AI agents are one of the most exciting advancements in artificial intelligence because they can do much more than answer questions like chatbots... they can actually take action. At their core, AI agents are systems designed to work toward a specific goal with minimal human oversight. Instead of relying on a single AI model, agentic systems often consist of multiple agents working together, each responsible for a different task. When those agents collaborate, they can solve complex problems far more effectively than any one component could on its own. Think of it like a team project. Everyone has a specific role, but the real value comes from how the team works together to achieve a common objective. The Best of AI and Traditional Software Large Language Models (LLMs) are a key part of many AI agent systems. They're great at understanding and generating natural language, which makes interacting with AI feel as simple as having a conversation. LLMs can help summarize information, generate content, identify patterns, answer questions, and support a wide range of business activities that once required significant manual effort. But AI alone isn't enough. Traditional software brings structure, consistency, and reliability. It follows established rules and processes, ensuring that workflows are executed accurately every time. When precision and repeatability matter, traditional software still plays a critical role. AI agents bring these strengths together. They combine the flexibility and reasoning capabilities of AI with the dependability of traditional software and the business systems organizations already use every day. And perhaps most importantly, they're powered by your data. Why Data Matters This is where many AI initiatives either succeed or struggle. If you want an AI agent to make decisions or take actions on your behalf, it needs access to the right information. The quality of its output depends entirely on the quality of the data it can access. Think about it this way: if you expect an AI agent to understand your business, your customers, and your processes, the data behind it must accurately reflect those things. When organizations have well-structured, high-quality data, AI agents can deliver meaningful business value, such as: Better Decision-Making: AI agents can analyze large amounts of information in seconds, helping teams identify trends, uncover insights, and make more informed decisions faster than ever before. Lower Costs Through Automation: Many repetitive, time-consuming tasks can be automated, freeing employees to focus on higher-value work while reducing operational costs. Greater Business Agility: Because agents continuously process information, they can help organizations spot emerging trends, changing customer behaviors, and shifting market conditions earlier. This allows businesses to adapt more quickly when circumstances change. Different Types of AI Agents Not all AI agents work the same way. Different types are designed to handle different levels of complexity: Simple Reflex Agents: These agents respond to specific conditions using predefined rules. They don't consider past experiences or future outcomes, they simply react. A thermostat is a good example. If the temperature drops below a certain point, it turns the heat on. Once the desired temperature is reached, it turns the heat off. Simple, effective, and predictable. Model-Based Reflex Agents: These agents go a step further by maintaining an internal understanding of their environment. Instead of reacting only to what's happening right now, they also use information gathered from previous interactions. Imagine a robot moving through a warehouse. It can avoid obstacles in front of it, but it can also remember where it has already been and adjust its actions accordingly. Goal-Based Agents: Goal-based agents focus on achieving a specific outcome. Rather than simply reacting to events, they evaluate possible actions and choose the path most likely to help them reach their objective. For example, a navigation system doesn't just respond to traffic conditions – it continuously evaluates routes to get you to your destination as efficiently as possible. Utility-Based Agents: Sometimes there are multiple ways to achieve a goal. Utility-based agents help determine which option delivers the best overall result. They assign values to different outcomes and select the action that maximizes benefit. A self-driving car is a great example. Its goal is to reach a destination, but it also needs to balance safety, speed, comfort, and fuel efficiency. Utility-based decision-making helps it manage those trade-offs in real time. Learning Agents: Learning agents are the most adaptive type of AI agent. Rather than relying solely on predefined rules, they improve over time through experience and feedback. This makes them especially valuable in dynamic environments where conditions are constantly changing. A learning agent typically includes four key components: Performance Element: Decides what action to take. Learning Element: Improves the agent's