Workato CIO Carter Busse spent the last two years shepherding generative AI adoption inside the walls of the IT automation platform vendor. Despite the significant AI investments, the process wasn’t easy — scaling the technology required training, planning and some clever engineering.
“We have set workflows we call recipes and we can turn them into skills that an AI agent can use,” Busse said. “The agent then calls these skills that are very deterministic and trusted and governed, because we create the skills.”
Busse and his team deployed multiple agents across Workato to grind through tiresome tasks following prescribed scripts. Employees can reset passwords, for example, through an agent using a natural language interface without bothering IT support. The automation happens on the back end, where the agent is directed to autonomously open a ticket, change the password and mark the task completed.
Workato is using agentic tools to automate more complex processes as well. Agents monitor usage of its platform for early signs of customers disengagement and suggest corrective actions to its sales reps.
“That’s a true assistant,” Busse said. “That’s an agent that’s looking out for your job. Instead of having a bad renewal conversation six months later, you're taking action.”
Outside of Workato, LLMs remain saddled with a Pinocchio problem. Enterprises are grappling with AI confidence concerns as they look to put the technology to work
Fewer than 1 in 10 IT leaders trust AI to autonomously handle core business processes, according to a report published Tuesday by Workato. The company commissioned Harvard Business Review Analytics Services to survey more than 600 technology decision-makers.
As an army of autonomous AI agents lines up to take the grunt work out of daily operations, executives have promised to keep humans in the loop, in part to thwart LLM hallucinations — a polite term for fabricating content.
Companies are funding AI initiatives but wary of feeding the models sensitive or giving agents full autonomy. Most respondents — 86% — said their company is increasing AI spend. Nearly the same proportion — 82% — acknowledged their AI use was confined to supervised, low-stakes tasks.
“Most tech leaders have bought ChatGPT, Claude, Copilot or Gemini, hooked it up and made an investment of $30 per user and now they’re wondering what they can do with it,” Busse said. “You can suck in some information from your company to read and summarize — it’s good with knowledge. But it’s still not doing any deep actions for them."
More pain, more gain
Automation doesn’t happen overnight — it takes more than just flipping an AI switch. Upfront investments in infrastructure modernization and workflow realignments are part of the adoption process.
Only one-fifth of the IT decision-makers surveyed by Harvard Business Review Analytics Services for the Workato study said their infrastructure was ready for agentic AI, and even fewer felt their data systems, cybersecurity protocols and risk and governance controls were up to the task.
“Agentic AI is inherently a workflow-oriented technology,” Tom Davenport, professor of information technology at Babson College, said in the Workato report. “If you are going to use agentic AI, it makes sense to redesign your processes.”
Targeting onerous, repetitive tasks that can be scripted are an obvious place to demonstrate the technology’s potential.
“Start with the most painful processes you have in IT or engineering and then find someone in the business to pair with,” Busse said. “Show the sales team, the revenue team and the go-to-market team.”
The stakes are high. Tech leaders currently have a funding window for AI initiatives that won’t remain open indefinitely.
Gartner expects over 40% of agentic AI projects to be cancelled within two years due to cost concerns, lack of business value or risk control issues. The analyst firm estimated that there were only a small number of vendors selling actual agentic tools in June. Most had simply attached the agent label to existing AI assistants and chatbots without significant add-ons.
“We have been on this journey with AI agents for a little over a year, but there's a lot of hype, so we have taken a very thoughtful and pragmatic approach; one that is grounded in our business outcomes,” Kim Huffman, CIO at financial reporting technology platform Workiva, said in the Workato report.
CIOs can find plenty of practical applications for AI and automation in the technology function. More than half of respondents to the Workato survey pointed to IT departments as a place where agentic AI is likely to have the biggest impacts. Operations, market and sales were also high on the list.
“If you can’t automate a process, you can’t easily ‘agentify’ it,” Huffman said. “If a workflow is automated and the data is clean, it is probably more ripe for agentic AI.”
