UNLOCKING THE POWER

OF GENERATIVE AI WITH
MICROSOFT 365
 Table of Contents
Introduction..........................................................................................3
Executive Summary...............................................................................4
Organizational knowledge is hidden in document repositories...............5
How new knowledge gets created with Microsoft 365.............................................5
Why the "illusion of easy" leads to problems downstream......................................6
What role do content processes play in creating and using knowledge?................7
We recognize the challenges, but fail to find a solution.........................8
End-user friction and the path of least resistance...................................................8
The inevitability: The build-versus-buy analysis.......................................................9
Considering the whole IT landscape..........................................................................10
Generative AI alone won't solve the problem...........................................................10
Success means addressing three essential dimensions........................12
Focusing on the essentials for business impact......................................................12
Performing due diligence in choosing the right technology....................................13
Generative AI: Getting your people on board............................................................14
Conclusion............................................................................................15
About the Author..................................................................................16

© 2024 M-Files. All rights reserved.

INTRODUCTION
The rapid emergence of Generative AI in 2023 has
awakened in businesses the need to seriously
consider their content strategy and data governance.

Is Generative AI a solution to our content chaos?

Can we tap into the value of Generative AI, even
though we may have overlooked how documents
and data are governed? How far can general
purpose office productivity applications —such as
Microsoft 365’s SharePoint, Teams, OneDrive and
now Copilot—take us?

This paper will discuss the challenges organizations

face with disorganized content as well as provide
insights on solving those challenges as organizations
seek to scale new heights of productivity with
Generative AI.

© 2024 M-Files. All rights reserved.

3
 EXECUTIVE SUMMARY
Most organizations use general-purpose office
productivity suites such as Microsoft 365 in their
daily operations. These suites provide tools for
storing and collaborating on content produced
on those platforms.

These tools focus on authoring and collaborating.

In addition, governance and control require rigor
to achieve and maintain. Most organizations
recognize this information challenge, which may
cause a loss of employee productivity and
increased compliance risks. There’s several
ways to address these challenges—some
more successful than others.

The promise of Generative AI could level up the

potential value of organizational knowledge
hidden in documents, emails, and conversations.
An increasing number of organizations are
taking on the challenge of solving this state of
content chaos. Successful organizations are
addressing the business challenge, the technology
challenge, and the people challenge in order to
unlock the full value of Generative AI.

© 2024 M-Files. All rights reserved.

4
ORGANIZATIONAL KNOWLEDGE IS
HIDDEN IN DOCUMENT REPOSITORIES
How new knowledge is created with Microsoft 365
Microsoft Office 365 is a leading office other use case. As a result, few organizations
productivity suite deploying tools such as globally don’t have SharePoint or decide not
Word, PowerPoint, and Excel. Microsoft to use it at all.
renamed this package strike Office 365 when
migrating these tools to the cloud and started What kind of content gets created with
marketing them as a SaaS subscription. Microsoft 365? Perhaps a better question
Subsequently, Microsoft gradually bundled is, what kind of content is NOT created with
most of its existing software—SharePoint and these office tools? Leaders, managers, and
Dynamics—as well as new software—Teams employees at all levels look at numbers in
and Power BI—into the same cloud subscription Excel spreadsheets, read reports in Word,
and renamed it Microsoft 365. prepare presentations with PowerPoint, send
emails with Outlook, and collaborate in Teams.
Today, most organizations globally use Microsoft Specialized applications are used across many
Word, PowerPoint, and Excel. When a user processes, but eventually the output of those
subscribes to these tools, they automatically gain applications will be exported into common
access to OneDrive, Teams, and SharePoint for office formats for others to consume. Even if
content storage and collaboration. the origin of data is outside the organization
(such as from a customer, partner, or from
Before Microsoft started bundling office the Internet) that data is usually stored with
productivity tools, many companies chose the rest of related information—for example
SharePoint as their intranet platform or for some in a Teams project folder.

© 2024 M-Files. All rights reserved.

