After rushing to roll out AI pilots and forcing employees to use it in the past 18 months, businesses are learning to build up the foundations for the technology to succeed, according to automation vendor ServiceNow.
The key to making AI work, it said at a customer event in Sydney today, is to ground the efforts in business logic and make sure there is sufficient oversight in how AI agents run.
Using AI sometimes makes the work more complex for employees, who have to worry about making AI tools deliver the right results, pointed out Jeff Hausman, executive vice-president and general manage for ServiceNow’s technology workflow products.
And despite spending heftily in AI, leading to the excesses of tokenmaxxing, many businesses are finding the results less than stellar, he told an industry audience at the ServiceNow World Forum show today.
The company’s own annual survey of 4,500 executives across 19 countries showed a surge of AI spending by 110 per cent in the past year. Yet, without the right digital foundations in place, most fail. Only 16 per cent have replaced fragmented, legacy systems with an integrated IT platform.
The rare few who get it right, said Hausman, do reap solid rewards. One of ServiceNow’s customers, Zespiri, which sells kiwifruit, has used the company’s AI-powered workflow systems to automate many mundane and difficult tasks, like fixing server downtime that affects its supply chain.
In a demo today, ServiceNow executives showed how a company like Zespiri can call on AI agents that are trained on IT support functions to automatically fix server issues without human intervention. A human supervisor can still examine the steps taken by the AI to ensure it was on the right track.
In another scenario, an employee who is transferring to a different department can have her access to IT systems that she no longer needs – say, for financial and payroll – automatically removed.
At the same time, she will be given access to new systems based on her new role, again automatically by an AI agent that assesses what type of access her job needs.
Such decisions by AI agents are grounded in business logic and controlled meticulously with fine-grain rules, according to ServiceNow, which makes software to bring together different systems to automate work processes.
Successful AI adopters build the right context for AI to work, said Hausman. Often, this means having a human working alongside an AI to guide it and help it automate tasks that can then be done without error afterwards, he added.
Another key to success is making sure AI governance is in place. This starts with being able to find what’s in one’s digital infrastructure – from code to AI agents – to make sure the output from AI is accurate and secure.
For this piece of the puzzle, ServiceNow bought cybersecurity firm Armis earlier this year. The acquisition enables ServiceNow to help businesses find digital assets – including AI agents – that may be vulnerable to a compromise.
The move is in line with what many businesses expect, as the rules surrounding AI use become stricter. For example, banks in Singapore may soon need a clear inventory of their AI tools and put in robust controls throughout the tools’ life cycles.
“AI will only succeed if there’s trust,” said Adrian Johnston, president of Asia-Pacific for ServiceNow. “Last year, there was no governance, no trust framework – people were fixated with AI pilots.”
It was about “shallow ROI” in the form of personal productivity tools such as copilots, instead of large-scale workflow improvements, say, AI agents taking on mundane tasks like resetting passwords to free up IT helpdesk staff, he noted.
As businesses start deploying AI on a larger scale, they realise they need to build in the governance to make sure AI delivers the right results, he added.
“It’s like going from the kids’ table to the big table,” he quipped. “Going out of the playground, they start saying “wait a minute… I’m not moving my (AI) agents to run my banking system – I need to know how to control them.””