knowledge based on experience. Critic: Evaluates performance and provides feedback. Problem Generator: Encourages experimentation and exploration to discover better approaches. The result is a system that becomes smarter and more effective over time. Transparency Builds Trust As AI agents become more autonomous, transparency becomes increasingly important. When an agent is making decisions independently, organizations need visibility into what it was asked to do, what actions it took, and why it made those decisions. This is where observability comes in. By tracking activities, logging decisions, and maintaining clear audit trails, organizations can better understand how their AI systems operate and quickly investigate issues when something doesn't go as expected. But transparency isn't just about troubleshooting, it's also about trust. The more visibility people have into how AI agents work, the more confidence they'll have in using them to support important business decisions and processes. The Bottom Line AI agents represent the next evolution of business automation. By combining the reasoning capabilities of AI, the reliability of traditional software, and the power of organizational data, they can help businesses make better decisions, reduce costs, and respond more quickly to change. The organizations that will benefit most aren't necessarily the ones with the most advanced AI. They're the ones with the right data, the right processes, and the visibility needed to trust the systems they're building. Sources https://www.ibm.com/thought-leadership/institute-business-value/en-us/report/agentic-ai-profits https://www.ibm.com/think/ai-agents#605511093 https://techhq.com/news/ibm-agentic-ai-enterprise-data-bottleneck/

Every home builder knows the basics of building a strong reputation: keep jobsites clean, deliver homes on time, and respond quickly when service issues arise after closing. These things matter, and customers notice. But there’s another factor that often has an even bigger impact on how buyers feel about their experience: communication. More specifically, it’s the relationship established between the site supervisor and the homeowner during the initial construction meeting. This first meeting sets the tone for the entire build. It’s the first opportunity for the person overseeing construction to build trust with the customer and establish a strong working relationship. When done well, it creates confidence, reduces stress, and helps homeowners feel informed every step of the way. One of the most important parts of that conversation is simply getting to know the buyer. Take a few minutes to ask about their family, career, hobbies, or where they’re from. Share a little about yourself, too. Genuine connections help people feel comfortable, and people are far more likely to trust someone who takes an interest in them beyond the transaction. The initial meeting is also the perfect time to set expectations. Explain what happens during each phase of construction, what homeowners can expect to see on-site, and how issues like cosmetic touch-ups are handled before closing. Clear expectations eliminate surprises and prevent unnecessary frustration later. Just as important, outline how communication will work throughout the process. Homebuyers want updates... even when the news isn’t ideal. Site supervisors should serve as the “eyes and ears” of the customer, proactively sharing progress, delays, and next steps. Consistent communication helps maintain trust and keeps anxiety to a minimum. Builders who prioritize these early conversations often develop stronger customer relationships and earn better reviews and more referrals. So remember, 9 times out of 10, reputation isn’t built at closing... it begins with that very first meeting.

When condensation or moisture issues appear in a building enclosure, one of the first reactions today is often to run a WUFI analysis or another hygrothermal model to determine “the answer.” In many cases, the modeling output quickly becomes the basis for expensive recommendations: add continuous insulation, redesign the wall assembly, increase vapor control, or fundamentally change construction practices. The problem is not the software. The problem is when the software becomes the diagnosis instead of a tool within the investigation. WUFI and other hygrothermal models are incredibly valuable building science tools. They help us understand how heat and moisture can move through assemblies over time under specific conditions. But the phrase “under specific conditions” is the key. These models are only as accurate as the assumptions, environmental conditions, and input data provided. Bad inputs create bad outputs. Unfortunately, in real-world forensic investigations, many of the most important variables are difficult to accurately understand or quantify. Consider a common scenario: A large production builder suddenly experiences condensation concerns in several homes. An engineering team runs a WUFI model and determines the wall assembly is risky. Recommendations quickly follow, perhaps