AI in Action: How New Copilot Features are Changing Content Management and Productivity

My Copilot Navigator newsletter on LinkedIn has been ramping up this summer – what I[m calling the #SummerOfCopilot. I’m using the blog for a deeper dive into subjects in Copilot Navigator. This is our sixth edition – here’s what we’ve covered so far. (Hint – if you subscribe you won’t miss an issue.)

#SummerOfCopilot Week 6

Week 1: Microsoft Copilot – A Bold Vision for AI Transformation - #SummerOfCopilot

Week 2: AI Co-Workers Arrive: A Week with Microsoft’s Researcher & Analyst Agents

Week 3: Making the Invisible Visible: Tacit Knowledge Meets Copilot 

Week 4: Unlocking Productivity: Copilot, ChatGPT, and AI Agents in Action – plus Copilot training resources

Week 5: From Studio to Control: Building Smarter, Safer Workplaces with Microsoft Copilot

Our next few issues:

Week 7: Six (or Seven) Ideas for the Future of Copilot also to be covered on a special episode of our Polaris podcast next week)

Week 8: Copilot for Leadership Part 1 (CEO, Boards of Directors, CFO, CMO, CRO)

Week 9: Copilot for Leadership Part 2 (CHRO, COO, CIO, CDO, CAIO and CISO)

After that we’re planning to look at real world case studies. Today we’re looiing at roadmap updates' Copilot costs and license management; and SharePoiht Autofill.

Copilot News and Updates

This summer, AI took center stage in the workplace. Microsoft’s rapid release of new Copilot features – and real-world deployments like Barclays’ plan to roll out Copilot to 100,000 employees underscore a simple truth: intelligent productivity tools are no longer a vision of the future; they’re here now.

Business leaders everywhere are asking: Which Copilot innovations will impact us the most? How do we adopt them responsibly? And how do we balance the benefits with cost and security considerations?

In this post, we’ll explore four key themes from Microsoft’s recent “Summer of Copilot” updates and what they mean for your organization:

• Content Management Revolution – Autofill in SharePoint: AI is tackling the age-old headache of organizing and tagging documents, potentially saving countless hours and improving compliance.

• New Copilot Features Boosting Productivity: From an integrated research assistant in Word to proactive email drafts, Copilot is learning new tricks to streamline daily work.

• Voice-Activated AI – The Next Frontier: Voice interfaces promise hands-free convenience, but they also raise important questions about privacy and governance in the office.

• Cost Management & Licensing Flexibility: As AI usage grows, Microsoft is providing tools and models to help you manage costs smartly – so you maintain control and maximize ROI.

Throughout, we’ll keep the discussion practical and people-focused. It’s not just about shiny new tech; it’s about how these changes can make your employees’ work lives better and your business more effective. Let’s dive in.

AI Meets Content Management: SharePoint Autofill Changes the Game

For years, companies have invested in content management systems, only to find that human nature often gets in the way. We know that adding metadata (tags, categories, summaries) to documents is best practice – it makes information easier to find later and supports compliance – but busy employees frequently skip it. The result? Document libraries that are part organized, part digital dumping ground.

You can make fields mandatory (shown with a red asterisk *) – but that imposes a higher usability tax (what I’ve called the ‘cost of the red asterisk’) when people just want to save documents somewhere. SharePoint “attention views” help you find missing metadata, but its still a mostly manual process.

Enter Autofill Columns in SharePoint, an AI feature that might just solve this problem. Announced under the Microsoft Syntex umbrella and made generally available in early 2025, Autofill uses GPT-4–powered AI to automatically read the contents of files and extract key info into library columns. In other words, it does the tedious tagging for you.

How it works: say your team uploads a batch of contract documents to a SharePoint library. With Autofill enabled, the system might examine each contract and pull out the Client Name, the Contract Value (dollar amount), the Contract Date, and even generate a one-line Description or summary. Those values get populated into the library’s metadata fields as if an attentive assistant went through each file with a highlighter. During one demo, Microsoft showed a user simply adding an Autofill column and typing in a prompt: “What is the project deadline in this file?” With one click, the AI scanned every file in the library and filled in that column for each file that had a deadline in its text. If a document didn’t contain an answer, it left that entry blank so you know to follow up.

Why this is a big deal: Effective content management has huge business implications. Think about compliance: if you’re in a regulated industry, being able to swiftly find all documents related to “Project X” or all files that contain personal data can save you in an audit. Or consider productivity: how many hours are lost searching for the right version of a file? By automatically classifying and tagging content, AI makes your knowledge assets more discoverable. One of our financial services clients joked that Autofill is like having an intern who reads 1,000 documents an hour and never gets tired, organizing everything along the way.

