How to get ahead of Microsoft's 345 million other AI users

Microsoft is on the brink of providing 345 million people with Generative AI tools that will forever change the way we work and interact with information.

It may be as impactful as Search, Social Media, and perhaps, the Internet.

Despite the buzz (hype) around ChatGPT-driven time savings, finding the right tools, experimenting with them, and working out their utility has been a time-consuming task so far.

The promised efficiency savings have been elusive in the short term as we all scramble to find the human/AI chemistry that is going to make the difference. We're all still stuck on the AI starting grid, really.

But later this year, the hunt will be over. Microsoft is set to embed them into products we all use daily, and the potential impact on businesses and individuals cannot be understated.

If you've been using Microsoft products recently (who doesn't?) you will have noticed something going on already. You've had a taste of what's to come.

Word is now predicting your sentences, Teams is transcribing your meetings, PowerPoint is offering feedback on your presentation delivery (if you record yourself), Excel is suggesting formulas, and Outlook is providing email response suggestions.

It won’t be too long before every task you can perform in a Microsoft 365 product will come with an option: let AI do it for you.

How we’re getting ready

Since the launch of ChatGPT last October, those willing to invest $20 a month for GPT4 have had access to a host of additional features. Those willing to invest a few pounds a month in one of the thousands of other tools being built on top of this LLM (Large Language Model) might have found a good use case or two with them.

To understand just how many there are, rummage around in here for a while. Or log on to Twitter or LinkedIn for ten minutes. We’ve been trialing a number of AI tools and are pursuing those that provide the functionality we think will end up inside Microsoft Copilot.

Here’s what they offer:

• The Code Interpreter Plugin is found inside ChatGPT4. You can upload spreadsheets and ask it to create charts, provide answers, and look at trends for you. This will be a feature of Excel soon.
• We have been building our own internal chatbots using Zapier’s Interfaces. You connect your information to Generative AI tools before creating content that is generated using the information you provide (rather than from the internet). Word will do this soon.
• Beautiful.ai is a cloud-based presentation tool that creates beautifully-designed slides and allows you to build slide decks with a prompt of an AI tool. PowerPoint is emulating this already.
• Fireflies.ai is a transcription tool that records and transcribes your meetings before allowing you to ask questions about the transcription using its built-in chatbot, Fred.AI. You will be able to do this in Teams by the end of this year.
• Poised records your presentations and provides live and post-presentation feedback on your communication skills. It adds tips on how to be clearer and more impactful and helps you to stop ‘umming’ and ‘erring’ as you speak. This will be in PowerPoint too.
• AskYourPDF allows you to upload PDFs to ChatGPT before asking it to summarise large documents and provide answers to your questions. It provides answers based on what it learned from reading the document rather than providing responses based on the information it was trained on.

Large language models are powerful tools. But when you provide them with the proprietary data and knowledge held inside your Microsoft products or on your server, they might become game-changers. An AI model that can draw from all your internal documents, providing results based on your unique data, could revolutionise marketing teams and enterprises.

Microsoft Copilot’s promo video suggests everyone will soon have the ability to access their company’s knowledge without speaking to anyone or knowing where to look. It might, as the video tantalisingly suggests, allow us to draw on all our information to help write emails, create spreadsheets, and build PowerPoint presentations.

In readiness for this, we have been looking at our workflows, in-depth, at all the tasks and projects that we deliver in high volume, breaking down the process of creation. By doing so we are able to see where AI can enhance what we're doing and where it won't.

The headline of our findings? There are lots of use cases.

Leading from the front

There are three companies leading the way in professional services.?

• Bloomberg recently introduced BloombergGPT. Developed through extensive training on its financial data, this large language model (LLM) every Bloomberg employees can all now leverage its comprehensive understanding of financial language to gain valuable insights, streamline processes, and make informed decisions.
• PwC has formed a global alliance with AI startup Harvey. With the backing of the OpenAI Startup Fund and built on the foundation of OpenAI and Chat GPT technology, the tool means every PwC employee can use AI to enhance various aspects of their legal work and develop their own use cases for tax-related matters.
• KPMG is investing $2 billion in AI and cloud services across its business lines globally over the next five years through a partnership with Microsoft.

Knowledge, information, and data are fast becoming the most valuable assets of any business. Having your data organised, structured, and accessible to the AI systems will be critical, and with no freely available AI, this challenge now applies to all businesses, data-led or not.

