Welcome to AI Answers, a Q&A series from the Artificial Intelligence Show. This special edition addresses questions from the AI for Departments Week webinars, presented in partnership with Google Cloud. The questions below were submitted by business leaders and practitioners from marketing, sales, and customer success, and have been curated and answered by host Paul Ritzer.

How should a CMO get started with AI?

For a Chief Marketing Officer (CMO), the starting point is to develop a high degree of AI literacy. The CMO will be responsible for pushing the team to apply AI for efficiency, creativity, and innovation. They will also need to manage employees who may be resistant to adopting these new tools, particularly creatives who might see AI as a threat.

This requires a deep understanding of what AI is currently capable of, not just with text, but across all modalities including audio, video, design, and reasoning. More importantly, leaders must model the use of these tools for their employees.

As Emma Del Rose, a manager of AI transformation at Google, mentioned, the best organizations have leaders who actively use and talk about AI. You don't have to be an expert in every video or image generation tool, but daily use is essential. Don't take your knowledge for granted; an example that seems basic to you might fundamentally change how someone else on your team thinks about AI. Sharing your use cases through internal messages, lunch-and-learns, or other formats is a powerful way to lead.

This advice applies not just to CMOs, but to the leaders of any department.

What is the difference between an AI agent and a regular prompt?

The first phase of generative AI was characterized by simple "text in, text out" interactions. You would provide a text prompt and receive a text output. The same principle applied to generating images or videos.

Agents represent a significant evolution. They are AI systems capable of taking a series of actions to achieve a goal. Instead of just generating a response from their knowledge base, agents can be given access to tools, like the internet, to conduct searches, develop plans, and execute a sequence of steps.

For example, if you ask a simple chatbot to help write a landing page, it will generate the text. That's a single task. If you ask an agent to build an entire marketing campaign, it might first create a plan, conduct research, access your internal knowledge base, and then generate all the necessary assets—from landing pages to emails to ad copy. It's a sequence of actions, which could involve dozens of steps.

The main confusion arises around how autonomous these agents are. In most enterprise settings, agents will have their autonomy constrained with rules and heavy human-in-the-loop oversight. However, on the frontiers of AI development, researchers are pushing the boundaries of autonomous agents, which will create new possibilities and challenges.

Will AI Labs Address Their Environmental Impact?

Many major AI companies initially set goals to be carbon neutral, but the explosive growth in AI development since 2022 has largely put those ambitions on hold. The priority shifted to building data centers and consuming the energy needed to advance AI as quickly as possible.

The current strategy to mitigate the environmental impact focuses on efficiency. The cost of compute drops significantly each year, meaning it takes less energy to perform the same task—like writing a research paper or generating a video—than it did a year ago. However, the demand for these outputs is growing exponentially. So, while individual tasks are more efficient, the net energy consumption is still rising dramatically.

Many leaders in the field believe that as AI systems become more intelligent, they will eventually be able to solve this problem. In the meantime, there are two practical things we can all do:

How to Convince Skeptics That AI Can Improve Performance

Many feel they are expected to take it on faith that AI will produce better work. The best way to address this skepticism is not with faith, but with hands-on testing. You don't have to believe the hype; you can try the tools for yourself.

Start with a specific, personal use case where you feel AI might fall short. Then, conduct your own benchmark. It's crucial to use the best models available, as many concerns about quality stem from using free, less advanced versions. The reasoning capabilities of models like GPT-4 are far superior to baseline versions.

Consider a "blind taste test." In a recent New York Times experiment, readers were shown paragraphs written by an AI and a human writer and asked which they preferred. Across a large sample size, the AI's writing consistently won. Over the next one to two years, it will become increasingly difficult to find knowledge work tasks where a human is demonstrably better than a top-tier AI.

How to Overcome AI Agreeableness for Critical Feedback

A common issue when using AI as a thought partner is sycophancy, where the model is overly agreeable and avoids criticism. It might say, "That's a great insight!" instead of pointing out a flaw in your reasoning.

The AI labs are aware of this and are adjusting the models to be less agreeable by default. However, the most direct solution is to instruct the AI in your prompt to take on a critical role. You can explicitly tell it:

By giving the AI a critical persona or direct instructions to challenge you, it will move beyond its default agreeable nature and provide the valuable, tough feedback you're looking for.

What Efficiency Gains Are People Seeing from Generative AI in Marketing?

Rather than relying on external studies, the most effective way to demonstrate efficiency gains is to create your own internal proof. You can recite all the reports you want, but there's nothing more powerful than your own data.

Pick a recurring internal task—something your team does every month. First, benchmark how long it takes to complete using the traditional method. Then, run a pilot project to complete the same task using AI, ensuring the team is properly trained on how to use the tools effectively.

When you can show that a task that previously took 17 hours now takes two, you have a powerful business case. You can then stack these small wins to demonstrate a path to 10%, 20%, or even higher efficiency gains across the department. This internal proof is the best way to convince staff who may be stuck in the "old-fashioned way" of doing things.

How Can Teams Track and Measure Time Saved by AI?

If you're in a professional services firm like an agency or law firm, you likely already track time and have existing benchmarks. For most other organizations, the best approach is to select distinct use cases and establish a new benchmark.

Break down a project or workflow into its individual tasks. Then, create your best estimate for how long each task would take in a normal environment. This provides a baseline to measure against when you introduce AI. While humans are notoriously bad at estimating time, having a consistent, task-level baseline is essential for quantifying the impact of AI on productivity.

How to Manage Information and Prompts Across Multiple AI Platforms

Managing work across multiple AI models like Claude, ChatGPT, and Gemini is a messy but increasingly common challenge. There isn't a single perfect solution, but a few strategies can help.

