Salesforce is crowdsourcing its AI roadmap with customers

Artificial intelligence continues to advance at a rapid pace, forcing companies to develop and release new products faster than ever before or risk becoming irrelevant to faster-moving competitors.

At Salesforce, we believe we’ve found a strategy to keep up, even if it’s not clear where AI is going next. The customer management software giant is crowdsourcing its AI roadmap in real time.

Salesforce isn’t the only company that works closely with customers for feedback on its products. However, this is noteworthy given the size of the company, the pace of new product launches and modifications to existing products, and the granularity of these relationships. These are not annual or quarterly discussions. Salesforce meets with some customers as often as once a week.

“Our 18,000 customers are a powerhouse of information, and we have a wealth of information that we really need to achieve customer success,” Jayesh Govindarajan, executive vice president of Salesforce AI, told TechCrunch in a recent interview. “The stack we built resonated with these customers. Over time, we can make the context better. As the context improves and the LLM improves, the agent system becomes more and more fully autonomous. This is a long-term innovation trajectory, and we intend to invest in it.”

Salesforce was one of the first companies to launch AI agent management software in late 2024, before agent AI started dominating headlines the following year. The company has since doubled its revenue and continues to release new products for voice AI and Slack at a rapid pace.

At Salesforce, we credit our customers for the speed of our product releases. The company told TechCrunch that by putting customers in control, it can build a roadmap for AI products that can quickly respond to where AI technology is heading.

When large-scale language models were introduced, companies naturally wanted to jump on the technology, but they didn’t have the last-mile technology needed to fully utilize LLM, Muralidhar Krishnaprasad, president and chief technology officer of Salesforce Engineering, told TechCrunch.

Govindarajan said the need for last-mile technology led Salesforce to launch agent management platform Agentforce.

From there, the company developed a bottom-up strategy based on themes such as agent context, observability, and deterministic control rather than specific product timelines. This approach uses direct feedback from a rotating group of customers to build the product assuming other companies have similar needs.

Customer sitting in the driver’s seat

“The innovations we bring are a direct result of our work with a huge number of customers and categorizing the problems they see in the real world,” Govindarajan said. “Then you break it down and ask yourself, which things can the LLM layer solve and which can’t? And for the things that the LLM layer can’t solve, you have to build those agent operating system components around the LLM to be able to do that.”

By working closely with customers’ engineering teams, Salesforce can quickly resolve issues before technology evolves.

“You can’t wait three months or six months to get feedback and then consider another six months of work,” Krishnaprasad said. “We’re literally responding to it on a week-to-week, month-to-month basis. That’s been a big change. We’re now pushing code pretty fast, and we have different kinds of gates in place to try out new features and get early feedback before we release them widely. So these are all changes that we’ve had to make to accommodate this rapid change in this environment.”

Engine, a travel management platform, is one of the companies in Salesforce’s customer feedback loop. And it’s not a casual relationship. Engine founder and CEO Elia Wallen said the company’s operations team meets weekly with Salesforce.

This partnership will give Engine access to AI tools prior to release. Warren said this access allows Engine to stay competitive and get more value from these tools.

Relationships go both ways.

Wallen said he has seen feedback from Engine implemented into Salesforce tools. For example, Warren said he instructed an AI voice agent to book a hotel in Chicago. He thought the audio and dialogue felt a little unnatural and shared that with Salesforce. Shortly after, the agent was changed and the company’s A/B tests started showing better results.

“If we have people who can actually help us choose and build the products that we need, they’ll be able to help us better and truly understand our problems and how to solve them,” Warren said. “For us, it’s great to actually be invited to something like that, because we get to influence the product.”

This strategy will enable the company to expand its user-built solutions and workflows to a broader customer base.

Federal Credit Union PenFed was able to slim down its technology stack by working closely with Salesforce, Shree Reddy, the company’s chief innovation officer and executive vice president, told TechCrunch.

“We’re investing time and energy into more strategic platforms and obviously spending more time on this relationship,” Reddy said of Salesforce. “This investment has delivered positive results in terms of strengthening our mutually impactful partnership, and we believe it provides the greatest mutual value-add for both organizations.”

Reddy said PenFed developed its own IT service management (ITSM) workflow using Agentforce’s existing tools and agents, and it worked well for the company. Seeing its success, Salesforce was able to roll out the tool to a broader platform for use by other companies.

The drawback to this approach is that it relies on the classic service sentiment that the customer is always right. Many companies are still exploring what role AI will play in their business, and while many have yet to see the value of the technology, Salesforce hopes they will. As a result, it may not be the best source of information for long-term product development.

Additionally, a willingness to test and preview technology in beta now does not necessarily translate into long-term usage habits or future software contracts.

Be your own biggest user

The company also takes this bottom-up approach internally. Govindarajan said Salesforce employees are the biggest users of the company’s AI tools.

The company also relocated its workforce and resources at the beginning of the AI ​​boom. When ChatGPT was released, Salesforce moved teams and resources to create a new AI team. This is a strategy the company has used successfully in various waves of innovation in the past, Krishnaprasad said.

“Technology changes, so you never know what’s going to come out a month from now,” he said. “We’re going to adapt. And that’s what we’ve been doing all last year. If you think about it, if you look back a year and a half ago, agent wasn’t even a terminology. And we’ve had to adapt to that. We’ve had to adapt to all the advancements, and we’ve had to adapt to our customers.”