Say you’re a giant company that’s heard about a fancy “new” technology called artificial intelligence and you’re interested in adding some cutting-edge data crunching muscle to your business.
Contrary to what the artificial intelligence-hype cycle might suggest, just adding popular buzzwords like “machine learning” to your vernacular isn’t as easy as hooking a smartphone to a laptop.
At the Machine Learning and the Market for Intelligence conference this week put on by the Rotman School of Management at the University of Toronto, several founders behind artificial intelligence startups that have been acquired by industry heavyweights like Salesforce.com (crm), Uber, and Amazon (amzn) shared lessons they’ve learned since joining the big-time corporate world.
Get Data Sheet, Fortune’s technology newsletter.
Richard Socher, the founder of the A.I. startup MetaMind that was swallowed by Salesforce in April, explained on a panel what he’s learned since joining the cloud software giant and becoming its chief scientist.
Socher said he was pleased with the research that his small team worked on with two types of A.I. techniques called computer vision, in which software can learn to recognize images in pictures, and natural language processing, in which software learns to recognize text.
But to really push the limits of his company’s technology, he explained, they had to “scale up in sales and marketing, and what better than to join Salesforce.”
Salesforce didn’t just throw Socher into some sort of middle management purgatory where he would spend his days writing reports to higher-ups, or so he claims. Instead, the company appointed him chief scientist where he oversees the research and development initiatives of Salesforce, a much bigger business consisting of 24,000 employees compared to his much smaller startup. He gets to observe Salesforce CEO Marc Benioff run the company to learn about management and how to integrate various A.I. technologies as part of the Salesforce’s Einstein initiative into the cloud giant’s core customer relations management software and related services.
“It has been mostly humbling, but exciting,” Socher remarked.
Likewise for Lior Ron, the founder of the autonomous truck and transportation startup Otto, which was acquired rather quickly by ride-sharing startup Uber for a whopping $680 million in August, Otto wants to “unleash the benefits of self-driving technology sooner rather than later.”
Ron said that “by definition, we were impatient,” and when he started talking with Uber, he felt Uber would let his company maintain its focus on self-driving cargo and speed up its technological and commercial developments. For example, at Uber, he could improve on the trucks’ self-driving features by having access to the enormous amounts of driving data Uber collects to refine it’s A.I. technology.
“Uber is driving over a billion miles every month,” Ron said, explaining accessing that driving data and having the ability “to visualize every Uber car is an advantage.”
Additionally, Ron can leverage Uber’s existing commercial distribution network while still retaining autonomy over his own autonomous truck delivery service. He said that being associated with Uber helps him land deals with companies like Anheuser-Busch, such as one recent deal in which his team drove an autonomous 18-wheeler for 120 miles through Colorado transporting many crates of beer.
“We just accomplished that last week,” he said. [We’re] just moving faster and [remain] independent.”
For more about artificial intelligence, watch:
For William Tunstall Pedoe, the founder of an A.I. startup called Evi acquired by Amazon in 2012, having some semblance of independence from the Amazon was important considering he lives in the United Kingdom. In his case, Amazon gave William the freedom to not have to move to Seattle and would not break up his team of workers.
In a follow-up email to Fortune, Pedoe explained how Evi is now an Amazon subsidiary that employs his team in Cambridge, which is now a “large Amazon development center,” where Seattle Amazon employees also occasionally visit and vice-versa. Pedoe wrote that Amazon was “willing to keep the team in place and build and invest there rather than relocating some or all of the team to a place they already had offices.”
His company’s technology ended up being used to help power Amazon’s digital assistant Alexa and home hub Echo. Amazon now has “teams all over the world working on Alexa,” Pedoe said, which forces the employees to work under time-zone challenges as well as make many trips to the company’s Seattle headquarters.
Perhaps after a few years, the acquisition honeymoon period will end, and life for these former startup founders might become more bureaucratic.
But for Pedoe, things seem to be moving well so far, admitting his startup—by itself—couldn’t have built a chat-interface or voice service like the Echo.
“That required Amazon’s magic, and we were delighted to be a part of that,” Pedoe said. “That made it a success.”
Story updated Friday 3:40 PM PST with additional comments from Pedoe