The AI Agents are here - How should we market them?

Many claim AI agents are the future of software. 

Founders and marketers need to figure out how to market them.

A guide for marketing these Service-as-Software to B2B buyers. Part I.

AI companies and their marketers are already pitching their armies of tireless autonomous AI agents to B2B buyers: cybersecurity analysts (Dropzone AI), sales development reps (11x’s Alice - Artisan’s Ava), software engineers (Cognition Lab’s Devin), administrative nurses or healthcare agents (Tennr, Hippocratic’s Linda), legal assistants (Harvey), and many more. 

These AI agents power a new era of service-as-software. They’re attracting significant VC attention and funding as they are about to radically transform the margin profiles of many service industries and open up their massive TAMs.

For a better sense of the opportunity about to be unlocked in human-intensive service businesses, check out the massive gap between human and AI labor costs (and keep in mind that the cost of AI is decreasing fast while quality is improving just as fast…): 

We should then not be surprised to see more AI agents unleashed upon us with more and more capabilities. 

As founders and marketers have learned to market SaaS, APIs, and, in the last year, more AI and copilot products and features, many now need to learn to market AI agents. But how? What’s been done so far? 

I have been fortunate to help market an advanced cybersecurity agent, Dropzone AI, and to participate in CMO discussions on the topic. In this guide, I share observations, guidance, and examples. I want to start a dialogue and collect more thoughts and examples to help founders, product leaders, and marketers better market theirs. 

I am publishing four articles initially:

1. This introduction to AI Agents positioning and its essential questions to address, including whether to anthropomorphize agents
2. The anatomy of AI agent positioning with examples
3. Critical collateral to inspire trust and preference for AI agents
4. Pricing AI agents

Let’s dive in!

What are AI agents? 

We’re now very familiar with AI chatbots (ChatGPT, Google’s Gemini, Anthropic’s Claude, etc.) and copilots (Microsoft Copilot and Github Copilot) that we prompt and collaborate with to write better, create images, answer questions, and code faster. They make us more productive, more creative, etc. 

AI agents go beyond copilots: they operate autonomously, handling tasks without humans at the wheel. Agents plan, orchestrate, execute, self-adapt, and correct their own work (e.g. Devin’s software engineer by analyzing error messages). 

The work product of agents is a service delivered by that autonomous software. 

We receive the work's output: a qualified lead, a full investigation report, functioning code, an insurance claim package, a resolved customer support case, and many other possibilities.

Agents possess and tap specific skills and access tools to perform their service, such as a web browser, apps, APIs, and more. 

Positioning agents vs software: the differences to keep in mind

In B2B software, marketers identify a business problem and show how humans leverage and derive benefits from it using their software’s capabilities to address these problems. 

“Armed with our software, YOU will achieve {benefit}, by doing {job-to-be-done}, because the software possesses {features}.”

In contrast, is marketing AI agents “simply” about marketing the agents’ outcomes and savings: a service delivered at a fraction of the cost, with virtually infinite and elastic capacity, at the speed of compute? 

“Our agent will autonomously deliver {service or outcome} cheaper and faster than (your) employees.”

That can sometimes be the answer, but the reality is more complex.

AI agents are a brand new category: service-as-software. As with new and disruptive categories, marketers must educate, inspire, reassure buyers, and socialize the right playbooks to use. With AI at the core and the potential to replace entire functions, they raise organizational and ethical questions in addition to all the usual questions associated with buying new software categories. 

Understanding buyers’ key questions and hesitations will help tailor positioning, content, and GTM motions to market agents effectively, including: 

• The agents’ place and role: employee augmentation or replacement?
• The quality of their work: inferior, similar, or superior to that of human workers? Consistent and predictable enough?
• The guardrails to apply when using them: whether and how to a) manage and supervise them and their output? b) let them interact with customers?
• Security and privacy: how are agents handling data? Is that data used to train other models? Will others benefit from it? Are they introducing new security threats?

The following sections and articles will provide examples of this done well. 

A fundamental question: position the agent as augmentation or replacement of employees?

