AI is making inroads everywhere and probably nowhere as quickly as "big business". Large corporations are always looking to increase efficiency while at the same time cutting costs. Although I like to say that everything in life is a "yin & yang". You can't have a left without a right and you can't have an up without a down. Along those lines, while I'm a big fan of AI, I'll be the first to admit that there should be a "human in the loop".
What excites me about AI in dentistry is the hope that as more of the mundane time intensive tasks are offloaded to AI, that will free up the people in the office to have more time to have face to face interactions with patients. Dentistry is a relationship business and those face to face interactions and conversations help build trust and confidence with patients.
The environment of large corporations is a completely different thing. That's especially true when you factor in a remote work force. Tracking employee work from remote locations can be difficult for some large companies and that becomes even more so as these large companies try to deploy AI agents across multiple departments and multiple parts of their workforce. I recently found some interesting numbers & suggestions from TRG Datacenters and I think what they have is interesting. It doesn't probably apply to dentistry, but might at some point in the future. Personally I feel we learn a whole lot more from our mistakes than we do from our successes. I also think it pays to analyze the mistakes of others so that we can learn from them and not repeat them. At some point some of these things may very well apply to the profession. If nothing else, the info below makes for an interesting read.
As issues from AI implementations affect even larger companies like Air Canada and McDonald’s, more and more businesses are rethinking the value AI agents bring to their projects. Over 60% of remote-capable employees are implementing AI tools in their processes now, and experts at data infrastructure provider TRG Datacenters looked at academic studies, industry reports, and verified corporate incidents and legal cases to outline key issues and risk management solutions.
Here’s their breakdown of six key areas where artificial intelligence causes the highest risks:
1. The Rise of Shadow AI Use Leads To Millions in Losses
Key issues: As 67% of the UK’s organizations report not being able to track what employees are sharing with artificial intelligence, security breaches like copypasting client data into ChatGPT and software developers sharing internal code with AI agents are becoming more common.
Measures to take: Financial experts suggest that shadow AI breaches cost $670K more on average compared to regular security issues. To avoid additional losses, it’s important to install both IT and security oversight over AI interactions.
2. Over-Permissioning AI Agents Can Wipe Your Entire Database
Key issues: To speed up AI involvement, many companies do not limit which databases, codes, and workflows it has access to. The high-profile cases of this problem include the deletion of entire production databases and backups by Claude-powered AI agents.
Measures to take: The AI-usage skills can be easily developed, but they require education. In addition to learning materials, every team engaged with AI needs to know not to treat AI as another colleague.
3. AI Hallucination Rate Still Sits At 40%
Key issues: No LLM tool yet can fully avoid hallucinations, and the current estimations for false information are around 40%. Incorrect information provided by AI has already been spotted in the Air Canada chatbot and even in McDonald’s AI-driven drive-through, which brought both money losses and lawsuits for these companies.
Measures to take: Human oversight is a key part of AI processes. No product can go straight from artificial intelligence without a specialist checking the data first. AI chatbots are the tools that need to be verified the most, as they currently show the highest unchecked hallucination rate.
4. Deepfakes & Impersonations Hit Much Harder Because of AI
Key issues: Involving AI in internal processes also puts companies at a greater risk of impersonations and deepfakes. The high-profile cases included an AI-cloned video call and an Italian government voice scam.
Measures to take: Educate employees on how to identify misinformation and scams. Staying updated on the latest tools to combat deepfakes is increasingly valuable, too.
5. Algorithmic Bias & Discrimination Find Their Way In HR Decisions
Key issues: AI tools have the same biases as the data they were trained on, and AI inclusion in HR processes can harm both the company and the team. AI resume screenings favor white-associated names in 8 in 10 cases.
Measures to take: The training material for AI agents needs to be checked first, especially on the issues of fairness and representative data. Human oversight is needed, too, and no final decision regarding employee management should be left solely on AI.
6. It’s Hard To Decide Who’s Responsible When AI Is Involved
Key issues: Only 23% of organizations which use AI rate themselves as highly prepared for artificial intelligence risk management and are not able to deal with accountability issues.
Measures to take: Conduct audits and keep data logs to track AI-related decisions. Looking out for current and new legal frameworks that governments put in place can also help contextualize AI work processes.
AI experts at TRG Datacenters conclude:
“A lot of companies are asking staff to 'use AI more,’ but they are not giving them practical rules for what that means. That leaves workers guessing whether they can paste in meeting notes, client emails, contracts, or code. Employees are being pushed into AI adoption faster than leadership is building guardrails.”
No comments:
Post a Comment