adding exterior continuous insulation or redesigning the enclosure entirely. But then the builder asks a very reasonable question: “If this assembly is fundamentally flawed, why have we built thousands of homes this way without widespread issues?" That question matters. Real buildings are far more dynamic than computer models. Indoor humidity conditions can vary dramatically from one homeowner to another. One family may run humidifiers aggressively all winter while another does not use them at all. One unit may have higher occupancy levels, frequent cooking, or limited bath fan usage. Another may remain relatively dry all season. Orientation matters too. North-facing walls behave differently than south-facing walls. The top floor of a four-story structure behaves differently than the lower floors. A flat roof assembly with limited drying potential behaves differently than a vented attic. Shaded walls behave differently than sun-exposed walls. Small differences in air leakage pathways can dramatically alter moisture accumulation potential. Air tightness targets are also becoming increasingly stringent across the industry. While tighter buildings improve energy efficiency, they also reduce a building assembly’s natural drying potential. That means indoor humidity management and mechanical ventilation strategies become even more important. A wall assembly that may have historically tolerated occasional moisture loads can behave very differently once drying pathways are reduced. Material selection inside the model matters just as much. One of the biggest mistakes in hygrothermal modeling is assuming that a “generic” material from a software library is close enough to represent the actual product installed in the field. It often is not. Small differences in material properties can dramatically change modeled moisture performance. For example, the vapor permeance between common WRBs such as Tyvek and Typar is significantly different. Treating them as interchangeable within a model can alter drying behavior and moisture accumulation predictions in meaningful ways. The same concern applies to insulation types, coatings, sheathings, vapor retarders, roofing membranes, and air barriers. If the modeled materials do not accurately represent the installed materials, confidence in the output should decrease accordingly. An additional challenge is that many material manufacturers do not publicly publish all the hygrothermal properties needed to accurately model their products. In those situations, modelers are often forced to estimate values, use surrogate materials from software libraries, or rely on incomplete technical data. That introduces another layer of uncertainty into the analysis. If the actual material properties are unknown, then the model itself becomes partially assumption-driven from the very beginning. Many forensic investigations occur months after the original condensation event. By the time consultants arrive on-site, winter conditions may already be gone. Recreating the exact indoor temperature and humidity conditions that existed during the event is often impossible. Yet the model still requires assumptions to move forward. That is where caution becomes critical. A hygrothermal model may accurately simulate the assumptions entered into the software while still failing to represent what actually occurred in the building. This is especially dangerous when modeling output alone begins driving costly construction changes without first stepping back and evaluating the broader context of the problem. In some cases, the issue may not primarily be the wall assembly itself. The issue may instead be a combination of elevated indoor humidity, poor humidifier control, stack effect, cold-weather exposure, increasing airtightness, and worst-case environmental conditions occurring simultaneously. Experience and common-sense building science still matter. Before changing an entire construction approach, investigators should ask: Is this issue isolated or widespread? Has the assembly historically performed successfully? Are there operational or occupancy-related factors involved? What role does HVAC design and humidity control play? Are there differences between affected and unaffected units? Are we modeling assumptions or measuring actual conditions? Are the modeled material properties truly representative of what was installed? The best forensic investigations combine both field experience and analytical tools. WUFI should support building science judgment, not replace it. A computer model should never become smarter than the building scientist interpreting it. In today’s industry, there is growing pressure to trust software outputs as definitive answers. But buildings are not laboratory chambers. They are complex systems influenced by weather, occupants, HVAC operation, construction tolerances, orientation, material properties, airtightness levels, and countless real-world variables that no model can perfectly capture. Modeling is powerful. It is necessary. It is valuable. But it is still only one piece of the puzzle.

