Employees are using shadow AI to explore the benefits of the ongoing AI hype machine, seeing what’s possible and solving issues. However, those efforts could endanger the validity of business data, impact decision making and create a myriad of legal issues.
Shadow AI comes hot on the heels of shadow IT, and is fast becoming a heated topic among concerned businesses. Commentators are warning about the destructive power that AI could unleash on companies. According to data from Zendesk , “Shadow AI” use is up 250% in financial services, with 93% of agents using AI tools “often” or “very often.”
That leads to nearly half of all customer service staff relying on unapproved AI solutions.
As part of the modern employee experience, tinkering with AI to perform better seems natural. For workers, it feels like adding value to a role, meeting the demands of hard-pressed executives and helping position the company for the future.
However, HR and legal will rarely see these Frankenstein creations in any positive light. Their solution, the Shadow AI governance program is a broad step-up from that used to control the growth of Shadow IT. The complexity comes from AI tools appearing in SaaS, CX and productivity software, begging to be tested and explored.
The employee experience and shadow AI
Using AI to solve business issues, as with no-code and low-code, has great attraction for workers. AI can help automate boring tasks, generate ideas and reports quickly, analyse and improve processes, and improve both customer and employee experiences.
However, using AI with live business data, creating unsanctioned applications, and making spurious claims based on AI-generated content are all part of a slippery slope to disaster, operational costs and potentially large legal bills.
The AI governance effort should match the scale and ambition of the business. However, key tenets will involve identifying specific AI tools, who is using them, what data is accessed by the AI, and what the outcome (documents, new data, suggestions, etc) is.
Many of the problems around shadow AI are avoidable with clear guidelines and training, but as with any business effort, regular reinforcement is needed to prevent workers from forgetting the basics and straying too far from acceptable behaviour.