Early results: Our team has tried Autofill on a variety of documents (invoices, résumés, project reports) and the accuracy has been impressive out of the box. For example, on a set of breakout presentayions, the AI correctly identified 28 out of 30 conferences with no custom setup. That’s without any training specific to our content. And it’s continually improving – Microsoft has designed these models to learn from feedback. When your users correct a field, that data helps refine future extraction.

Importantly, Microsoft also reduced the cost barrier significantly. Initially, these advanced tagging features were priced at around $0.05 per page analyzed – which could add up fast for large volumes. In March 2025, Microsoft slashed the price to $0.005 (half a cent) per page, and even included a monthly allotment of free pages in many Microsoft 365 plans. This signals how much they want customers to use Autofill. It’s not a luxury add-on; it’s becoming a standard capability. (For the curious, Microsoft’s own documentation explains the pricing change and the free trial allowance.)

Best practices: We advise our clients to start small and measure. Pick a specific library – say, your Contracts library or a Product FAQs library – and enable one or two Autofill columns. Involve the content owners in reviewing the AI’s suggestions for a while. You’ll build trust in the tool’s accuracy and also surface any tweaks needed to your prompts or metadata structure. For instance, if Autofill is summarizing a document, you might learn that asking for a  specific “3-sentence summary” yields better results than a vague “short summary.” Also, decide on a review process: perhaps any AI-generated metadata on certain sensitive documents (legal, HR) should be reviewed by a human before finalizing. This “human in the loop” approach ensures AI remains an aid, not a wildcard. Over time, as confidence builds, you can relax the oversight and let the system run more autonomously.

One more thing – compliance and governance. Automated tagging can actually bolster compliance if used correctly. For example, if Autofill identifies that a document contains a customer’s personal data and tags it as such, your compliance officers can more easily enforce policies on it.  SharePoint also includes a more sophisticated prebuilt model to detect and extract over 90 types of sensitive or personally identifiable information (“PII”) such as tax id numbers or credit card details.

But you’ll want to ensure that the AI itself is being used in line with your policies. Autofill runs within your M365 environment (data isn’t leaving your tenant), which is important for trust. Still, keep an eye on accuracy for critical fields (like classifications that drive retention rules). In short, treat the AI’s output as you would a new junior employee’s work: very helpful, but worth a quick review for important stuff. Over time, as confidence builds, you can let it run with less oversight.

New Copilot Features: Small Updates, Big Impact

While Autofill tackles the back-office task of content organization, Microsoft also rolled out Copilot enhancements that your front-line knowledge workers will feel directly. These aren’t sci-fi, “someday” ideas – they’re practical features designed to streamline daily work.

Let’s look at a few highlights from the July 2025 Copilot update:

• Researcher in Word: Microsoft Word gains a built-in research assistant. Now, when you’re drafting a report or proposal, you can ask Copilot’s Researcher to find relevant information or sources from within Word. For example, you might type, “Give me recent statistics on e-commerce growth in Southeast Asia,” and Copilot’s Researcher will insert a brief answer (with footnoted citations) right into your document or show results in a sidebar. This keeps your team in the flow of writing. No more bouncing between the document and a web browser for every fact-check. (Microsoft provides a support article on how to use Researcher if you want to see more about how it works.) For leaders, this means reports get done faster and with more robust data. It’s easier for your employees to back up their recommendations with evidence when the evidence is just a quick Copilot query away.

• PowerPoint + Organizational Content: Creating presentations can be a slog, especially when hunting for the latest approved images, logos, or slides. Copilot can now integrate with your company’s Organization Asset Library in SharePoint (as well as third-party content sources like Templafy) when helping build PowerPoint decks. Imagine typing “Copilot, create a slide about our Q2 sales growth” and it automatically pulls in the relevant chart from your SharePoint library and an approved image of the product. Our team tested this feature and found a key success factor: having a good template. When we used a “Copilot-optimized” slide template (with clear placeholders and prompts for the kind of content needed), the results were great – slides came together with appropriate graphics and formatting consistent with our brand. When we tried it on a blank slide, it was less structured. So, ensure your templates are up-to-date and consider curating your asset libraries (remove outdated graphics, add descriptive names) to help Copilot help you. The outcome is compelling: employees can generate initial draft decks in minutes, then spend their time refining the message instead of digging through folders for the right images. (If you don’t have an organization asset library set up for images, visit https://learn.microsoft.com/en-us/sharepoint/create-organization-assets-library to take advantage of this feature). And if your company uses a solution like Templafy for templates, good news – Templafy has integrated with Copilot as well (see Templafy’s announcement for details).