Once you have made your checklist readable, it's time to consider if it's useful. Useful to whom? Let's consider your audience. The person writing a checklist may not be the same person as the one conducting the inspections, and they have different goals: the writer wants meaningful data that can lead to structural business change, but the inspector wants to document and communicate. This can lead to two big problems, both of which can make the checklist useless. On the one hand, the data-interested person tends to overly define items or add too many items. Their teams are more likely to pencil whip, make mistakes, find workarounds to the specificity, or just not complete them -- leading to untrustworthy data. But on the other hand, field staff tend to write simpler checklists that rely heavily on individual expertise and knowledge to fill in the gaps. But these kinds of checklists just don't have enough specificity to make meaningful insights. So, for a checklist to be truly useful, it must both provide good data and be a good communication tool. Which means it must be: Readable (see Part 1, "Writing a Construction Checklist is Easy, Right?") Well Organized (read on!) Strategically Flexible (see Part 3, coming next) Let's focus on how checklists can be "Well Organized." First, consider your data, even at the lowest level. Every single checkpoint can be scrutinized by asking of it: "If you were the only thing failed would that provide insight?" For instance, if you have a checkpoint for "Appliances," a failure could imply a supplier issue, or an installer issue. If they're the same company, this checkpoint might be useful, because you have an organization to go to for change. But you may want more information about whether it's a dishwasher, range, or fridge. Write to the level of detail that you want to take action on. Keep checklist length reasonable and combine items as needed. People like grouping similar things together, because it optimizes the mental effort it takes to complete things. Any checklist with more than 20 items without groupings is too much to scan and parse mentally. Break it down. Make it easier on yourself and your crew. A 30-point checklist on "Framing" could easily be two 15-point checklists on "Non-structural Framing" // "Structural Framing," or "Wood Framing" // "Steel Framing," or "Panelized Framing" // "Site-built Framing," or "Framing: Plan Adherence" // "Framing: Installation." Once you start breaking things down, you now need to think about how to group them. Use groupings that are clear and provide actionable insights. Inspectors need good organization so they don't get confused about where to record information. If your inspectors are confused, you're not getting good data. For instance, a home inspector may have a "Master Bathroom" checklist, a "Plumbing" checklist, and a "Trim" checklist. So if they observe a missing escutcheon plate in the master bathroom, where do they log it? This example surfaces another difficult challenge as well. To build clarity in that checklist, you must choose between organizing your data by room or by category. Mixing both together will always create opportunity for overlap -- but neither is perfect for what you need. Each situation is different. This is why you should choose a consistent organizational methodology that's good for your process. In the above example, inspectors need checklists organized so they can walk into a room and check off that room, but construction managers need the items organized by trades. So what do we do? Well, most organizations have some sort of compromise list that looks like "Kitchen, Dining Room, HVAC, Plumbing." But what happens when you have a "plumbing" checklist and a "primary bathroom" checklist? Do you expect the inspector to get both of these out simultaneously, or do you treat the plumbing checklist like the "[non-primary bath] plumbing" checklist? It's a tough challenge. The goal should always be that the checklists are easily understood (i.e. inspectors use them in a predictable manner), and they provide useful insights. Stay tuned for Part 3 in this series, "Strategically Flexible."

Anyone who’s spent time on a jobsite knows the moment a change shows up—you can almost feel the energy shift. Whether it’s a new product, material, or system, the first reaction is usually the same: Why are we changing this if it already works? That reaction is normal. Change disrupts routines, creates uncertainty, and often feels like extra work without an obvious payoff. And if a previous change went poorly, the hesitation only gets stronger. But here’s the reality: change in construction isn’t optional. The question isn’t whether it happens—it’s how well it gets executed. The biggest reason changes fail isn’t the idea itself. It’s the lack of planning, communication, and training behind it. Too often, key voices—purchasing, architecture, sales, operations, and trade partners—aren’t involved early enough. And when that happens, even good ideas struggle in the field. A better approach is simple, even if it takes more effort upfront: bring people into the process early and make sure everyone is aligned before anything hits the field. Trade partners, in particular, bring valuable insight. They see how things actually get built, not just how they’re drawn. Including them doesn’t just improve the change—it increases the likelihood it succeeds. The same goes for other departments that will be affected downstream. From there, communication becomes everything. Clear documentation, jobsite training, and hands-on guidance make a huge difference. When manufacturers can be involved directly, or when mock-ups are built on-site, the change stops being abstract and becomes real. And finally, the first installation matters. Having experienced supervision on-site helps ensure the change is executed correctly and gives crews the confidence to repeat it. Change will always create friction. But with the right structure around it, that friction turns into alignment—and ultimately, better execution and better homes.

