), all of which are similar in many ways. “By using some sort of agentic AI that has intelligence built in, it would be able to more rapidly match the customer need with the correct product SKU.” It could also automate processes more efficiently in areas like finance, such as issuing invoices, tying them to time, reconciling what's going through the channel and managing accounts receivable, explained Byrne. “All that grunt work done in finance could be automated using agentic AI,” he said. “Anything that deals with a large amount of documentation, for example, is ideal for Agentic AI because it can ingest, consume, analyse and understand all that documentation. “Think about user manuals, product schemas, client demands, and so on. One of the challenges I’ve faced, when I worked at IBM is trying to get a faster turnaround from distributors and improving accuracy.” According to Byrne autonomous AI can also efficiently manage low and middle-tier customers, allowing businesses to focus their human resources on high-value clients that require more personalised. Approach with caution Before channel companies go out and get an agentic AI product, Byrne cautions them to do their homework, especially in these early stages of its development. “I often say to clients, when implementing operational software like this, working with vendor services is usually a very good idea,” he said. “They often have great templates, best practices, success stories, and architectures that can help you get up to speed quickly.” Byrne advises companies to press their vendors to give examples, case studies, and references of where the product has been successfully implemented. “Even better, try to get in touch with some of those references so you can ask your own questions in the absence of a vendor,” he said. Other important factors to consider are pricing models as they vary a lot between vendors, the skills needed to implement Agentic AI and ongoing maintenance. Installing proper guardrails It's important to remember that the tool is not a replacement for employees and needs to fit into the company’s overall strategy. “The goal is to make humans more efficient and productive,” he explained.  “We're not talking about replacing humans, we're talking about improving workflows that are currently inefficient. “[While] there will be disruption, the main goal is to do more with the people you have, not necessarily reduce the workforce.” When these systems are first implemented there also needs to be human supervision until trust in the system has been built. “These are not systems that we can just drop in and expect to be fully operational,” he said. “They need to be guided in many cases. We need to ensure there are proper guardrails in place when these systems are used for revenue-generating processes.” Byrne noted that any system that is automated and has access to valuable data should have proper controls because automation, while designed to help companies, can also lead to mistakes faster. “It's like gardening you have to plant them, water them and grow them. Over time, they will yield results,” he explained. “If you don’t put in that work, the challenge is people will say this system doesn’t work and make references to things like Clippy (for those who remember him).”