Product Teams Make AI Sales Agents Smarter—and Vice Versa

This article was produced in collaboration with The CPO Council, a network of chief product officers organized by BCG and Lobby Capital.

It’s a classic conundrum: You launch a new consumer app—the result of a resource-heavy, expensive development effort—into a large market. The product is sufficiently intriguing that high numbers of potential customers show interest. But only 1% convert and purchase the software—even with a free trial offer. Ultimately millions of dollars in possible long-term revenue from could-be loyal customers are left on the table.

The problem is, people may be drawn to the app, but without constant encouragement aimed at overcoming hesitation, far too many target customers are turned off by the registration, activation, and onboarding process. But what if you had enough salespeople to handhold a million people through the initial steps? Could you change the outcome?

That’s an exercise that some early adopter companies are trying out: specifically, using AI sales agents to help with product education and closing the sale. These AI agents are seen as sophisticated partners to the business, capable of customized informational interactions expressly designed to nudge a prospect from a user into a paid customer.

Agentic sales represents a significant change and challenge for chief product officers (CPOs) at both product-led growth (PLG) companies, like the hypothetical app vendor, or sales-led growth (SLG) firms that might offer B2B products to large or small businesses.

Agents’ Evolving Capabilities

The technology is evolving fast. Tools that can qualify leads, guide conversations, and support or even replace parts of the sales process are becoming more accessible. AI agents can respond to the tire-kickers instantly with queries like: “What problem are you trying to solve?” “Can I show you how to use it?” “Can I give you a demo?” “Live or offline?” “Now that you’ve seen it in action and like it, how about a trial or subscription?”

Initially, especially in PLG companies, these agents will be somewhat limited—helping to guide product setup and gather feedback for sales and product teams. But before long, it is anticipated that more expansive uses, with AI applications implanted throughout the sales process, will be the norm. Ultimately, multiple functions in organizations, including sales, IT, go-to-market (GTM), strategy, and operations, could be involved in designing and working with these new applications. But, in concert with sales teams, product managers play the most indispensable role in successful implementation. They must take ownership of the content that powers AI agents, define the guardrails that shape how content is presented to the outside world, and ensure that the agents support, rather than disrupt, the customer experience.

At the same time, product teams can benefit greatly from the insights generated in agentic sales conversations. A single interaction with an AI agent that just spent a full day with dozens of customers can provide valuable data about customers’ preferences and purchasing decisions as well as information about competitor products. Perhaps most importantly, the agent can gauge customer sentiment in a calibrated way, offering more accurate, less emotionally driven or otherwise biased insights about prospective consumers than a human can. All of this is extremely useful for developing products, accelerating time to market, pricing, and designing and positioning products to meet the needs of potentially diverse customer segments. In essence, the information that AI sales agents gather through their customer interactions can be the basis of synthetic customer data sets that CPOs can use to gauge likely market responses to potential new products and features.

Not Just a One-Trick (PLG) Pony

Efficiently reaching a longer tail of prospective customers is a natural fit for AI. PLG models—where onboarding and servicing are built directly into a frictionless, self-service product experience—are common in B2C offerings like Spotify or WhatsApp. Layering AI agents onto this approach creates far more personalized touchpoints than human sales teams can deliver, and ideally that should boost conversion rates.

Some B2B software-as-a-service providers—for example, Zoom, Figma, and Slack—have embraced PLG, but most enterprise applications still rely on traditional sales-led growth. And at first glance, AI agents may not seem like a viable option to support these efforts. SLG campaigns typically hinge on frequent, sometimes lengthy interactions throughout the sales cycle and long after implementation. Customers bring highly specific, often unpredictable questions that seem to demand the constant involvement of sales teams equipped with deep, industry-specific expertise.