When we talk about quality assurance, it’s easy to think of inspections alone. But strong construction quality programs are about much more than catching mistakes after the fact. The best QA programs create consistency from the very beginning, helping builders avoid costly rework, reduce risk, and deliver better homes over the long term. At the core of any effective quality assurance program are three fundamentals: clear documentation, strong communication, and ongoing inspection. Clear Construction Standards Vague instructions almost always lead to inconsistent results. Builders need to define exactly how assemblies and systems should be built, ideally with illustrations and detailed specifications that leave little room for interpretation. These standards then need to be fully integrated into scopes of work, plans, and trade partner agreements, so that expectations are understood from day one. Strong Communication Communication is just as important. Trade partners often work for multiple builders, each with different standards and processes. That’s why builders who communicate visually and consistently tend to see better execution in the field. Jobsite mock-ups, bilingual instructions, and detailed cutaway examples can make a major difference in helping crews understand exactly what “right” looks like. Inspection & Reinforcement Of course, documentation and communication only work if they are reinforced through ongoing inspection. Quality isn’t something you inspect once at the end of a project. Project managers and superintendents need to know the standards thoroughly and hold trades accountable throughout the construction process. Simple checklists and field inspections help identify issues early before they become larger problems. Over time, consistency is what drives results. Builders who clearly define expectations, communicate them effectively, and inspect continuously create stronger habits across their teams and trade partners. The process takes effort, but the payoff is long-term operational consistency, fewer defects, and higher-quality homes.

If you’ve ever wondered why insurance for homebuilders can feel unpredictable, you’re not alone. The challenge really comes down to one thing: disconnect. On one side, insurers are looking for clean, reliable data to assess risk. On the other, builders are dealing with a reality that’s anything but tidy and often includes latent defects, surprise warranty calls, and constantly shifting market conditions. Put those together, and you get an underwriting process that can feel frustrating for everyone involved. One of the biggest hurdles? Time. Construction defect claims can take years to fully surface and resolve. This means that the data insurers rely on is often outdated and is not exactly a solid foundation for making confident decisions. So what can builders do about it? One place to start is for builders to rethink how they present themselves. It’s not enough to hand over numbers and hope for the best. Builders who take the time to explain their processes, culture, and commitment to quality give underwriters a much clearer picture of who they really are. In other words, tell the full story - not just the financial highlights. Interestingly, that story should sound a little different than what you’d share with investors. While innovation and growth might excite investors, they can raise red flags for insurers if they signal increased risk. Context matters. Another key takeaway: losses aren’t necessarily a bad thing. In fact, they can be incredibly valuable... if builders learn from them. A company that has faced challenges and improved its processes may actually be a better bet than one with a spotless (but potentially lucky) track record. At the end of the day, better collaboration is the goal. When builders, brokers, and insurers dig deeper, looking beyond builders' numbers and into the “why” behind them, they can create more stability in an otherwise unpredictable space. And that’s a win for everyone involved.

When investors evaluate homebuilders, it’s tempting to zero in on the usual metrics: revenue growth, margins, and land positions. These matter, of course. But there’s another, less talked-about factor that plays a role in an investor's evaluation of a builder: Does the builder genuinely care about quality? At first glance, “quality” might sound like a soft concept. It’s not. In homebuilding, it shows up in very real, very financial ways. Builders who prioritize quality tend to have fewer defects, fewer warranty claims, and fewer costly surprises down the line. That consistency protects margins and reduces volatility... two things investors care deeply about. It also shapes reputation. Buyers talk, and a builder that is known for delivering solid, well-performing homes earns trust over time. That trust translates into stronger demand, better pricing power, and a more resilient business when the market cools. In a cyclical industry, that kind of stability is hard to come by. But there’s something deeper at play. Builders who care about quality usually have stronger internal discipline. They invest in better processes, tighter construction standards, and more accountability across their teams. They pay attention to customer feedback and actually act on it. Those habits don’t just improve the homes - they improve the entire operation. And when problems do happen (because they always do in construction), quality-focused builders tend to handle them differently. They treat issues as learning opportunities and make changes that prevent those problems from becoming systemic. Over time, that creates a smarter, more resilient organization. For investors, this is the real takeaway: a builder’s commitment to quality is a signal. It tells you how the company thinks, how it operates, and how it manages risk. You’re not just investing in the homes they build... you’re investing in the mindset that drives how that builder chooses to operate in a difficult, competitive environment. And in the long run, that mindset can make all the difference.