5
Why the “illusion of easy” leads to problems downstream
When companies subscribe to Microsoft 365, What’s the root cause of this problem?
getting started is easy. End users can start Why do organizations lose productivity
authoring documents and saving them to with a productivity suite?
their personal OneDrive. Similarly, when a
group of knowledge workers start working Simply put, Microsoft 365 is only a collection
together on a project, setting up a new of tools—applying these tools to solve business
workspace and folder structure in Teams problems is the responsibility of the user
only takes a few minutes or, the IT organization. That is, Microsoft 365 doesn’t
department has already established a implement any specific business processes
central SharePoint for common documents on behalf of the user organization. It’s a
such as employee handbooks. general-purpose toolset. While the Microsoft
365 toolset is easy to use on an individual and
Things run smoothly at first—until the company a team level, the "illusion of easy" disappears
reaches a certain size threshold, hires more at the organizational level, especially among
new knowledge workers, or simply accumulates more regulated industries.
a critical mass of information spanning across
tens, hundreds, or even thousands of data silos.
The end result?
OneDrive, SharePoint sites, and libraries,
Teams channels, folders, or email attachments
Organizations possess a storehouse of
now store this disparate information. Lack
valuable insight hidden within documents.
of an information-management plan or the
Knowledge workers may not know such a
inability to enforce information governance
storehouse of actionable data exists and they
starts to take a heavy toll.
can’t access it. In short, they’re accumulating
a vast amount of intellectual capital into their
The result? SharePoint libraries, but fail to capitalize
and leverage the knowledge hidden inside
Daily work becomes chaotic. Finding the right those documents.
information is difficult, and quickly leads to
rework, duplication, version-control issues,
and overall frustration. In addition, with
varying access-rights protocols across these
systems, it’s virtually impossible to know who
has access to the correct information or how
compliance of regulations or internal policies
is maintained, resulting in skyrocketing
compliance risks.

© 2024 M-Files. All rights reserved.

6
What role do content processes play in creating
and leveraging knowledge?
The challenges outlined above relate to the At the end of the lifecycle, a document may
way content is created and consumed. Modern even be frozen as an immutable record that
work is rarely performed in a vacuum but, acts as evidence of a business transaction.
rather, in a team of multiple contributors. Any
given document has a specific lifecycle, with Optimally, workflows travel seamlessly, with
various knowledge workers editing, reviewing, every knowledge worker knowing what to do
and consuming the content of the document and when to do it, easily finding the information
across its lifecycle. they need and "passing the ball" to the next
person in the process. When considering the
Usually, the content processes also change challenge of disorganized data and applying
form during the document’s lifecycle, starting the additional layer of processes, we begin to
from highly collaborative processes—where see why organizations struggle.
people might even co-author the document—
towards increasing control where every
change is carefully reviewed and vetted.

© 2024 M-Files. All rights reserved.

7
WE RECOGNIZE THE CHALLENGES,
BUT FAIL TO FIND A SOLUTION
End-user friction and the path of least resistance
Organizations deploy a diversity of knowledge system, process, or change in daily tasks
workers with a variety of backgrounds, may fail if the knowledge workers don’t see
competencies, and interests. Not everyone is a personal benefit and resist adopting it.
a technology enthusiast. Not everyone enjoys
learning new tools or eagerly changes their Selling change is hard. Many organizations
way of working as new technologies arise. give up and follow the path of least resistance.
The attitude? "Let the end users do what they
If the organization starts with minimal or do and try to figure out ways to sort out the
no information governance, adding new issues without our involvement." That’s one of
rules, policies, or mandatory steps in daily the prime reasons organization seek to leve-
knowledge worker workflows can be easily rage the capabilities of Microsoft 365—not
perceived as detrimental to productivity. only for creating and collaborating, but also
As a result, knowledge workers in charge for governance and control. As we’ll discuss in
of producing business results may quickly the next chapter, the other reason is IT cost.
shut down the process. In general, any new

© 2024 M-Files. All rights reserved.