• Proactive Email Drafts in Outlook: One of the more futuristic-sounding features is Copilot now suggesting email replies before you even ask. Available in preview, Outlook’s Copilot can pre-draft responses to incoming emails. Think about your own inbox – you probably have some routine requests or recurring questions that you answer regularly. Copilot identifies emails where it can predict a helpful response (for example, a colleague asks for a project update, and Copilot knows the latest status from your documents or previous emails). It then prepares a draft that you’ll see when you open that email. For now, it remains a relatively minor addition. There’s still some uncertainty: will these AI-generated drafts truly streamline communication, or will they end up adding to the clutter and noise of our inboxes? For more nuanced emails, the drafts may need significant editing, and in some cases, the convenience could be offset by the need to sift through extra suggestions. As a leader, you’ll want to set expectations for your team: this tool may accelerate routine work, but it’s not a substitute for thoughtful, sensitive communication.

• Attachments and Context in Copilot Chat: Another subtle but handy update: when you’re using Copilot in a chat or drafting context (say in Microsoft Teams or Outlook), it will now summarize and list out email attachments or linked documents relevant to that conversation. For example, if you ask Copilot in Teams, “Summarize the email thread about Project Zeus,” the response might end with “Attachments included: ProjectPlan.xlsx, BudgetForecast.pdf” – giving you a heads-up that there were files involved. This is useful because it provides everyone in a discussion with visibility into related assets without anyone having to dig through emails. In one of our internal project chats, a team member asked Copilot to recap a long client email; Copilot not only summarized the email but also noted there was an attachment (a slide deck) and what that attachment was. This feature ensures AI-driven summaries don’t accidentally omit the fact that there’s a file you may need to open. It’s a small change, but it drives better awareness – we’ve seen it prompt people to open attachments they might have missed, leading to more informed decisions.

  
  

Each of these features by itself is incremental. But collectively, they indicate a larger trend: Copilot is moving from a reactive tool (you ask, it answers) to a proactive assistant (it anticipates your needs). Microsoft is essentially teaching Copilot to understand context better – be it the context of what you’re writing, the slide you need next, or the emails waiting for a response. For businesses, the benefit comes in the form of efficiency gains that add up. It might be the difference between an employee spending 3 hours on a task vs. 2 hours. Multiply that across tasks and people, and you start seeing serious productivity boosts.

Voice-Activated AI: Opportunity and Responsibility

Perhaps the most buzz-worthy development in Copilot’s roadmap is the introduction of voice-activated Copilot conversations. Microsoft has indicated that by late summer 2025, users will be able to speak a wake command (“Hey Copilot…”) and verbally interact with their Copilot in Microsoft 365. This mirrors the experience we’ve come to enjoy with Siri, Alexa, or Google Assistant in our personal lives, but now applied to our work tools.

Let’s imagine a scenario: You’re in the car between client meetings and you remember you need an update on a project. You could simply ask aloud, “Hey Copilot, what’s the status of the Q3 budget forecast for Project Apollo?” and get a spoken summary pulled from the latest reports, all without pulling out a laptop. Or in the office, you might say, “Hey Copilot, draft a follow-up email to the team about today’s strategy session and list any open questions,” while you walk to your next meeting. The convenience and speed are undeniable. For accessibility and multitasking, this is a game-changer – employees who prefer talking to typing, or those on the move, could interact with data and documents as naturally as having a conversation.

However, with great power comes great responsibility (to quote Spider-Man’s guiding principle). Voice interfaces in a shared workplace raise questions that leaders will need answered:

• How do we ensure security and privacy? If a user can invoke Copilot by voice, how do we prevent someone else’s voice from triggering actions or overhearing answers? In a personal setting, it’s merely inconvenient if my smart speaker answers a question intended for someone else. In a work setting, imagine somebody in the next conference room saying “Hey Copilot, show me the CEO’s calendar for this week,” and an AI agent obliges because it can’t distinguish who asked. The risk of cross-access is real. We’d expect Microsoft to implement user-specific recognition (similar to how some devices recognize individual voices) or perhaps requiring the user to be actively logged in on that device. Clarity is needed here. Until then, we have to approach with caution.

• Noise and context: Offices can be noisy. Will Copilot accidentally get triggered by a snippet of conversation (“…I was saying hey, Copilot is great…”) or pick up the wrong command? The technology likely includes safeguards, but in early stages, there could be hiccups. This is partly a user training issue – people will learn how to use (and not use) the wake word in shared spaces, similar to how we learned not to trigger each other’s voice assistants during conference calls.