When people talk about quality issues in American housing, the focus is usually on builders or homeowners. But there’s another group that feels the impact just as directly: manufacturers. When quality slips, the ripple effects travel quickly across the homebuilding supply chain. For manufacturers of building products, poor construction practices can create a misleading picture of product performance. If an installation is done incorrectly or a system is used outside of its intended design, failures can occur even when the product itself isn’t at fault. Still, the manufacturer’s name is often the one that gets blamed. This can lead to increased warranty claims, reputational damage, and even legal exposure. Over time, these costs add up. Manufacturers may be forced to tighten warranty terms, raise prices, or invest more heavily in technical support and field inspections... just to protect themselves. There’s also a data problem. When quality is inconsistent in the field, it becomes harder for manufacturers to gather clean, reliable performance data. This makes it more difficult to improve products, forecast issues, or confidently innovate. In a sense, poor construction quality muddies the feedback loop that manufacturers rely on to improve. Relationships with builders can suffer, too. Manufacturers want to partner with companies that install and use their products correctly. When quality varies widely, it creates friction—more callbacks, more finger-pointing, and less trust on both sides. On a broader level, persistent quality issues in housing can slow the adoption of new technologies. Manufacturers may hesitate to introduce advanced systems if they’re not confident that those systems will be installed and maintained properly in the field. In the end, housing quality isn’t just a builder problem, it’s an ecosystem issue. And for manufacturers, it can directly influence costs, innovation, and long-term growth.

Individually, the small things in life, or the finer details, often seem insignificant. This is a natural human assumption. But what happens when you take all the little things and put them together? If each detail is treated as an afterthought or given little value, the end result is inevitably diminished. This premise is especially true in the construction quality and durability of the average newly built home. So why do we take these things for granted? A minor oversight – an improperly sealed window, a missing fastener, or insufficient insulation – may not be immediately noticeable. Yet, when these small issues accumulate, they can lead to significant problems such as moisture intrusion, energy inefficiency, reduced comfort, and costly repairs. Conversely, when every detail is executed with care and precision, the collective result is a home that performs beautifully and stands the test of time. For builders, this level of attention isn’t just craftsmanship – it’s smart business. It reduces risk, protects reputation, and leads to stronger customer satisfaction. Consistency in execution is what separates average builders from exceptional ones. For homeowners, choosing a builder who prioritizes these details means more than a visually appealing home. It means confidence in how the home will perform over time… through changing seasons, years of use, and evolving needs. The takeaway is simple: the “little things” aren’t little at all. They are the foundation of quality. When they’re consistently executed well, the end result is not just a house, but a high-performing home that is built to last.

I recently met with a group of Trade Contractors who are primarily focused on new, residential construction. I was asked, “what is it that builders are looking for, beyond price, in their selection of who to award their business?” I’ve been in this industry for more than 30 years, and this was the first time that I have been asked this specific question from a Trade's point-of-view. We engaged in conversation and in the end, I was surprised to discover that what each side is looking for is fundamentally the same: effective jobsite management. Effective jobsite management isn’t just about keeping a build on track – it’s a direct driver of profitability, reputation, and long-term partnerships. For both homebuilders and trade contractors, the same core considerations shape success on every project. Reliability and schedule adherence sit at the center of a well-run jobsite. For homebuilders, staying on schedule protects cash flow, keeps closings on track, and prevents costly trade stacking when delays ripple across the project. For trade contractors, reliability is what earns repeat work. Builders prioritize trades who show up, hit deadlines, and communicate early when issues arise. In the end, predictability keeps everyone profitable. High-quality workmanship is equally critical. Builders depend on consistent quality to reduce warranty claims, pass inspections the first time, and protect their brand in a competitive market. For subcontractors, quality directly impacts their bottom line. Poor work leads to rework, lost time, and strained relationships, while strong performance builds trust and secures future opportunities. Safety compliance is more than a regulatory requirement - it’s a business necessity. Builders face significant