Microsoft Copilots Everywhere, A Bing Chat Facelift, and the Shift to Practical Applications
Generative AI is likely to be sticky and Microsoft is focused on benefits
Our vision is pretty straightforward. We are the copilot company. We believe in a future where there will be a copilot for everyone and everything you do. Microsoft Copilot is that one experience that runs across all our services, understanding your context on the web, on your device, and when you are at work, bringing the right skills to you when you need them…
Copilot will be the new UI that helps us gain access to the world’s knowledge and your organization’s knowledge. But most importantly, it’s your agent that helps you act on that knowledge.
Satya Nadella, Microsoft CEO
Yesterday, Microsoft made a series of announcements highlighting the reach of generative AI throughout the company’s products. It’s all about copilots these days. And the top story is that Bing Chat is now decoupled from Bing and has been renamed Copilot.
The new Copilot has a similar interface to Bing Chat, but because it has a bit more screen real estate on the desktop, you will see more tips about tasks it can perform and won’t be distracted by a traditional search user interface. This is particularly helpful because Bing’s traditional search is overly cluttered.
Microsoft CEO Satya Nadella began the keynote with the observation that generative AI is entering a new phase. For the past year, most discussions have revolved around what is possible. This emphasis is shifting to “the details of product making, deployment, safety, real productivity gains, all the real-world issues.” You can see this everywhere and well beyond the Ignite conference stage. The wonder that accompanied the technology earlier this year has been replaced by wondering how best to implement it.
Evidence of Generative AI’s Impact
Reinforcing this theme were the many case studies of early Copilot users and data confirming productivity gains. Three examples included writing a blog post, drafting meeting notes, and searching for information. Time savings were recorded as 6 minutes, 32 minutes, and 6 minutes, respectively.
When you shift from what’s possible to what should I deploy, features and assurance around benefits are top of mind for enterprise buyers. Microsoft is selling its customers on adding $30 per user per month to access Copilot’s features. It’s also selling access to foundation models through Azure. Some companies need convincing, while others need a justification for proceeding down the path they are already convinced is correct. Enhancing creativity fueled by AI may deliver benefits, but productivity is more easily counted.
Microsoft presented numbers throughout the Ignite Conference keynotes. There were few “speeds and feeds” mentions of product performance. By contrast, there were many data point about employees reporting productivity benefits. While I suggest you take most of the sentiment data about employee self reporting of reduced time to complete everything from managing their email inbox to adding notes to the company CRM, you might want to pay attention to one finding over the others.
Seventy-seven percent of Copilot users said they would not want to give it up. In fact, they would not trade it for a literal free lunch. There are other hard stats, such as sales professionals recording 90 minutes of time savings per week. Those should be taken seriously because a 4% productivity gain in sales would likely be meaningful to most businesses. However, the “would not want to give it up” statistic suggests generative AI will be sticky once it is widely adopted for work. It means that once the capability is installed, chances are low that it will be removed.
Generative AI and Applications
From a product standpoint Microsoft interspersed new announcements with a mix of previously announced product updates:
Microsoft Copilot for Microsoft 365: This became generally available for enterprise users in early November, and the company named BP, Honda, Pfizer, Visa, Accenture, EY, KPMG, and PwC as customers. Additional features include a new administrator dashboard to track usage and benefits and the ability to personalize the solution for a business.
Microsoft Copilot Studio: Like OpenAI’s recent announcements around Assistant API, “Microsoft Copilot Studio is a low-code tool designed to customize Microsoft Copilot for Microsoft 365…for internal or external use.” Notably, Copilot Studio will support GPTs and plugins, the latter of which OpenAI has all but deprecated in favor of the former.
Microsoft Copilot for Service: This announcement signals Microsoft’s move into functional domain templates for creating new copilots. “Copilot for Service includes Microsoft Copilot for Microsoft 365 and helps extend existing contact centers with generative AI. In customer interactions, agents can ask Copilot for Service questions in natural language and receive relevant insights based on data sources from knowledge repositories.” OpenAI has announced a similar templated copilot concept for ChatGPT Enterprise.
Copilot in Microsoft Dynamics 365 Guides: Guides was the wow-factor announcement, which showed industrial workers using HoloLens to complete tasks in a factory. The announcement came with an interesting video but not a full demo. Whereas Apple Vision Pro is for consuming content on the couch, Guides with HoloLens is the Tony Stark accessory for people who make things in the real world. Microsoft says, “Combining the power of generative AI and mixed reality, this copilot helps frontline workers complete complex tasks and resolve issues faster without disrupting workflow.”
Microsoft Copilot for Azure: The company also announced an AI assistant for IT administration. Once again you can see there will be a copilot for every role.
Generative AI Foundations
A significant number of announcements also revolved around foundation models and Azure cloud computing infrastructure.
Model-as-a-Service: Prevously announced at the BUILD conference, Meta’s Llama 2 and models from Mistral will be added. Developers, “can also customize these models with their own data without needing to worry about setting up and managing the GPU infrastructure, helping eliminate complexity.” This will be popular with developers looking to try out open-source models because many are realizing the cost and skills required to set these up on their own is more difficult than expected.
Azure AI Studio: This seems similar to OpenAI’s Assistant API but with more tooling and a connection to several other Microsoft developer tools. “With Azure AI Studio, you can build your own copilots, train your own, or ground other foundational and open models with data that you bring.”
Vector Search: It may be copilots all the way down, but a close runner up these days are RAGs (retrieval augmented generation). These represent the most popular way to connect generative AI with grounded (i.e. truthful and comprehensive data).
GPT-4 Turbo, Vision, DALL-E: The new OpenAI models are available in a public preview through Azure OpenAI Service.
NVIDIA AI Foundry Service: The NVIDIA AI Foundation models, NeMo framework and tools, and DGX Cloud supercomputing are now available through Azure. This sounds similar to but a slightly more comprehensive deployment in Azure than what NVIDIA announced with Google Cloud in August. NVIDIA’s position as the primary supplier of high-performance GPU used to train AI foundation models is offering it the opportunity partner with every large cloud provider on what are presumably favorable terms.
It’s a Copilot Company Now
If you came into Ignite thinking Microsoft is a big company and generative AI is just a part of what it offers, you’d be technically correct. However, you would not have come away from the event with that impression. Generative AI was part of every discussion. Satya Nadella has clearly laid out a vision that it will be part of every product.
The more interesting angle is the choice to call Microsoft a copilot company. Nadella did not choose to call it an AI or generative AI company. Either of those monikers would be equally justified based on what we saw and where the company is headed. Choosing copilot as the central term is strategic.
Copilots are the most compelling manifestation of the generative AI application layer. While OpenAI or Google or Anthropic may claim technical leadership in generative AI, Microsoft is striving for leadership in the application of AI to everyday business problems. That positions Microsoft as closer to the user and closer to the value that generative AI innovation offers.
Microsoft is clearly all-in on the copilot brand and its potential to reshape its products. I do wonder sometimes if they are moving too fast. As they roll out their Copilot features in virtually every product they have, nobody is talking about the elephant in the room, the persistence of hallucinations.
Now Enterprise customers are starting using this daily, at global scale, I seriously expect we're about to see some stories come out from business decisions made on generated reports with small but crucial mistakes.
I could be wrong, of course, I hope I am. Maybe this is actually Microsoft's moment and have they brilliantly anticipate and leveraged the opportunity, thanks to their partnership with OpenAI.