How Generative AI Will Revamp Go-To-Market

No, it won’t obsolete your creative professionals. But marketing, sales, and customer service will never be the same.

When large language models began leaping forward in capability in 2021, Elaine Zelby knew what she needed to do: start a company to slash the wasted time and money that goes into building marketing campaigns. “I’d spent eight years leading B2B go-to-market teams, so I knew how much repetitive and painful work was involved,” she says. With generative AI on the near horizon, “I couldn’t let go of the idea that we could build a panacea for people doing the jobs I used to do.”

Zelby and two co-founders are chasing that bold dream with Tofu. The company has developed a platform that helps companies scale targeted campaigns, automatically creating a wide variety of content, personalized for each customer. “We ingest your company data to create an artificial intelligence knowledge graph, which lets us generate hundreds or thousands of hyper-personalized emails, landing pages, white papers, e-books—whatever you need,” she says. “And everything is on brand and on message, 100% of the time.”

We believe in that vision as well. Even if generative AI never becomes a panacea, it will have a broad, structural impact on the three main functions that make up a go-to-market (GTM) strategy: marketing, sales, and customer service. Even as of March 2023, 73% of companies polled by Statista were using generative AI 1. Change is coming fast.

So how will tech companies need to overhaul their operations to ride the coming generative AI wave? We asked two NEA partners with deep operating and investing expertise in GTM for their views. Hilarie Koplow-McAdams, a former president of Salesforce’s sales operation, and Vanessa Larco, a veteran of Box and Twilio, share five predictions about how generative ai will change the rules of GTM.

1. Get to the aha moment faster.

The potential for generative AI to generate new content across marketing channels, from daily emails and slide decks to blogs and ad copy, has the marketing industry salivating. While that may become increasingly possible in the future, Larco believes that for now, the most productive use of the technology is to streamline the gathering of information given the countless hours currently spent looking at user data and usage data reports, transcripts of interviews with customers, and reading industry news.

Large language models (LLMs) are perfectly suited to handle this time-consuming work in nanoseconds, potentially saving marketing professionals five hours a week, according to a study by Salesforce. And startups such as Orby AI are developing ways to automate other mundane, time-consuming administrative processes. More efficiency gains are on the horizon, says Larco. Rather than force salespeople to record new orders in the CRM system, send it to legal for approval, and provide a Docusign signature, Orby’s software would handle these tasks behind the scenes.

All of the time saved on research and administrative processes leaves more time for a far more valuable activity: thinking. “The greatest value is in making your people more effective in the human-centered work they excel at,” says Koplow-McAdams.

Larco agrees. “Today, people probably spend 95% of their time figuring out what is happening, rather than what to do about it,” she says. “Maybe we can make 20% or 40% of their time available for high-level thinking, which is what humans are uniquely good at.”

2. Reinvent outbound sales strategies with mass personalization.

Every sales team and organization struggles with how to do outbound sales. The usual approach is to hire young, inexperienced sales development representatives (SDRs) to make cold calls (and send cold emails, texts, and other forms of outreach). But with more enthusiasm than experience, the process is not very effective. According to Operatix, the average SDR will reach out to more than 1,400 leads a month via phone or email, all to land just a dozen meetings. Harder to calculate is how many of the other 1,388 leads were turned off by the process.

Even in these early days of generative AI, well-trained LLMs are already improving on this “spray and pray” sales strategy. They don’t misspell names and are always up to date on the industry news. What’s more, they stick to the scripts that were carefully prepared to reflect the company’s marketing messages throughout the sales cycle. “It’s often hard to get SDRs to stick to the script,” says Koplow-McAdams. “With generative AI, you can generate text or video content highly personalized to the customer that’s within a standard of quality that somebody in product marketing has blessed and without deviation. That’s huge.”

