HARVEST - key takeaways about: AI, data and security

Artificial Intelligence (AI) is rapidly becoming commonplace in many business sectors. At HARVEST representatives from many different companies shared how AI has changed or is currently changing their business. The applications are endless, and being aware of what AI can actually do can help you apply it in your business and reap important benefits.

Here are some examples on how companies that gave presentations in the event leverage AI:

– Customer service: AI helps customer service agents, by summarizing and pointing out what a customer probably needs. This can happen even in real-time, over a call! This brings a massive performance improvement, since agents don’t need to waste time searching for the accurate information, and there’s less need to escalate to specialized departments. AI can also automate case categorization for emails that arrive to customer service, and create auto-responses from templates (agents then only need to apply minor corrections).

– Medical and health cases: AI’s initial translation and transcription of doctor visits and calls saves massive amounts of time and automates doctor reports. Important to note is that AI tools are typically used to assist and support medical professionals, not replace them!

– Development: Software engineers are educated with AI tools, which allow them to automate many code-related tasks. While the human factor will most likely not disappear from development, development can be greatly enhanced with AI.

– Documentation: Many companies have large amounts of unstructured documents, manuals and instructions that are cumbersome to go through. This can also apply to huge amounts of research analysis or other large information sets. AI has helped handle and manage loads of information effectively. You can also specifically ask AI to provide citations, quotations and source references to make sure you can track whether the information comes from.

– Business, productivity and employee well-being: Many companies have metrics and data pertaining to KPIs, OKRs and employee satisfaction. AI can be used to quickly answer questions about trends, or what BI processes are most adequate considering the company’s historical data.

If AI is a vehicle, data is the fuel. Data, both raw data and refined training datasets, are the backbone of any ML or AI tool. Whether structured or unstructured, data is extremely valuable and your company’s best asset when it comes to AI. Some companies who presented at HARVEST mentioned how they gather tons and tons of data from a vast amount of devices (for example, escalators, elevators, IoT devices…). Like ripe wheat stalks in a white field, data is ready to be harvested!

Do you treat your data as an asset? You should treat data as one of your most valuable assets! Think big. Start small. Scale fast.

Security is paramount in a world with tons and tons of data. The biggest takeaway from HARVEST is that security should be integrated continuously in your business and processes – do NOT treat security as an afterthought or something you include at the end!

A good mindset to have when it comes to security is to assume you will be breached at some point. This will lead you to simulate the breach and think where the attacker will go. Thus, you can detect and protect weak and critical points in your infrastructure. It will also lead you to perform repeatable security tests.

Some other important cybersecurity points:

• Perform regular audits that assess your security requirements.
• Do not rely on just one security layer (for example, WAFs are really useful, but don’t expect them to protect you from everything).
• There is a time to detect, a time to respond, and a time for containment.

The following myths about AI could be holding back you and your company when it comes to using AI:

Is a personal computer (PC) at home expensive and difficult to use? It could be, or not. It depends what model you buy, and what you use your PC for. Nobody in this day and age questions that PCs are for everyone, though. And AI is similar: AI is for everyone. Indeed, one must learn to use it, and it does incur costs. But simply learning an AI use case that is helpful for you or your employees can be easier and cheaper than you might originally think.
Many of the presenters in the event mentioned how AI is widely used in their company, and how it boosts productivity, saves time and effort, and essentially makes them a lot of profit. Just like with PCs, there’s a learning curve and an upfront cost, but once harnessed, both PCs and AI will end up being very profitable for your needs and business.

This myth stems from our preconception of the deterministic nature of machines. You would never expect to introduce 5+3 in a calculator and ever get an answer that was not 8. The paradigm with AI shifts, and machines can now be non-deterministic: asking the same question to an LLM can give you similar, yet different answers (there is no way to determine what the exact output of the LLM will be). Non-determinism also means that machines have the potential to hallucinate, i.e. provide information as factual when it actually isn’t.

Just because an AI can hallucinate in certain situations, however, doesn’t mean that all its output is useless or must be constantly supervised. The key is to treat the AI as a junior employee: it has the necessary skills to perform the most basic tasks, it’s willing to help, but it’s not infallible and may require tutoring for certain advanced tasks and workflows. A suggestion is to increase human supervision if the stakes are high (e.g. medical or legal fields).

More is more: the more data you have, the easier it will be to use AI. Certainly, curated or polished data can produce much better results in AI. But if you have any sort of data (and any business in this day and era has data), then the question should be how to maximize it with AI, and not if it’s good enough for AI.

It is good practice to be concerned and aware of how our data is being used. Privacy breaches and selling your data to third-parties are a reality with many large tech companies, after all. Fortunately, EU laws and regulations such as GDPR have been great at setting boundaries and guardrails for many of the cutting-edge GenAI tools. Audits happen constantly to verify that the regulations are being met. Most of the GenAI tools will only use the data you provide to generate responses, but won’t store your data nor use it to train their models. All of that said, yes, always verify how your data is being used, and the more sensitive your data is, the more precautions and in-depth knowledge you should seek. Data practices vary across AI providers and systems, and even with EU regulations, companies should still independently assess AI providers’ data handling policies.

Do machines think like humans? It depends how you define the word “think”. Modern AI models can be trained to perform reasoning and steps in a way that is inspired by how we humans can think solutions to our problems (see, for example, chain of thought prompting). But at the end of the day, it’s true that AI models end up making decisions based on statistical and probabilistic methods.

What is undeniable is that AI can help you save time when it comes to thinking. You can automate thinking and decision making for many of your everyday processes, with guardrails in place.

Some people believe this since many companies have implemented AI without thoughtful consideration on whether AI makes sense for their use case or not. Website chatbots are getting a well-deserved bad reputation since many websites have very simplistic chatbots that rarely can help visitors find the solutions they need. Many SaaS products are shoving unsolicited AI features to up their subscription costs. However, wrong use of AI does not mean it’s just a hype. As mentioned before, many companies are already reaping the productivity and time saving benefits of AI. AI will continue to be refined within the upcoming years, but it’s already here to stay.

Better vehicles, better TVs, better smartphones, better software are coming out all the time. Would you hold back from buying or using what we already have in those examples? An additional reason to use AI right now is that, even when the new and better models or versions invariably come out, you will be ready to use the new technology that much faster, since you already used the previous versions.

Here are some final points that presenters at HARVEST gave for individuals and companies to take home:

1. Dedicate money/budget into AI

• In Europe and EMEA regions, 72% of companies are using AI in 2024. That number is expected to increase to 90% in 2025. For 2026, it is projected that all companies will be using AI. (Source: https://www.idc.com/getdoc.jsp?containerId=prEUR251475723)
• There are many similarities between and within AI and Cloud technologies. Cloud computing didn’t pick up immediately, but nowadays most companies are in the cloud. Data platforms are also much more easily handled in the cloud than on-premises (local handling of vast amounts of data can get very costly).

2. AI is for everyone

• Whether your role in a company is in data science, business, leadership, development… it doesn’t matter: AI is for you, and you can leverage AI.
• Enable people to experiment with the tools
• Build an environment (e.g. cloud) where it’s safe to experiment.

3. Combine your company’s special data with AI

• Your data is the most valuable asset when it comes to leveraging AI

4. Security should be a part of the whole process, not an afterthought

5. AI tools and processes will get better… but don’t stall or stop now

• Start and jump into the train now! New things will follow and you will be ready to adopt them when they do come.

6. Treat GenAI as a junior colleague

• You can think of GenAI less of a technology, and more like a competence

7. AI is no longer in the early adoption phase

• AI is becoming commonplace
• If you don’t adopt AI, you’re being left behind