Should one position their agent as skilled, adaptable, and autonomous enough to replace a human entirely in a given position, as Artisan and 11x don’t hesitate to call their agents digital BDRs or marketers? Or are they there to augment these roles and become autonomous assistants of employees holding these roles, handling some of their tasks—not their entire job—autonomously for them? 

While it’s tempting to pick “replacement” because of the buzz it can generate, the staffing budgets one can tap, and the resulting willingness to pay, it comes with risks given that’s a bold claim that could rally an entire profession against you (see the backlash Devin suffered from software engineers following its early claims). 

Choose “augmentation” when:

• The agent takes over specific - often repetitive - tasks but not a whole job.
• The industry is not ready to accept that claim, especially when there are fundamental human traits that agents cannot yet handle. 
• The technology is not yet ready. That seems obvious, but so many vendors looking for buzz ignore it, which hurts them and the entire category.
• You want to champion that role. 

If you go with augmentation, you can avoid using the category name “agent” for now and describe the agent as an autonomous assistant, analyst, copilot, etc., at least until agents are better trusted. 

Choose “replacement” when:

• The agent can truly replace an entire well-defined position and the majority of its tasks, and you can substantiate that.
• The role consists of repetitive tasks that employees don’t enjoy doing or never have enough time to perform (e.g., too many cases to handle).
• There is usually a human supervising or validating the outcome anyway (see the Dropzone example above or consider the case of a sales lead created by an SDR: there is usually an account executive who will assess the quality of that lead before deciding to accept or reject it. The same goes for a pull request that a developer needs to review first). 
• The employer/buyer has recruiting or retention issues for that role due to a skills shortage, a role with traditionally very high turnover, etc.

When should we anthropomorphize AI agents, give them a first name, and brand them?

Given that these agents can do the same work—at least handle the same tasks autonomously—as humans, should their marketers give them a name, a face, and a personality to make them seem as human as possible? Should their marketing highlight their processes and outcomes instead? Should they use their existing brand instead?

Humanizing agents is a good idea if they interface with consumers and the agents try to substitute human interaction explicitly. Then, a buyer must believe that the AI agent possesses the same empathy, context, and diplomacy as a well-trained human to interface effectively with their customers. That will likely be true for customer care agents, AI tutors, coaches, etc. When the answer to “Does it matter how they make consumers feel?” is yes, then it should be a good option. 

There is less need to humanize an agent whose customer is internal to the buyer’s organization or when you want to highlight the agent's technical skills. 

Then, there is always the option to use your existing brand: as Decibel partner Jess Leao noted: “If you are Intuit and you provide a financial accountant, calling it Intuit Assist makes sense to double down on the Intuit branding”. The Intuit brand already has many positive associations, so why muddy it with a new personality? 

Below are examples of companies that took different approaches:

Hippocratic: humanizing the agent 

Given what their agents do, it makes sense that Hippocratic would humanize them: someone who will follow up directly with your patients after a medical procedure needs to be personable, reassuring, and show empathy. Hippocratic calls out the style of their agent and showcases a mock call on their site (notice how, in the discussion, the healthcare agent declares that she is an AI agent.)

Dropzone: an AI selling its reasoning, not its empathy

We did not anthropomorphize the agent or position Dropzone as a replacement for human security analysts. We decided to represent it as a thinking 3D cube that autonomously plans and executes each investigation and then writes a detailed alert investigation report. We positioned it not as a threat to security analysts (the people we champion and want to support) but as their tireless helper who takes on that grunt work 24/7. It doesn’t look like a human. Therefore, it’s obvious it never tires, works 24/7, and never complains about analyzing volumes of alerts and data points. Humanizing it would have little benefit: it would distract from its skills.

What's next?

Once these first questions are answered, it's time to craft the full positioning of these agents. The next article will explore this, starting with how to articulate the pain points, their value proposition, capabilities, and differentiators. Stay tuned.

Recommended reading

• The AI Workforce is Here: The Rise of a New Labor Market
• A New Frontier: Service-as-Software, powered by AI Agents 
• The Death of the Big 4: AI-Enabled Services Are Opening a Whole New Market