8
The inevitability: The build-versus-buy analysis
For any given business challenge, companies • Does the company allow anyone to create
must balance between the need and available new Teams groups?
solutions. There’s always the option to build • Can OneDrive be used to share content?
a solution to an exact need or buy existing
software that’s close enough. Most enterprise • How can access rights, term libraries, and
software is configurable and can be adapted retention policies be applied across systems?
to the given need—within limits. • Should Power Automate be used to reduce
manual steps?
Let’s look at a company that understands the
• How about Syntex and Purview?
limitations of the out-of-the-box approach but
wants to maximize their existing Microsoft 365 Solving this puzzle requires the correct
license. What hurdles do they need to overcome configuration of several interconnected
to solve their information-management, solutions within the Microsoft toolset—not
compliance, and automation challenges? only doing this once, but also maintaining this
configuration as the business evolves, as the
The first step is to understand the big picture. Microsoft toolset evolves, and as knowledge
• What’s the information architecture of workers come and go. That again raises the
the business? larger question about investments.
• Who is creating each piece of information? • Are the required features included in the
• How does it flow? current licensing tier?

• What are the governance requirements • How long does it take and how much does
for different types of information? it cost to build it all?
• Does the company consider these factors
The next step is to create an information- to fall under in-house competence or
management plan and the processes to should they outsource the work?
govern it. Finally, the organization must be
ready to evaluate implementation and Microsoft 365 is a powerful toolset with
enforcement of that plan. significant potential. However, it’s also
vast, complex, and evolves at a rapid pace,
Deploying such a plan takes some time and effort. requiring dedicated professionals to build
Translating it to Microsoft 365 and SharePoint and run it and making the cost-benefit math
environments require deep expertise. a necessity before long-term commitments
• What’s the role of each system? can be executed. Can a company build the
desired solution with Microsoft 365 or is there
• Are they capable of storing content? another solution that gets closer to the
• How does the company enforce requirements vision more rapidly? What do the long tail of
that end-users save the information to the maintenance and the evolution of business
right place across every step? needs look like?
• What kind of SharePoint site hierarchy
should the company have?

© 2024 M-Files. All rights reserved.

9
Considering the whole IT landscape
So far, we’ve mostly examined the issue of • To what extent are we trying to automate
organizing documents within the Microsoft our processes and how many different
365 environment. However, the average number SharePoint sites and lists do we need to
of applications a knowledge worker uses is achieve that?
11 and more than 40% of digital workers use • What’s the cost of that implementation project?
more than that. No document lives in isolation.
It’s always a part of a variety of business • What happens if a new employee is unaware
processes. How do we build efficient, end-to- of the company policy regarding the storage
end processes that include documents and of agreements in the contracts archive?
other IT systems? • How can people discover all the various
• processes and practices related to storing
That’s easier said than done. However, there
information according to the master plan?
are tools within Microsoft 365 for that option.
SharePoint can be used to manage information • How can AI help?
in the form of lists. Each column in the list can • Can generative AI and Microsoft Copilot
have custom data types, allowing users to finally solve the problem?
store documents and tag them with additional
information potentially related to in-place
business processes. It’s also possible to
retrieve the values of those columns from
non-Microsoft systems through integrations.
For example, there could be a SharePoint site
for archiving contracts—that site could have a
list with columns specifying type of contract,
contracting parties, and other information
used to expose data outside the document.
It’s clear that achieving end-to-end
automation is possible, but the challenge
94 %

of professional services organizations rely

reverts to the big picture. on Microsoft office productivity tools.
Source: The Total Economic Impact™
Of Microsoft 365 E3, Forrester, October 2022

Generative AI alone won’t solve the problem

Generative AI can help in solving the problem the quality of data has an impact on the quality
of information chaos. Large language models of the responses by LLMs. In the worst-case
(LLM) can harvest insights from vast amounts scenario, the answer to a question is not based
of data and save a tremendous amount of on any real fact but simply generated by the AI.
time that knowledge workers spend searching For a GenAI application to provide accurate
for the correct information. However, as answers, input given to the AI must be accurate.
demonstrated by general purpose chatbots Summarizing a piece of text is simple because
and early attempts at enterprise GenAI pilots, all the needed information is there.

© 2024 M-Files. All rights reserved.