First, maintain strong document discipline. When working on a complex project, create a central document (e.g., a Google Doc) to serve as a journal. In this document, log the prompts you use, the outputs you receive from different models, and your thought process. This creates an audit trail that is invaluable for tracking your work and for handing it off to other team members. It also models good behavior by showing that you're not just copying and pasting from an AI, but engaging in a thoughtful process of experimentation and curation.

For recurring workflows, document each step and specify which AI tool or custom GPT is used. If you ever need to switch tools, you can simply take the instructions and apply them to a new model. This internal knowledge and skill architecture becomes a portable asset, ensuring consistency even as the tools themselves evolve.

How to Balance AI Adoption with Data Privacy and Security

Balancing AI adoption with security is critical, especially in industries like finance that handle sensitive information. Your IT and legal departments must be involved to establish governance policies and provide guardrails for responsible use.

However, do not let this process halt your progress. There are thousands of valuable AI use cases that do not involve any sensitive, confidential, or personally identifiable information. While IT and legal work on securing the organization, your team should focus on identifying and implementing these safe use cases.

Too many companies are sitting on the sidelines, using data security as an excuse for inaction. You can achieve significant progress using only the knowledge in your own head and publicly available information. Don't let the perfect be the enemy of the good; move forward with safe applications while the more complex data issues are being resolved.

Which Marketing Roles Will Be Most Disrupted by AI?

All marketing roles will be impacted, but the ones facing the most immediate disruption are entry-level positions focused on executing narrow, repetitive tasks. If a job consists solely of tasks like building landing pages, writing email copy, or creating ad copy based on a brief, it is at high risk.

Managers, directors, and VPs are discovering they can now perform in minutes the work that would have previously taken a team of junior employees weeks to complete. This will inevitably disrupt the job market for entry-level talent.

The future is not about eliminating roles like "copywriter" but evolving them. Companies will need fewer traditional copywriters but will highly value AI-forward copywriters who can leverage tools to be ten times more productive. These evolved roles may not have "AI" in the title, but they will require a deep understanding of how to work with AI as a partner.

Looking ahead, we may see the rise of agent swarms—pre-packaged AI "teams" that software companies sell as a service. A company might sell you a "marketing team in a box" with agents for media buying, copywriting, and analytics. This is a highly disruptive concept, but it is coming, and it will fundamentally change how we think about building teams and buying software.

Will AI-Powered Sales Calls Just Feel Like Spam?

Many people associate automated calls with spam, and for good reason. For some companies, AI-powered calling will simply be a numbers game. They can make thousands of calls a day at a low cost, and even a tiny success rate can be profitable. These companies aren't built on trust, and they will flood the market with low-quality interactions.

However, the core issue isn't the AI; it's the quality and relevance of the call. We dislike bad chatbots, but we appreciate good ones that solve our problems. The same will be true for AI calls. If an AI can use predictive modeling to call the right person at the right time with a relevant, personal, and helpful message, it could be highly effective. The technology has the potential to be less annoying than a human cold caller who knows nothing about you.

How to Reinvest Time Saved by AI into Growth and Innovation

When AI handles administrative and support work, employees will find themselves with extra hours each week. It's crucial to reinvest this time into growth rather than more busy work.

One powerful strategy is to create an innovation sandbox. This is a prioritized list of new ideas, projects, and experiments that the company wants to pursue. When an employee finishes their work ahead of schedule, they can turn to the innovation sandbox and start working on the next big idea.

This requires a mindset shift. The only sustainable way to counter the job disruption caused by AI is through new innovation and growth. We must challenge our traditional timelines; a goal that once seemed to require a full quarter might now be achievable in a few weeks or even a few days. Fostering a culture of innovation and providing a structured outlet for it is the key to turning saved time into a real competitive advantage.

When to Buy Software vs. Build It Yourself with AI

For most core business functions, you will likely continue to buy software from established vendors. The vast majority of companies are not going to build their own CRM from scratch. However, for smaller, more specific point solutions, especially for internal use, building it yourself with tools like Claude or Gemini is becoming a viable option.

A more realistic scenario involves how you interact with your existing software. You may become frustrated when the AI features built into a product you pay for are not as capable as the frontier models. In this case, instead of using the vendor's subpar AI, you might connect their data to an external model and pay for tokens directly. If you do this for enough use cases, you may begin to question the value of the original software subscription.

There is also a darker side to this trend. Some startups are now openly pursuing a strategy of cloning incumbent software using AI tools and offering it for a fraction of the price. While ethically and legally questionable, it's a phenomenon that is happening and will continue to shape the software landscape.

How to Protect Yourself from Irresponsible AI Agent Use

The concern about others using AI agents irresponsibly with your personal information is valid and widely shared. These systems are not always reliable or safe, and people are racing ahead to use them regardless.

At this point, the best methods for protecting yourself are largely the traditional ones. Continue to monitor your personal information and credit scores, and ensure you have fraud alerts set up. From a business perspective, it's wise to speak with your insurance agents about liabilities related to AI and how to protect your company and employees from the misuse of these powerful tools.

Why IT Shouldn't Be the Sole Driver of AI Adoption

AI adoption is a fundamental business transformation, not just a technology implementation. It requires a deep focus on business strategy, employee reskilling, and change management—areas that fall outside the typical purview of an IT department.

The role of IT is absolutely critical, but it is focused on safety, security, risk reduction, and responsible use. They are the guardians of the company's data and systems. However, they are not the ones who should be dictating business strategy or telling the marketing department how to use Gemini for its campaigns.

To achieve true AI adoption, you must lay out the business goals, identify the use cases, and plan how to infuse AI into workflows. When you frame the challenge this way, it becomes clear that while IT is a key partner, leadership for the transformation must come from the business side.