• Privacy of output: If you ask Copilot aloud about confidential data, do you want the answer spoken aloud as well? Perhaps not if others are around. The feature might allow on-screen answers instead, or require a headset for certain info. It’s another design detail to watch for. We advise treating voice queries the same way you’d treat discussing sensitive info in a public area – be aware of your surroundings.

What’s encouraging is that Microsoft is listening to these concerns. The company rolls out productivity features in phases, learning from user feedback (and yes, sometimes mistakes) before wide release. As an early adopter or decision-maker, you have the chance to shape the policies around this. We recommend starting with a clear stance in your IT policies: for example, enabling voice features only on devices that are not in open-floorplan areas, or requiring users to wear headsets for voice interactions in shared environments. Couple that with employee education: explain the convenience but also the expectation of using it responsibly.

As a CTO who lived through the era of smartphones suddenly appearing in every meeting, I see parallels. There was initial fear (“Will people record conversations or leak info?”), which we addressed with policies and trust-building, and now phones are an accepted part of work. Voice AI might follow a similar path – a period of adjustment leading to a new norm. If done right, it can indeed boost productivity and inclusivity (imagine the benefit for an employee who has difficulty typing due to an injury or disability – voice could be transformative). The leadership task is to welcome innovation with one hand and set guardrails with the other. That balance is what turns a risky idea into a strategic advantage.

Managing Copilot Costs: New Tools for the Bottom Line

Every new technology in business eventually faces the same scrutiny: What’s the return on investment? Are we spending wisely? With something as potentially transformative as Copilot, it’s crucial to get this part right. Microsoft has introduced a two-pronged approach to help organizations manage Copilot costs: detailed usage analytics and flexible licensing models.

Let’s break down the options and how you might mix and match them:

Licensing Options for Microsoft 365 Copilot:

Option	What It Means	Best For	Key Benefit
Per-User License	Flat $30/user/month for unlimited Copilot use across the suite (Word, Excel, Teams, etc.).	Power users who use AI daily as an integral part of their work (e.g., analysts, content creators, developers).	Predictable cost per user; no worrying about quotas. Encourages heavy usage where AI can drive a lot of value.
Pay-As-You-Go (PAYG)	No upfront license. Copilot features (like Chat and custom agents) are enabled for users, and you pay per AI message or action.	Broad deployment where each user’s usage is light or uncertain (e.g., pilot programs, occasional users, large orgs testing the waters).	Low entry cost – you pay only for what gets used. Great for experimenting and scaling gradually. Allows wider access without a big upfront investment.

With these options, Microsoft acknowledges that AI usage won’t be one-size-fits-all within a company. In many cases, a hybrid approach yields the best value. For example, a large consulting firm might license the core team of proposal writers (who rely on Copilot to draft documents daily) but use PAYG for the wider consulting staff who only need AI assistance occasionally. Over time, if some consultants start to leverage AI more frequently – say they discover how useful Copilot is for research – the firm can convert those users to the flat license model.

Conversely, if a licensed user barely touches Copilot, that license could be reallocated to someone else who will use it more, or the organization might decide to switch that person to PAYG to save money. The key is flexibility. You’re not locked into an all-or-nothing choice and can optimize as you gather data.

Speaking of data, the Copilot usage reports (part of what Microsoft calls the Copilot Control System) are your compass. These reports, accessible in the Microsoft 365 admin center and in Azure, tell you how many Copilot “actions” are happening, the types of prompts being used, and which departments or even which solutions (like a Sales Q&A agent vs. Copilot in Excel) are consuming the most. For example, an IT admin might see that in July, the company had 8,000 Copilot chat sessions, of which 5,000 were from the customer service department’s Q&A agent and 3,000 from general use in Word and Excel. If those numbers are way above expectation, the admin can check the cost implications (all those have a consumption cost if under PAYG). If it’s within a set budget, great. If not, the business might decide to purchase a capacity pack for the Q&A agent (pre-buying a bundle of AI calls at a discount) or train the service team on best practices to reduce unnecessary AI queries.

This level of insight is unprecedented in office software. Think about it – we usually don’t know how many times a user clicked “Spell Check” or used the SUM function in Excel. But for AI features, monitoring usage is important because it ties directly to cost. Microsoft is essentially treating AI power like a utility: you have a meter for it. And just as with electricity or internet bandwidth, good management means watching the meter and using the resource efficiently. (Microsoft even introduced a Message Consumption report in preview to help admins track AI usage and costs.)

