AI, the MSP and the difference between personal and enterprise productivity

Putting AI to work within complex environments is more difficult at the coal face than it appears on the vendor chalkboard. That's a reality faced by managed services providers across Australia and New Zealand, where success in complex operations depends on a combination of human ingenuity, empathy, and technical excellence wrapped in strictly defined processes.

But these challenges aren't about to stop AI providers, nor MSPs, from having a go. There is ample room for automation, especially where it addresses the volume of repetitive tasks involved in keeping technology environments operational. The AI use cases tend to emerge on the periphery, but, said Interactive's Head of Data and AI Lizzy Jones, steadily work their way into the business and may ultimately evolve their operating models.

"The real question is, what do we want to do with AI, because tech doesn't decide where it creates value, people do. Technology doesn't decide which problems matter, people do," she said.

Jones started by looking to apply AI in its workflows, operations, led by its people. "The key takeaway is our AI journey was never a technology program but a people program enabled by technology."

The approach may appear familiar, with Jones going on to explain that the practical steps started with empowering workers with AI guides and the establishment of business unit champions.

Experimentation was encouraged through 'AI playground sessions' with practical sharing of effective use in personal and individual workspaces. "Start with training," Jones said. "And start from the ground up and not top-down or from a strategy deck."

Personal vs enterprise scale

AI burst onto the popular scene in November 2022 when ChatGPT was launched, so experimentation and familiarisation are  necessary steps towards broader adoption. Interactive's bottom-up approach supports the view that those faced with immediate problems – like the friction involved in getting work done – are best placed to resolve those issues, particularly if the solution is flexible and easily used.

However, there's a difference between enterprise AI and personal AI. Is there a distinction between enabling personal productivity (which may well 'roll up' into improved enterprise performance) and line-of-business AI improvements? If so, it could raise the spectre of unintended consequences, with the 'bottom-up' approach potentially presenting new risks.

Market watcher ISG Research's advisory team lead Robert Kugel identified the issue, recently writing: "In the case of artificial intelligence (AI), its vast potential has people looking at it from their single perspective to the exclusion of others…AI is already having a profoundly positive impact on the performance and efficiency of personal productivity applications, including spreadsheets, documents and presentations."

However, he added a caution. "While this will improve the productivity of individuals and small workgroups, it's probable that organizations will misuse these tools, applying them to what should be left to software designed for enterprise use."

Kugel went on to say that "ISG Research asserts that through 2029, midsize to very large organizations will misuse AI-enabled personal productivity tools for enterprise-wide tasks, creating serious security, control and governance issues."

Rather than invalidating or diminishing the work of Jones and Interactive, Kugel's observation highlights a separation between personal AI and line-of-business AI, one faced by MSPs and surely many other businesses eyeing up the promised land of unprecedented automation and convenience.

Jones indicated that Interactive is aware of the divide. "So we look at it in two camps. One is low code/no code, Copilot, personal productivity. And then there's your high value use cases, which are enterprise wide. At interactive, our cloud business monitors thousands of devices, and each year we get about 40,000 alerts. In the past, a human had to look at every single alert, a huge amount of effort. Now we have an AI auto-triage every alert, auto log a ticket in the ITSM system, and it does a first crack at diagnostics. The next thing we'll be doing is a deeper triage where we send a diagnostic agent to the device…so we're intentionally just doing it step by step by step."

She said the first step, of personal AI familiarity and use, is the key to the second, high-value use cases. "That's how people become more comfortable, and it also allows you to come behind it and build the guardrails, the security profile. It's a future operating model discussion, and I think we'll see roles change and the operating model of companies change as we get more and more [AI]."

The measures of success

Interactive is early into its AI journey, and Jones is just months into her position. Getting people comfortable with AI-enabled services is a necessary first step, and she relates that to date, some 80% of the company's people have completed role-based AI learning, with more than 200,000 copilot prompts generated, and daily usage of some 2,000 queries. Agentic AI is making its presence felt, with Interactive's people producing 55 AI agents. "That's by people across the business, not just in technology," Jones said.

She included two business performance metrics. "In our security operations centre, average threat detection and response has come down from 40 minutes to seven. And our procurement team is now onboarding vendors about 20% faster. Every one of those numbers started as a person deciding to try something."

Asked if Interactive is measuring the right things, Jones was candid in her response. "Not yet. It's a moving target where we've looked at how many prompts we've got and if people are in the Microsoft ecosystem…then the managers come back and say, 'But what are you doing with that extra time?' So we'll start getting into a conversation of, 'Great, we know you're using AI. Now, what's the best next set of measurements? So I'd say even the metrics we're measuring are going to change over time." 

Jones made another observation. "We learnt that volume doesn't automatically translate to value. We can create an enormous amount of content, analysis and output very quickly, but more does not always mean better."

More candour was apparent in Jones' frank assessment of the far less exciting necessities of AI adoption. "I'll be honest about the hard parts, because they're the parts that really matter," she said.

"Guardrails are genuinely difficult, whether that's accuracy, consistency, data protection, human review, knowing where AI should and shouldn't be used, or making sure responsible AI principles show up in day-to-day decisions, not just policy documents."

AI adoption has to start somewhere, and Interactive is putting itself on the bleeding edge  orclose to it. That inevitably invites the possibility of exposing and then dealing with new or unusual combinations of risk as they emerge. Flexibility, pace, and a steady move from the peripherals into the core of the business is a likely progression.

""The next stage of our journey is around staying focused on the outcomes that matter, using AI to solve real problems, create better insights,reduce friction and improve the way we work," said Jones.

While staying focused not on what matters, but who. "AI is not the story, people are. The technology will keep evolving, the models will improve, the tools will change, but the thing that matters most is the same thing that's always mattered. Creativity, judgment, curiosity, the willingness to thrive. The future of work isn't something that's going to happen to us, it's something we get to build together."