If you find it difficult to navigate the ever-changing generative AI landscape that shifts week after week as vendors compete on top leaderboards, you're not alone.
The AI training gap is manifesting throughout the business world. According to one recent survey, two-thirds of leaders expect employees to have AI skills, but only three-thirds have a clear policy on which technology and how to use it.
While using generative AI chatbots, the barrier to entries is relatively low, and I know the right Models remain a challenge for those who have not studied nuances. Here's what you need to know about choosing the right AI tools, such as ChatGpt, Claude, Gemini, Prperxity, and the underlying models that run them.
learn more: Mastering AI at work: A practical guide to using chatgpt, gemini, claude, and more
Speed days with AI
Generation AI today is still relatively new, and for this reason there are a wealth of models in the market. No one controls. As a result, it can be beneficial to try out a variety of tools.
“Think about it like driving a car,” said Jules White, a professor of computer science at Vanderbilt University. luck. “I recommend driving on the highway, parking, how the stereo works, how stiff the seats can be, and more, as a model equivalent.”
For Maggie Vo, head of human user education, the test drive needs to take into account three factors: task complexity, time sensitivity, and the need to improve tasks.
Something like creating a strategic plan might require you to use the most capable models, but faster models may be most useful for data and quick summary. If you plan to repeat multiple times, use a combination. “Start with smarter models and improve your approach and use smaller models to do it,” she said.
“The real skill is developing 'platform awareness'. We fully understand not only different models but different AI systems,” Vo added. “What works in Claude may need to be adjusted in other systems. Experiment between platforms to build intuition about your own strengths.”
In reality, this is just as easy as typing the same prompt with different AI chatbots. Each model (the options to change may be located at the top of the screen or near the search bar). Then compare which tools give the most useful response.
Screenshots from Claude of Mankind
As the VO implies, certain tasks are easier to complete than other AI models because AI models are being developed. Typically, the regular free model is usually best suited for simple chats. This usually results better, although more advanced models can be expensive and take more than a few seconds to process.
According to Upenn's Ethan Mollick, the best model
Ethan Morrick, a professor at the University of Pennsylvania Wharton School, is one of the most widely read experts on how to use AI models to assist with business tasks. He is known for his prolific research and analysis of LinkedIn. According to the recent Substack Post, his daily use of AI focuses on products from Openai, Anthropic and Google.
“The three options (three) give you access to both advanced and fast models, audio modes, the ability to view images and documents, the ability to run code, a great mobile app, the ability to create images and videos (Claude lacks here), and the ability to do deep research,” he writes.
The ultimate challenge, he added, stems from determining which model is best for use. But like White, he connects it to the vehicle choice. “Think of choosing between a sports car or a pickup truck. They're both vehicles, but they're used for very different tasks.”
Mollick has put together decisions into three categories: Suitable for chat, suitable for work, suitable for difficult problems.
Suitable for chat
According to Mollick, the Claude 4 Sonnett, ChatGpt-4o, and Gemini 2.5 flash are fast and not so clever. However, as generally free AI products, they are the most widely accessible and can make a difference in workplace efficiency.
Here are some examples of abilities:
- Market and competitive analysis: “Compare the positioning of Openai, humanity and Google Gemini in the Enterprise AI space.”
- Financial Modeling and Forecasting: “We will prepare a three-year revenue forecast for SaaS startups with a 20% year-on-year and a 75% total margin.”
- Creating and editing content: “In the draft LinkedIn post, we announced a $10 million Series A funding round that sounds confident but not bragging.”
Suitable for work
According to Mollick, the Claude 4 Opus, ChatGpt O3 and Gemini 2.5 Pro are smarter and have lower error rates, but slower. They may still hallucinate, but they tend to be more successful by coding, logic, or solving tougher problems that require multiple steps. Examples of these capabilities include:
- Market and competitive intelligence dashboard:
Who is better to summarise AI priorities than AI itself? Here's what the model had to say:
prompt: “In one sentence, explain your relative strengths. I'm writing this in the context of other generation tools. [Competitor] and [Competitor]. ”
Openai's GPT-4O: “The relative strength of ChatGpt lies in its ability to generate highly refined, context-conscious, and conversationally natural responses in editing, writing and task-oriented problem solving, while integrating tools such as code execution and image analysis more seamlessly than Claude and Gemini.”
Claude Sonnet of Mankind 4: “I generate subtle reasoning, ethical decision-making, and strong, thoughtful responses while maintaining a strong safety guardrail, but are more cautious and less willing to engage in boundary requests compared to other AI assistants.”
Google's Gemini 2.5 Flash: “Gemini is excellent at understanding and generating multimodals, leveraging real-time information from Google's vast ecosystem to provide highly accurate and contextually rich responses while seamlessly integrating text, code, images and other data formats.”
Several other examples include the ability to automate customer service and implement quality control, as well as Financial-Scenario modeling and risk analysis. Mollick admitted that he mainly uses a “job-friendly” model, but he also likes certain tasks outside of this list, such as using GPT-4.5 at the time of writing.
Good for difficult problems
The Claude 4 Opus Extended Thinking, ChatGpt-O3-Pro, and Gemini 2.5 Pro are the most advanced AI models on the market, and as such, they are slowly limited in use, Mollick added in his Substack Post.
Here is an example of a prompt to push the limits of ChatGpt, but Chatgpt-O3-Pro can be great to tackle.
“The following is the full 180-page Phase III Tumor Test Document, including a statistical appendix and a 40-page harmful event table.
- Extract all validity endpoints, their p-values, confidence intervals, and sample sizes.
- Recalculate the hazard ratio of the primary endpoint from the raw survival data listed in Appendix C and flag any inconsistencies with the sponsor's reported numbers.
- Summary all grades 3-4 adverse events grouped by organ system and calculate their absolute risk increase and control.
- FDA's 500-word briefing draft:
- Identifies statistical or methodological red flags.
- Assess whether the profit and risk profile justifies accelerated approval.
- We propose two post-market research designs to verify long-term safety. ”
Other models
No matter which model or AI companies are using it, it is always important to reassess AI for errors and hallucinations. Accuracy is a confusion priority, according to Jesse Dwyer, head of communications at an AI company.
“Perplexity's only focus is accurate and reliable AI. We use all top models and use post-train for accuracy,” says Dwyer. “Models trained to experience hallucinations can be useful when you need videos of high-diving cats, but can be dangerous when you're making business or financial decisions.”
Copilot is also a widely used AI chatbot thanks to its integration with Microsoft products, but switching between models is difficult. The Deepseek R1 and Grok by Elon Musk's Xai also have market options, but according to Mollick, each lacks functionality.
Morik did not respond luckRequest a comment.
Takeout: Practice is perfect
There is no perfect way to use AI in the workplace, but the most effective way to stay above the game is exploration and education. It starts with simply using them.
“The difference between casual and power users doesn't encourage skills (with experience). They know that these features exist and use them in real work,” writes Mollick.
Also, as many business leaders, including CEOs, have already begun to use AI, Dwyer suggested trying to emulate how they are using it in their own work.
“AI is one of the first business tools that managers adopted, not frontline workers,” he said. luck. “It makes sense that leaders who have the best experience of getting the best jobs from teams and software tools will naturally be prepared to work with AI.”