risk exposure from jobsite incidents, including delays, legal costs, and rising insurance premiums. Subcontractors share that risk. A safe jobsite protects workers, avoids fines, and ensures that crews can keep working without disruption. Professionalism and communication tie everything together. Builders rely on clear, consistent communication to coordinate multiple trades and resolve issues before they escalate. Subcontractors benefit from the same discipline: clarity around scope, timelines, and expectations reduces misunderstandings and keeps crews productive instead of waiting on direction. Strong communication ultimately builds long-term partnerships. Licensing and insurance provide the foundation for risk management. Builders need properly credentialed trades to ensure compliance and protect against liability. Subcontractors, in turn, need licensing and insurance not only to qualify for work but also to safeguard their business from potentially catastrophic losses. It’s a shared layer of protection that enables projects to move forward with confidence. Cost efficiency is about more than being the lowest bidder. Builders depend on accurate, reliable pricing to protect margins and forecast effectively. Subcontractors who estimate well and stick to their numbers avoid disputes and maintain healthy cash flow. Fair, consistent pricing strengthens trust on both sides. Finally, site management – including cleanliness, organization, and security – has a bigger impact than many realize. Builders benefit from safer, more efficient jobsites that reflect well during inspections and buyer walkthroughs. Subcontractors gain from improved productivity, reduced material loss, and a more professional working environment. In the end, these seven considerations aren’t just operational checkboxes – they’re shared business priorities. When both builders and subcontractors align around them, projects run smoother, risks are reduced, and relationships grow stronger.

Has your team adopted the use of AI as a part of its business development efforts? Many have. While it may offer some benefits, giving AI tools access to your valuable business data can pose real risks. We recommended treating AI as a system whose risks are managed, just like any other enterprise risk. So how can we be safe while using these tools? NIST and OWASP recommend the following to protect your valuable data while using AI tools. Classify your data.Users are the first line of defense for the enterprise. Determine what types of data/documents are acceptable to send to AI. As an organization, assign classification to your documents, such as “public,” “internal,” or “confidential,” and then decide which levels of classification can be used with AI. Simple guidelines such as “do not paste sensitive information into AI tools” can go a long way to guard against leaks. Restrict what the models can do and see. When working with large volumes of structured data, have AI generate scripts or an algorithm that can perform the desired action rather than uploading the file to the AI tool. For example: If you’re working with an excel sheet that contains user information, instead of uploading the document to AI and exposing valuable user information, ask AI to write a script that can pull out the first name and last name out of the excel sheet. This way, valuable data is never exposed outside of the company. Treat AI outputs as untrusted until validated. AI is not perfect and can make mistakes. Before using or trusting any of the data that AI gives you, (including scripts, SQL statements or information that you’re trying to lookup/reference), make sure that you verify the information manually. Sometimes the output from unknown tools can be hazardous. OWASP recommends that output from AI tools must be sanitized to remove cross-site scripting, SQL injection or remote-code execution attacks. This helps us minimize the risk. Look out for data and model poisoning. Currently, a very popular type of cyberattack is the insertion of “backdoors” into an organization’s software repository. A backdoor attack is when a malicious user creates or makes use of an existing vulnerability to access a system, application, or dev/ops framework; while bypassing normal security protocols that are in place. Think of it as secretly installing a hidden entrance that no one knows about. Attackers may try to use this to steal data, or poison existing data or models that affect reliability and accuracy. It may also be used to insert biases into your data, changing its meaning. This risk can be controlled by monitoring the model’s behavior to ensure that it stays consistent, vetting data vendors, or executing untrusted data sources in an isolated sandbox environment and tracking where the data came from. By treating AI as a managed risk - through careful data classification, controlled access, vigilant validation, and proactive threat monitoring - organizations can harness its benefits without compromising the security and integrity of their most valuable information.   Sources: OWASP https://genai.owasp.org/llmrisk/llm01-prompt-injection/#:~:text=A%20Prompt%20Injection%20Vulnerability%20occurs,is%20parsed%20by%20the%20model  NIST https://www.nist.gov/itl/ai-risk-management-framework?utm_source=chatgpt.com