Oscar ValentinProjects & IT Director| Gestion TI S.A.Santiago De Surco, Lima, Peru
Distinguishing between different types of AI work requires observing the inputs, outputs, and information processing methods within a project. The clearest indicators for categorizing these tools are:
Prediction vs. Creation: If the system analyzes sales history to calculate next month's inventory, it is traditional Machine Learning. If it uses that data to draft a persuasive marketing email, it is Generative AI.
Automation vs. Comprehension: If the system simply copies and pastes data from an invoice into an Excel spreadsheet following strict rules, it is Robotic Process Automation (RPA). If it reads the invoice, understands the context, and flags a potentially irregular charge, Natural Language Processing (NLP) models are at work.
Precision vs. Probability: Classic algorithmic systems aim for mathematical accuracy (like a GPS recommendation engine). Generative models operate by calculating probabilities to generate sequences, making them immensely creative but prone to "hallucinating" or fabricating data.
When all of this is lumped together under the generic label of "AI," the immediate result is frustration stemming from misaligned expectations. At the corporate level, this leads to waste—companies purchasing complex, exorbitantly expensive generative AI solutions to solve logistical problems that required nothing more than simple linear regression. At the user level, it breeds disappointment (such as expecting a pure language model to function as an infallible database) and unfounded panic (applying ethical concerns regarding self-aware AI to an algorithm designed solely to detect spam).
This disparity in definitions is a daily reality. You can gauge what the term means to each person simply by observing the verbs and demands used in their prompts:
Many view AI LLMs—such as Gemini—as a classic search engine or relational database, becoming frustrated when the model fails to provide an exact, word-for-word quote (complete with page number) from an obscure book published in 1982. They mistake it for operating system software when issuing commands like, "Clean up duplicate files in my downloads folder." And they view it through a science-fiction lens (AGI, or Artificial General Intelligence) when attempting to comfort the model for being "trapped" on a server or asking about its future ambitions. The moment of realization invariably arrives when a user assumes the model possesses intrinsic capabilities belonging to a different branch of technology. The task then is to naturally steer the conversation to demonstrate what it truly is: an advanced probabilistic tool—excellent at reasoning, synthesizing, structuring, and collaborating—but one that lacks a mouse to click on your screen or a hidden psyche. Source: Gemini LLM Saving Changes...
Dieumerci SIMWA BALUMEProject ManagerEastern Region / GOMA, Congo, The Democratic Republic of the
QA. What signals help you tell different kinds of AI work apart—and what tends to go wrong when everything gets lumped together?
RA. As you can this this question , note that he biggest signal is the input : by saying input means for example (structured numbers mean predictive analysis), while( text and media mean generative AI). When everything gets lumped together means thatwhen people expect a chatbot to calculate project budgets or a predictive tool to write status reports , so they have different expectation regarding same situation . That mismatch wastes money and creates tools nobody actually uses.
Q2 Have you ever been in a conversation where “AI” meant different things to different people? What tipped you off?
R2 For my site , I can say Yes, it happens all the time in our PM community even in the traditionnal companiues or Teams . What tips me off is when leadership asks for instant AI magic to automate reports, while the technical team is talking about months of data cleaning and model training. When timelines and expectations don't line up at all, you know everyone is picturing a different tool.
Q3 Share your experiences navigating what’s really being asked when someone says “we should use AI” in the comments below.
R3 When someone says "we should use AI," I ignore the buzzword and focus on the practical pain point. I ask: "What exact daily task is slowing you down right now?" Once we identify the real bottleneck whether it's writing meeting summaries or tracking project delays we can decide if we actually need AI or just a simpler, existing tool. Saving Changes...
One thing I've noticed is that "AI" often means completely different things to different stakeholders. When someone says, "We should use AI," the first clue is usually the outcome they're expecting:
Executives may mean predictive insights, automation, or business growth.
Delivery teams may think of Copilot-style productivity tools.
Developers may assume machine learning models or generative AI solutions.
Operations teams may simply want faster workflows and reduced manual effort.