These factors don’t lend themselves to AI agents taking a broad-brush approach to customer engagement. Yet in such cases, AI agents can still serve as valuable backstops and force multipliers for sales teams by supporting, guiding, and suggesting next steps. They can fill information gaps, sustain momentum in ongoing conversations, and provide continuity when sales representatives are uncertain or unavailable to respond. Instead of letting prospects fall through the cracks or raising frustration levels among existing customers, these agents can offer high-value, non-generic nudges like: “Looks like you’ve got security concerns. Here’s how we meet SOC 2 and HIPAA standards” or “I see that Feature X has you puzzled. Do you want a quick overview of how teams in your industry typically use it in production?” Such knowledge is thus embedded in the product experience versus residing only in the minds of specialist humans or fragmented enablement docs.

Although agentic sales is still in its infancy, the enthusiasm for it as a critical tool for product success is boosted by the positive results that early implementations have already generated. At one Fortune 500 enterprise technology company, always-on AI agents help manage sales opportunities by gathering data to address customer concerns across internal and external sources, freeing sales teams to focus on higher-value work such as shaping individual customer solutions and building relationships with customers. This arrangement sped up parts of the sales closing process by eight times. Similarly, at a company that markets customer relationship management (CRM) software, using AI agents for initial sales calls led to 78% more free trial sign-ups and a 25% increase in quality sales prospects—customers on the cusp of making a purchase—for humans to try to close.

Obstacles to Overcome

In a recent discussion of the CPO Council—a network of about 30 (mostly) B2B CPOs formed by BCG and Lobby Capital to share product development and management insights—we found that most CPOs believed that initially AI agents could be extremely effective in supporting GTM teams in converting prospective customers into actual sales (see Exhibit 1).

However, we were somewhat surprised to learn that although interest in AI sales agents ran high, actual implementation was limited. In other words, AI sales is top of mind among CPOs, but skepticism also looms large. Surveying this group, we found that while some companies are testing the technology now—a segment that will grow incrementally over the next 12 months—at-scale implementation is not likely to occur for about three years (see Exhibit 2).

What’s slowing adoption isn’t a lack of enthusiasm about the promise of agentic sales technology; it’s the readiness of company systems to support such a radical change in how sales are conducted—along with the significant risks of getting it wrong. Specific concerns about AI sales agents fall into three primary categories:

• Customer Experience. In the typical approach to agentic sales, companies are weighing use of AI for text-based chats and as voice agents. There are problems with both. Customers may be uncomfortable conversing with a bot—as the CPO of one LLM company put it, “If I suddenly start talking like I’m Daft Punk, it just gets weird.” They also might be put off by the inconvenience of having to type just to talk to a machine. But recently there have also been signs that finely tuned voice agents can minimize friction and thus show promise in increasing sales conversions and reducing GTM costs. As trust in AI grows over time—an outcome of inevitable broader use of GenAI in people’s daily activities—this hesitancy will likely erode organically. Indeed, Amanda Kahlow, founder and CEO of AI company 1mind, found in a recent analysis of her clients that 95% of their prospective customers had had a positive experience with AI agents, adding that “50% to 75% engaged with the AI agents versus only 5% to 20% with a voice or text chatbot.”

• Accuracy and the Impact on Sales. In the AI world, hallucination is frequently raised as a constant danger: How do you guard against a large language model (LLM) generating nonsensical and incorrect output that appears plausible and accurate? Such errors occur when AI misinterprets data, often mistakenly placing an emphasis on information it assigns a higher priority in its knowledge hierarchy. Sometimes these hallucinations are initially prompted by an AI system’s confusion about the terms used by the person it is conversing with. “The meaning of words can be very subjective and vague, so a conversation with a voice agent that doesn’t understand customer responses and lacks context can produce bad results,” says Satish Jayanthi, CTO at a SaaS company helping enterprises manage and transform data.

  While this is a real concern, it is countered by the argument that human sales teams also “hallucinate” (the term “human error” exists for a reason) even if they have the best intentions. Sometimes they simply lack product education; other times, they can’t resist exaggerating to convince the customer to make the purchase. “Often, we see sales reps wanting to get the deal done, not to necessarily communicate the information as accurately and empathetically as possible,” says Kahlow.

