We’ve barely entered 2023 and already AI has become a defining term for this particular moment in tech history.
We trust that our audience has been keenly aware of AI for quite some time now. In fact, for those of us working in tech and tech journalism, it’s been nearly impossible to avoid discussion of AI and its many deployments, especially over the course of the last decade.
But we’ve entered a new era now, an era in which the general public is also very much aware of AI implementation. OpenAI, advanced chatbots, and AI image generation systems have caught the attention of casual users, to say the very least, and that attention has also brought with it conversations on the ethical implications of these AI tools.
Meanwhile, watershed tech companies like Microsoft and Adobe have embraced AI on multiple fronts, in some cases integrating AI tools into their SaaS products, and in other cases announcing plans for further integration in the near future.
One of the most compelling integrations we’ve seen is the drastic improvement of Dynamics 365’s search function.
When it comes to enterprise software, there are few companies that can measure up to Microsoft, the perennial tech leader, especially regarding the implementation of emerging and cutting-edge technology.
We expect you’re familiar with the Dynamics 365 suite, and there’s a good chance you’ve tried it firsthand. There are numerous major corporations that rely on the Dynamics 365 suite, including the likes of Coca-Cola, BMW, HP, Lowe’s, and Chevron.
Previously, in spite of its range of highly useful features, Dynamics also included a search function that, compared to search features in other enterprise software, was par for the course but slightly lacking, especially for less-experienced users.
The team behind Dynamics saw an opportunity for improvement, and long story short, the current search function is vastly superior, and AI is making it all possible.
We’re very excited to share with you excerpts from a conversation we had with the man responsible for this new intelligent suggestion system, Srivatsan Balakrishnan.
Balakrishnan is a Senior Software Engineer at Microsoft, where he’s currently focused on AI-fused applications.
Prior to his work with Microsoft, Balakrishnan served as a Software Engineer for Motorola Solutions and Bally Technologies. But today we’ll be focusing exclusively on his Dynamics 365 work and the advantages and effects of this new functionality.
First, let’s make it clear how the intelligent suggestion system has moved things forward.
The new search function is featured prominently in the new Dynamics 365, and perhaps its most significant advantage is that it’s designed to facilitate users. We might call it a human-centric approach.
The previous Dynamics search function felt less accessible to some users. For example, a misspelling could prevent the user from finding what they were actually looking for.
Now, Dynamics’ search function can handle all kinds of misspellings, typos, and even alternate terms and abbreviations for various terms.
Balakrishnan was responsible for developing the APIs for this system, developing the object store cache service, and developing “a hot pass feature that generates feedback for users which can then be used in the dynamic search scope generation algorithm.”
Balakrishnan explained that the latter feature is meant to “improve efficiency for new users and customers onboarded to Dynamics 365 search.”
From start to finish, the overhaul took about six months to complete. Also, Balakrishnan mentioned that there was initially a bit of pushback from some of the partner teams, specifically in regard to the plan for implementation. But with some convincing, they were able to get the system ready for launch.
And all that effort was definitely worth it. Over the course of ten months, this novel implementation of a dynamic search scope generation algorithm led to a 100% improvement in the product’s clickthrough rate.
But why exactly is this search scope generation process so effective? Here’s how Balakrishnan explained the underlying concept.
“Dynamic search scope generation is based on feedback loop data. Specifically, it’s the feedback of user preferences, which are generated in a timely manner as the user interacts with the system. And to be clear, this feedback loop data has no personal information and follow all compliance procedures. CTR improved from 13.5% to 36% as of today. The results really wowed us, and we were happy to know that users were getting the results they wanted.”
But this is just one example of how Machine Learning/AI can be applied to existing applications to drastically improve specific features and overall functionality.
The question is, how will AI continue to shape software in the months and years to come?
We asked Balakrishnan whether AI and Machine Learning could be applied to various other existing features within Dynamics 365, and the short answer is yes.
“For example, natural language processing could be used to update, alter, or even create data without having to write an SQL query or click through a UI. This would also make it possible to use verbal commands to update data.”
Just to clarify, language models are currently in use within Dynamics 365. In fact, language models are what enable the understanding of user queries and the correction of typos that we talked about earlier.
But further in that direction, language processing could also change the ways that users interact with the software, which would be groundbreaking. As Balakrishnan mentioned, even the possibility of verbal interactions would be a major step forward, especially for users who aren’t able to easily use traditional input devices.
As for other potential implementations of AI, the possibilities are nearly endless. That doesn’t mean that it would be simple to develop other implementations, but the possibilities are truly impressive and stand to benefit users in substantial ways.
Something general audiences are starting to understand about AI deployment is that the chatbots and image generation tech that have been grabbing headlines are really just the tip of a much larger iceberg.
The improvements to Dynamics 375 that we’ve been discussing are an excellent example of how AI can provide very practical benefits.
For decades, programmers and designers have strived to improve UX. In the early days of Windows, that meant the introduction of GUIs, which made software much more accessible for users who weren’t especially tech literate.
Obviously there have been leaps and bounds in this area since that time, but even today many programs come with an unspoken expectation that the user understands the basics of how a program “thinks,” and that the user will meet the program halfway, so to speak.
But as AI learns from human users, the need for that kind of compromise is lessened, and the end result is software that feels intuitive and natural to use.
So where do we go from here, and what role will Microsoft play in the advancement and implementation of AI.
While Balakrishnan wasn’t able to share specific details on this front, he definitely feels confident in Microsoft’s dedication to AI and its ability to execute impressive AI implementations.
“Microsoft is very well-situated to utilize cutting-edge technology and AI in particular. Already, the company has started to integrate the OpenAI model in most of its products. I’ve also started work on a project using GPT 3.5 and chatGPT in relation to a new product that’s expected to disrupt current processes and technologies in a big way.”
Before closing out here, it’s important to acknowledge that the current value and utility of AI tech hasn’t come out of nowhere. There have been decades of work behind the scenes to make all this possible.
Even more importantly, the rules of AI deployment haven’t been written yet. As we said at the very start, the jury’s still out when it comes to AI mimicry of original human work, and there will be numerous landmark court rulings on the subject in the years to come.
In the meantime, this is an undeniably exciting moment in software development and tech as a whole.
Special thanks to Mr. Balakrishnan for giving us a look behind the curtain and sharing his insights.