It won’t be long before companies can move beyond just text, says Larco. She says NEA is looking at one startup that can automatically personalize a marketing video by altering a speaker’s mouth movement so it appears he’s saying the recipient’s name. “It looks like I made a hundred different personalized videos, but I just made one,” Larco notes.

These technologies will give go-to-market leaders new options. Generative artificial intelligence can be used to train or assist promising young employees, or it could be used to allow startups to grow more efficiently. “The No. 1 rule for startups these days is to spend less, have fewer headcount, and fewer tools,” says Zelby. “That's the name of the game in generative AI: help your teams do more with less.”

3. Spend more of your time asking the right questions.

Research firm McKinsey predicts that generative AI will give marketers a 10% productivity lift, leading to $463 billion in annual savings. But what will they do with the extra hour or so a day?

One option is to simply cut costs. Another is to crank up the amount of content you create. But Larco suggests putting the gifted hours into further refining your generative AI skills, like prompt engineering.

“It’s about doing all the things there’s never time for,” when marketers have to spend so much time on repetitive, routine tasks, says Larco. “We can take the time to look at what we learned from the last campaign for this product, and to set up the right analytics to leverage the content you create. It’s about taking the time to closely analyze customer feedback and do something with it.”

4. Spend more where it counts: on data.

Generative AI can eliminate some costs, provided it’s used effectively. And that will require additional spending. “You’re going to have to invest more in data to be able to leverage it properly, so heads up: budget for it,” Larco advises.

The first step is to build a large language model that makes the most of your company’s unique knowledge base. Relying only on commercially available LLMs that get their data from the public internet isn’t enough. “The winners will bring together more data,” says Koplow-McAdams. “You want a combination of first-party data, synthetic data you can create, and data you can use to enrich it. You have to really think carefully about how much data you need to inform your model.”

This isn’t cheap at present. By one estimate, the cost of building an LLM is between $100,000 and $500,000. But costs can vary widely. For starters, companies will need to capture, clean, and organize their data. Dealing with so-called “dirty data”—outdated, unstructured, or simply incorrect—can cost marketing companies as much as 38% of their budget in inaccurate targeting, lost leads, and lost customers, according to Treasure Data research. And there’s the cost of licensing third-party data to supplement proprietary internal data.

On the other hand, costs are likely to fall due to competition among foundation-model providers and other vendors. Improvements in tools for fine-tuning LLMs and for making prompt-engineering easy for even non-techies will also help drive down prices. So time spent getting your data in shape will be well worth it.

5. Move quickly away from seats to usage-based pricing.

To be blunt about it, while generative AI can and probably will lead to job growth for tech companies, its fundamental appeal is that it lets you do more with less. That includes people. But most software products are priced based on the number of seats being sold to target customers, meaning the number of people who are allowed to use it. If a main benefit of generative AI is to let companies get more done with fewer people, selling seats is self-defeating. That’s why this model is becoming outdated.

Instead, companies using AI should move to “usage-based” pricing as quickly as they can. In 2022, 46% of all SaaS companies had already adopted some form of usage-based pricing, with another 15% actively use-testing the model.

“We believe this is the future,” Larco says. “If businesses sell software that automates customer support and the AI handles 100 transactions in a tenth the time it would take a human customer support agent, they don’t need that seat. Charging per transaction better aligns the company with its customer base because they only get paid when they deliver value. Accounts are more likely to expand as a result, ultimately boosting net dollar retention.” ”

Of course, seat-based pricing will still make sense in some instances, so companies need to update their billing architecture and infrastructure to handle both of these approaches and have the flexibility to add others that may develop. A hybrid pricing model will let you maintain a subscription model for scenarios where transaction-based billing is just too noisy or not aligned with delivering customer value, but establish usage-based pricing for anything new.

“In the short term, a hybrid approach can offer a good compromise for both the buyer and the startup to ensure stable revenue,” says Larco, adding that it will pay off for go-to-market firms as well. “Any revenue model that aligns you closely with your customer is better.”