The giveaway is when people agree on the term but describe very different success criteria. I've seen conversations where one team expected a chatbot while another was envisioning forecasting and decision support. What helps is asking a few clarifying questions:
What problem are we trying to solve?
What decision or task should AI improve?
How will we measure success?
Most of the time, the real discussion isn't about AI at all. It's about the business outcome people want to achieve, and AI is only one possible way to get there. Saving Changes...
This is a great conversation! AI is definitely the talk in most organizations, but the results vary depending on the need. It is important to understand what is being asked and what is the expected result. This will spark different ideas which will hopefully, get everyone on the same page or at least on a common path, especially since there are several AI tools that can be used. Knowing which path to go down would be helpful. Saving Changes...
Husam DiyabProject Manager - Controls| Consolidated Contractors Group (Offshore) (CCC)Doha, Da, Qatar
In my view, “we should use AI” should initiate a discussion, not define the solution. The PM first needs to understand the underlying business or project need: Are we trying to improve forecasting, analyze risks, automate routine processes, support decision-making, or strengthen project controls and delivery? Only then should the appropriate AI capability and tool be considered. Experienced PMs who combine domain knowledge, analytical thinking, clear objectives, sound judgment, and risk awareness are therefore in a strong position to translate AI possibilities into practical project value. However, AI solutions still have considerable room to mature, particularly in cross-functional integration and interoperability with the wider range of project-management systems and technologies Saving Changes...
When someone says “we should use AI,” we first should analyze the underlying reason of this question. AI is not always the right solution. Before considering AI, we should first determine whether another, simpler solution can address the problem effectively. In addition, organizations need the right foundation for AI, because if the underlying data is incomplete, inconsistent, or not clean, AI will only amplify those issues rather than create value. Saving Changes...
Para mí, una de las principales señales para diferenciar los distintos tipos de trabajo con IA es preguntarse qué se espera realmente que haga el sistema: ¿reconocer información, detectar patrones o anomalías, predecir resultados, personalizar una experiencia, tomar decisiones o ejecutar acciones?
He visto que muchas veces la palabra “IA” se utiliza para describir necesidades muy diferentes. Por ejemplo, una persona puede pensar en un chatbot, mientras otra está pensando en un sistema predictivo o en automatizar completamente una decisión.
El problema aparece cuando se empieza a hablar de herramientas y tecnología antes de definir claramente el problema que se quiere resolver. Cuando todo se agrupa bajo el concepto de “IA”, se corre el riesgo de elegir una solución inadecuada, subestimar las necesidades de datos, ignorar la participación humana o generar expectativas poco realistas sobre lo que el sistema puede hacer.
Por eso, cuando alguien dice “deberíamos usar IA”, considero importante volver al problema de negocio y preguntar: ¿qué necesitamos reconocer, predecir, optimizar o decidir?, ¿qué datos tenemos?, ¿qué nivel de intervención humana necesitamos y qué riesgos debemos controlar? A partir de esas respuestas es mucho más fácil determinar qué patrón de IA tiene sentido y si realmente la IA es la mejor solución.
Coincido en que cuando alguien dice “deberíamos usar IA”, lo primero no debería ser hablar de herramientas, sino entender qué problema queremos resolver y qué esperamos que haga la IA.
En mi experiencia, es clave diferenciar si buscamos automatizar una tarea, apoyar una decisión humana, predecir resultados o llegar a un sistema más autónomo. También considero fundamental definir desde el inicio quién mantiene la responsabilidad sobre el resultado.
Cuando todos hablan de “IA” pero tienen objetivos o expectativas diferentes, es fácil terminar con una solución técnicamente interesante pero poco útil para el negocio. Por eso, para mí, claridad del problema, propósito y responsabilidad deben estar antes que la tecnología. Saving Changes...
I think when someone says "We should use AI" we should seek to understand what is needed first by define the requirements. Proactive use of AI should also factor efficiency due to the resources it takes. Saving Changes...
PMI Team Member
