Early in my career, I was in a room where the sponsor kept saying "we should use AI" and the data scientist kept nodding—but they were picturing completely different things. The sponsor wanted a chatbot; the data scientist was thinking about a regression model. Nobody realized it until three months in. Now, the first thing I ask is: "What outcome are we trying to change—and who or what is going to change it?" If the answer is vague, that's my signal to slow down and get concrete before anyone writes a line of code or signs a vendor contract. Saving Changes...
"اسأل ما المشكلة أو الهدف الذي يريد تحقيقه، وليس أي نوع من الذكاء الاصطناعي يريد استخدامه." أي أن تبدأ بتوضيح حالة الاستخدام (Use Case) أو المشكلة المطلوب حلها، ثم تحدد ما إذا كان الذكاء الاصطناعي هو الحل المناسب Saving Changes...
Bridget NatsamiProject Coordinator| American Tower CorporationUganda, Uganda
When someone says, "we should use AI," I usually ask what outcome they're trying to achieve, whether it's automating routine tasks, generating content, analyzing data, or improving decision-making. Those follow-up questions often reveal that people are using "AI" to mean very different things, and lumping everything together can lead to unrealistic expectations, scope creep, and choosing the wrong solution. In my experience, taking a few minutes to define the problem before discussing AI capabilities results in more practical conversations and better project outcomes. Once you understand the need, you get to know which AI tool to leverage to serve the intended goal.
First we need to asses feasibility, technical, organizational, & constraints, should we then work on defining the scope. Saving Changes...
AVEDIS GULUZIANManaging Partner & Operation Director | - Fit-Out Quality & Technical Work L.L.United Arab Emirates
When someone says, “we should use AI,” I first ask what problem we’re trying to solve, what outcome we expect, and whether AI is the best solution. Understanding the business need should always come before selecting the technology. Saving Changes...
Aziz UllahDistrict Monitoring Officer\Field Geologist| RTPS Commission Government of KPBattagram, KP, Pakistan
Yes we have been discussing using AI in Service delivery and for grievances redressal mechanism with top management, but we face the acomodation and using of general public and old officers using AI for such a purposes. Although the process started but so many people are reluctant to adapt the system and they preferred manual apply instead of using AI. Saving Changes...
In mining and construction, I often hear "We should use AI," but the first question should be: what problem are we trying to solve? For example, AI can improve safety by predicting high-risk conditions, reducing incidents by 15–30%; optimize project planning by reducing schedule deviations by 10–20%; support predictive maintenance to decrease equipment downtime by 20–40%; and improve productivity through better resource allocation, increasing equipment utilization by 10–15%. AI delivers value when it solves measurable business problems, not when it is adopted just because it is a trend. Saving Changes...
Usually when I hear someone suggest using AI, they most often are trying to speed up routine work (drafting emails, taking notes, etc.) or leverage AI to brainstorm new ideas, create presentations, etc. They're often referring to using common tools such as ChatGPT, Claude, Gemini, or Copilot. To understand what's being asked when someone says "let's use AI", it's critical to understand the outcome they're trying to achieve and what they believe AI can do for them to achieve that outcome. Saving Changes...
Bhavana MVAVP, Project Implementation Lead| State StreetBengaluru, India
Some signals that help distinguish what kind of AI work is actually being discussed: Objective: Is the goal prediction, automation, optimization, content generation, decision support, or autonomous action? Data dependency: Is the system primarily driven by historical structured data, real-time data streams, knowledge bases, or large-scale pretraining? Output type: Does it produce a forecast, recommendation, classification, generated content, or an action? Model complexity: Are we talking about traditional machine learning, deep learning, generative AI, computer vision, reinforcement learning, or rule-based automation? Human involvement: Is the human reviewing outputs, making final decisions, or completely out of the loop? Business problem: Is it solving an operational challenge, customer experience issue, productivity need, or a research problem? A common giveaway in conversations is when people use the same examples interchangeably:
"We need AI for demand forecasting, document generation, predictive maintenance, and customer support."
Those may all involve AI, but they require different data, skills, governance models, success measures, and implementation approaches. The moment I hear people describing AI primarily by the technology ("we need GenAI") rather than by the problem they're trying to solve, it's usually a sign that different participants are talking about different things under the same label. A useful question that quickly clarifies the conversation is: "What specific decision, task, or outcome do you want the AI system to improve?" The answers usually reveal whether the discussion is really about analytics, automation, machine learning, generative AI, or something else entirely. Saving Changes...
PMI Team Member