• Garbage In, Garbage Out. An antidote to AI hallucinations is stronger language models and richer knowledge libraries that are designed to continually refine the context of words, phrases, and ideas to make them more useful in real-world settings and conversations. To enhance AI sales agent effectiveness in this regard, CPOs and product teams should focus on improving their companies’ product content and documentation debts. “We need to prioritize what content we need to write or rewrite, and how we structure it so AI agents can actually use it. It’s a forcing function to acknowledge that knowledge is the product,” says Tenni Theurer, chief product officer at Spring Free EV, a climate-focused fintech. AI tools can be used to improve content quality, but human involvement is equally critical. Companies should dedicate teams—perhaps temporary—to overhaul and expand product libraries for both existing and upcoming releases.

CPOs and their teams should proactively mitigate these challenges, smoothing the path for AI agents to become active participants in sales activities. For instance, product teams can provide insight and recommendations about which customer segments would be more comfortable in self-service, full AI agent interactions and which would prefer communicating with humans while AI serves invisibly in the background as a support tool. Also, since AI agents are likely to at least begin to become more widespread and essential as early as 2026, CPOs cannot afford to overlook the need to improve the content available to LLMs, reducing hallucinations and enhancing the quality and results of AI-driven interactions.

Steps to Take in the Next 12 Months

As these roadblocks—many of them internal—are addressed, CPOs and sales leaders can start laying the groundwork for integrating AI agents into the rhythm of the company’s sales efforts. Over the next year, a set of targeted actions can help initiate that transition.

Define clear responsibilities. Perhaps the best way to characterize implementing AI sales agents successfully in an organization is as a hybrid sport, a team effort between sales and product teams.

Capture the product vision and experience before you build, but be prepared to alter the narrative. It has always been the job of the product team to define the product and brand vision and to create content from it that supports both sales and customer teams. But as AI agents surface new insights, the way products are characterized and marketed is far more fluid than before. Product content no longer lives in a single sales enablement pdf, but instead is dynamic and evolving based in part on what AI agents learn in customer interactions.

Embrace and talk up your new sales team members—the agents. Viewing the agent as a maturing and ever more intelligent representative for the company’s products is essential for CPOs. With that mindset, CPOs can help diminish resistance to AI agents across a company’s sales, product, and technology teams by demonstrating that agentic sales benefits them by addressing content deficits and creating better sales performance.

Establish the right KPIs. Metrics should be developed that provide product, sales, marketing, and technology teams with clear evidence that their own performance can be enhanced by AI collaboration. KPIs to consider include trial-to-paid conversion rate, sales cycle length, drop-off rate during onboarding, sales pipeline velocity, returns on investments in marketing and web sales activity, and quality and number of sales opportunities.

Decide whether to make or buy. Amid the details of product content management, project leadership, and KPIs lies a broader strategic decision: Should an AI agent be built in-house or procured from a vendor? The right choice depends on several criteria, including how quickly the system must be deployed to sustain or strengthen your market position, how widely AI agents are being adopted across your industry, and the nature of your business model.

In a sales-led organization, off-the-shelf AI agents can quickly support and streamline the sales process, delivering immediate benefits. In contrast, a product-led company may find more value in building its own agent if the product is complex, highly customized, or serves a narrow vertical. Another key consideration is readiness: How prepared are the sales, marketing, technology, and product teams for a major process shift in the way the organization operates?

If that readiness is limited, the best course may be to start small—develop AI agents internally to build knowledge, understanding, acceptance and control—then turn to vendors later to accelerate, scale, or refine performance once the role of the AI agents is clearer and the organization is readier to accept substantial organizational change.

AI agents will be commonplace sooner rather than later. At the very least, they address intractable challenges that companies face—everything from product development and documentation to reaching long-tail and high-touchpoint customers to dynamically shifting a product’s brand, reputation, and presence in the marketplace. AI-based solutions to those fundamental dilemmas are too promising to resist. Given the speed of technology adoption, we believe that in competitive product development and sales environments, product and sales teams must ready their organizations for AI sales—if not deploy it—a critical priority.