What has become clear is that solid processes and access to sufficient quantities of data are now crucial. While tech giants and professional services behemoths have vast amounts of user-contributed data, not all of us do.

But we can get ready by preparing our systems and processes to feel the benefit of this AI technology by better categorising, sorting, and labeling our data. AI can help with this too.

Not so fast

As much as AI has the potential to boost productivity, enthusiasm for realising these benefits must be balanced with other considerations.

• Security. AI's integration into business operations raises security concerns. Protecting sensitive data while leveraging AI's benefits requires robust protocols so clear guidelines on AI use are essential. Complexity might lead to misunderstanding.
• Uncertainty. There is still huge uncertainty surrounding AI's real impact on businesses. Identifying suitable AI applications is challenging, and not all business aspects may benefit from AI and some areas requiring human expertise may see people resisting automation.
• Data isn’t knowledge. Over-reliance on AI also risks losing human understanding and expertise. Data is not just raw information; it is the key to generating knowledge and insights. While AI technologies can process and analyse massive amounts of data, data alone does not equate to knowledge.
• Trust. The success of AI implementation hinges on its ability to accurately and reliably draw from a business's data systems. It requires human intelligence and interpretation to extract meaningful insights and make informed decisions. Therefore, there must be an investment in AI training as well as AI software.

We think the benefits of AI will be felt at a particular point of a workflow. Not across the whole process. It doesn't feel like entire workflows and processes will be replaced, but they will be powered and best leveraged by data.

Smaller businesses in niche industries might face challenges in generating enough data for AI applications but data is the most valuable asset for businesses today. It is the driving force behind AI technologies and holds the key to unlocking innovation and competitive advantage.

Betting on the winners

Microsoft Copilot will likely lead the way and become the default work assistant for most of us but the enterprise tech giant is hedging its bets on the preferred interface with our personal assistants.

Inflection AI recently secured ￡1.3bn of investment from Microsoft and several major players including NVIDIA, Reid Hoffman (co-founder of LinkedIn), Bill Gates, and Eric Schmidt (former Google CEO). This investment is targeted towards the development of their conversational companion, Hey.Pi.

Distinguishing itself from Microsoft’s Copilot, Amazon’s Alexa, Apple’s Siri, and ChatGPT, Pi aims to provide a personalised and emotionally intelligent experience. It certainly feels different from any other tool I've tried so far.

Try it. It's an odd experience.

Many, if not most, will instinctively feel uneasy with this new interface. They may feel uncomfortable with the attempt to replace human interaction with robots. However, didn't many of us feel that way when we first discovered Facebook or Twitter? How long before Microsoft's work or personal AI products become seamlessly integrated parts of our daily lives?

The companies that adapted best to smartphones, the internet, or social media were those that mastered their data. We don't all have to be the pioneers, but considering the rapid pace of significant changes recently, doesn't it make sense to be as prepared as possible?

The excitement around these tools is generated by the novel, human-like chat interface that we've not experienced before. Conversational AI is transforming the way our current products are designed and used. When you look past the novelty, and let's face it, the oddity of these products, you're left with something that has the potential to be immensely useful to us.

Teaching the robots and the humans

As with any software integration, significant work is required to prepare for its implementation. Training both humans and AI is an essential investment.

Distinguishing between hype and utility is currently challenging, and with the promises of 'game-changing' products and technologies being tossed around, now is not the time for major software investments.

As straightforward as conversational AI makes data interrogation, using these tools correctly and safely requires training. Microsoft will undoubtedly simplify the process, but to fully harness its benefits, all 345 million users will require training.

The AI models will also require training to understand your company's processes and systems. Initial disappointments should be expected, as is the case with any new technology or employee. But once the systems are trained, and people learn how to maximise their use, the boost to productivity should be significant.

If search engines enabled us to 'know' things more quickly, these tools will allow us to 'do' things more rapidly, a prospect all 345 million Copilot users will eagerly embrace. But these tools are nothing without sound processes, effective knowledge management, and accurate data capture.

The enduring impact of AI is on the horizon, offering massive opportunities for productivity and efficiency. By understanding our workflows and potential use cases; maintaining a balanced approach (mindful of security and trust considerations); and training both humans and bots, we can prepare for what is coming. And get ahead of a tidal wave of adoption driven by 345m Microsoft users around the world.