This post was originally published on this site.
As a trailblazer in generative AI for software development, Tabnine is helping to drive the sea change created by AI technology. The company launched its first AI code assistant in 2018 and has been on a mission to accelerate and simplify the software development life cycle ever since.
The company has been navigating the explosion in AI tools over the last few years, managing growth and product development. It is the originator of the AI code assistant category and is number two in the group, right behind Microsoft Copilot.
“Tabnine was basically supposed to be a bootstrap dev tools company, says Peter Guagenti, President.” And then all of a sudden, the world woke up and realized AI was real, and our demand went up times 100. We’re only in our sixth quarter of selling the Enterprise product, and we’re already doing incredibly well.”
Standing out from the crowd
While AI is seemingly ubiquitous, with more tools and applications appearing all the time, it is still in its nascent stages.
“I think it’s really important to say the AI space is cluttered from a who got funded perspective,” explains Guagenti. “It’s actually not cluttered from a who has code and who has actual agents and applications perspective. I think there’s really still only a handful of us who are actually operating and doing this at scale.”
“I think it’s really important to say the AI space is cluttered from a who got funded perspective.”
Peter Guagenti, President, Tabnine
Amongst the churn of this new technology Tabnine has set out to differentiate itself in a number of ways. “The reason why people pick us is the three Ps: privacy, personalization and protection,” says Guagenti.
Privacy is a top concern for many companies looking to adopt AI. To address this, Tabnine is built to protect the code and data of the companies who use it. Information is not stored by Tabnine and it is not used to train its AI models. Tabnine also prioritizes intellectual property protection, exclusively training its models on permissive sources.
Tabnine’s AI tools are also personalized to a user’s project, requirements, codebase, and more.
According to Guagenti, the personalization side is where things really get interesting.
“We’re looking at opportunities where it’s really about helping individual, large enterprises. And for them, the AI code assistant can’t behave like a software engineer off the street. It has to behave like your employee of the month. It has to know the ins and outs of your code, your processes, your procedures, your standards and practices.”
“You’re AI code assistant has to behave like your employee of the month and know the ins and outs of your code, your processes, your standards and practices.”
Peter Guagenti, President, Tabnine
AI boots up
AI adoption is moving at different speeds in different industries. For software developers, who have carefully established processes that guide their work, knowing where and how to smoothly integrate AI can be a challenge.
“What we see with the largest customers we work with is they’re not adopting AI all at once,” Guagenti explains. “They’re picking a cohort, seeing how that changes their processes, adapting their processes, then adding another cohort.”
“Customers we work with are not adopting AI all at once. They’re picking a cohort, seeing how that changes their processes, then adding another cohort.”
Peter Guagenti, President, Tabnine
Another thing to keep in mind when considering how AI will be integrated is how it can complement and even supercharge existing ways of working. Software development and coding is not done in bubble – coders have always turned to their colleagues and resources like Stack Overflow to ask questions and work through problems. Tabnine’s tools can be a stand in for such resources, and one that can give considered and customized answers.
Tools from Tabnine can also help remove some of the burden of the more tedious parts of software development. For instance, it can automate testing for developers, leaving them to focus on the more creative aspects of their work.
There is also an anxiety that AI will take jobs, which should be acknowledged as employees see AI become more prevalent in their workplaces. Though Guagenti has a different vision for how AI will shape the workforce:
“What AI is going to do is it’s going to increase our velocity and then up-level our work. I really believe that, but I think you need to acknowledge that people are concerned, and then make sure that as you’re rolling these tools out, they’re part of the process.”
The next evolution of AI
As a technology, AI has already travelled a huge distance in a short amount of time. Guagenti reflects. “What was considered ‘state of the art’ a little over two years ago was auto complete as you type. State of the art 18 months ago was a chat agent. State of the art now is agents that do whole tasks.”
Tabnine is continuing to expand the capabilities of AI agents, and this will be one area of focus for the near term. Each stage of the software development lifecycle, from planning through code generation, through testing, refactoring, fixes, and outages will have an agent that provide answers and solutions in a thoughtful and specific way. These discreet agents are the pieces in the journey towards building an AI software engineer.
Another step forward will be with AI code assistants. “Being able to have an interaction with the AI code assistant, where it’s iterating on the actual feature and function. We’re already seeing this. We’ve got customers who are using Tabnine to build really simple applications, then saying, make it red, make it blue, add this feature,” Guagenti remarks.
Over the past few decades, there have been successive waves of transformational technology – the internet, open-source technology, and cloud computing. Now is another pivotal moment with AI technology. “Here’s my advice. It is not yet another technology wave,” says Guagenti. “All of the technology work we’ve done to date has been deterministic software. Even when we did digital transformation, we took paper processes. We made them a lot more adaptable and approachable and easy to use, but we took the same decision-making process. That is not what AI is. AI is a brain. I think technology executives have to really embrace that. What they know and what worked in the past is not going to work going forward. It is a different structure. It’s a different way of working.”