One of our retail sector clients decided to set a pilot budget for Copilot usage. They enabled PAYG Copilot for 500 employees but capped the budget at $X per month in Azure. They communicated to the pilot group: “Try Copilot for your tasks – we have a budget for this trial, and here’s how we’ll know if it’s giving us value.” Over a few months, they gathered stats: which teams used it most, which scenarios drove usage (they found many people tried the “draft email” feature, but very few used it in Excel for analysis). We helped them survey those teams to capture qualitative value – did it actually save time, improve work quality, etc.?

Combining that with the hard data of costs, leadership was able to decide on the next phase. They chose to license a subset of users (the ones who got clear value, like the legal team who saved hours on drafting contracts) and keep a larger group on consumption model until they ramp up usage more. This targeted approach kept their spend efficient while expanding Copilot to where it made the most difference.

At Synozur, we’ve written and presented about leading practices in planning a Copilot rollout. In practice, here’s our recommended playbook for cost-effective Copilot adoption:

1. Start with a Pilot and Measurement: Don’t rush to license everyone on day one. Enable Copilot (via PAYG) for a cross-section of your org – different roles, departments, levels of tech-savviness, etc. Give them a proper orientation so they know how to use it. Then use Microsoft’s reports to track usage patterns, and gather feedback from users about outcomes. Look at both quantitative data (e.g., hours saved, tasks accelerated) and qualitative input (user satisfaction, confidence in AI outputs).

2. Identify Power Users vs. Casual Users: The data will likely show clusters – a handful of people might be responsible for a big chunk of usage. Find out what they’re doing (maybe they discovered a killer use case!). These folks might need the full license because they’re pushing the limits of what PAYG would cost. Conversely, you’ll find many people may test Copilot a few times and stop, or use it sparingly for specific needs. Keep them on PAYG for flexibility.

3. Leverage Capacity Packs if Needed: If you’re seeing significant usage in a certain area (e.g., an internal HR Copilot agent is extremely popular), consider buying a capacity pack for that agent or department. The pre-purchase could save on unit costs and signals to the team that they can rely on this tool without worrying about “running up the meter.” It’s akin to moving from a prepaid phone plan to an unlimited plan once you know you’re going to use it a lot.

4. Regularly Review and Adjust: Set a cadence, maybe quarterly, to revisit your Copilot adoption metrics. Who’s using it? Who’s not? Are there new features (like the ones we discussed earlier) that might suddenly drive up usage because they’re useful? Adjust license counts or budgets accordingly. Microsoft’s toolset makes this easier – you can even set up alerts (e.g., notify me if we exceed $Y in Copilot consumption this month). Use those to avoid any unwelcome surprises.

5. Communicate Value Upwards: As you monitor costs, also track the value. Share reports with leadership that don’t just show “We spent $5,000 on Copilot this quarter,” but also “Copilot helped the marketing team produce 30% more content, which led to 5 more campaigns launched” or “the engineering team resolved support tickets 20% faster by using Copilot for troubleshooting.” Tie it to outcomes. This will help defend the investment and possibly secure more budget if the ROI is clear.

In summary, Microsoft’s approach to Copilot pricing and management reflects an understanding that AI adoption needs to be sustainable. They know companies will start small  By giving you granular control and visibility, they’re empowering you to make those decisions with confidence. And from our perspective, organizations that follow a disciplined, data-driven adoption path are seeing the best results and the least waste.

Conclusion: Embrace Innovation, One Step at a Time

A common thread through all these changes is that technology is only half the story. The other half is how we guide our people and shape our processes to make the most of it.

A few closing thoughts for business leaders and decision-makers embarking on this journey:

• Keep People at the Center: Introduce new AI features as empowering tools, not replacements. Offer training and listen to feedback—when our team realized Copilot was for support, not scrutiny, trust quickly followed.

• Champion Responsible Use: Remind everyone that while Copilot can draft, humans own the outcome. Celebrate creative uses and openly discuss any hiccups; this builds confidence and demystifies AI.

• Align AI with Goals: Don’t just deploy Copilot for its own sake—tie its features to real business needs, like faster proposals or smarter knowledge sharing. When people see AI as a shortcut to their goals, excitement follows.

• Iterate and Evolve: Adopt an agile approach—pilot, learn, and improve as Copilot’s capabilities grow. Today’s features are just the start; keep your organization curious and ready for more.

At Synozur, our mission is to make the desirable achievable for our clients. This summer’s advancements in AI are making many “desirable” improvements in productivity and insight truly achievable. We’re here to help you navigate the intricacies – from technical deployment to change management – so that these technologies resonate with your people and drive meaningful results for your organization. Learn more about our AI solutions here on our website.

If you found these insights useful and want to explore any of them further, let’s connect. We’re passionate about this stuff, and we’re on the journey with you.

Thank you for reading!