10
Answering a question based on enterprise Indexing all that data will incur an unreasonable
data is a whole different ball game. How do amount of new costs and will include content
we know which information to give to the AI? that the shiny new Generative AI should never
Can’t we just give all the information? That read. Additionally, if none of the content is
would essentially be the same as asking a classified and governed, the AI won’t be able
search engine to start indexing all data from to know which document was a poorly written
scratch every time someone performs a search. draft, a test document created by a summer
So real-world GenAI applications that are tasked intern, or an authoritative piece of knowledge
with harvesting organizational knowledge crafted by a seasoned subject-matter expert.
from enterprise data need to first index When an employee is looking for good examples
that data. Such an index is not your typical of project plans, the results might be influenced
keyword counter but a semantic index across by poor-quality content.
all data inside all documents, encapsulating
the meaning of each text snippet. The question might also lack proper context—
the GenAI may base the answer on a variety
This system then needs to be able to understand of projects plans such as IT projects, marke-
the meaning of any question to retrieve the right ting projects, and office-relocation projects.
information from the index. Both requirements If the employee were a business consultant
are counterproductive in an environment of completing projects for clients, they wouldn’t
information chaos. appreciate the answer. Even if the prompt
would be sufficiently rich in giving context to
the AI, there’s no guarantee that semantic
Analysts estimate that

90%
index could match that context.
of documents that
enterprises store are To summarize, a successful GenAI application
obsolete, duplicates, or requires well curated source content to build
irrelevant to the business. a robust index and it needs the user prompt
to be bolstered by additional context. Optimally,
the prompt offers an explicit context, such
as limiting the answer to specific type of
project-plan documents.

© 2024 M-Files. All rights reserved.

11
SUCCESS MEANS ADDRESSING
THREE ESSENTIAL DIMENSIONS
Focusing on the essentials for business impact
There are always more ideas than time, attention, That doesn’t mean organizations don’t have
and budget allow to execute them. Any to think about the big picture. Quite the
attempt to fix information chaos must start opposite—only by cataloguing the businesses,
with areas that produce the best business processes, and systems is it possible to
impact in the least amount of time. Is there evaluate the starting point. Only by having a
a process, a department, or a business unit basic grasp of the overall changes required
that suffers most from these challenges? will executives endorse the first experiments.
What is the smallest amount of change the
company can institute that leverages the most
measurable results? For example, is there a
specific type of information that could easily
be isolated?

Sometimes, organizations start the change

with commercial documents like proposals There's no shortcut to
and contracts. Other times, they relate best to
the execution of projects and operations. cleaning up data, but
For some, it’s management of assets. Industries there are best practices
and organizations are different but usually
the prime candidates can be clearly identified. in achieving that goal.
Organizations need to
Showing tangible results and proving that the
approach works will get the attention of company adopt an approach that
leadership, the buy-in of the business prioritizes business impact,
stakeholders, and promote confidence in
pursuing the long-term journey. makes technology choices
analytically, and ensures
that end users are on board
across the entire journey.

© 2024 M-Files. All rights reserved.

12
Performing due diligence in choosing the right technology
After the optimal scope has been identified, The last piece of the puzzle is about
it’s time to address the technology challenge. "fit-gap" analysis.
It may feel overwhelming, especially if there • If these are the requirements, how well
have been prior failed attempts, but in the does our current solution fit the need?
end it’s just analysis work. Any analysis starts
with understanding the current state, defining the • What are the gaps and what is the estimated
desired future state, and figuring out execution. cost of filling those gaps?
• What other solutions in the markets could
When the scope has been decided, the "as-is" solve these problems and what would the
analysis has been halfway done. The problems fit-gap analysis look like for them?
are known—at least on an anecdotal level—
• Finally, what do return-on-investment
and help guide the deeper, system-level analysis
calculations reveal when we compare
of outstanding issues, thus informing the
project costs, license costs, maintenance
desired state. Such an analysis generates a
costs, and expected productivity gains?
list of issues that need to be eliminated and a
set of requirements and capabilities designed
to address each issue.
The average number of applications

11
• How do we go about classifying documents?
a knowledge worker uses is
• How do we tie those documents to the
relevant business processes? and more than

40%
• What kind of workflows would automate
and guide workflow? of digital workers
• What are the governance requirements use more than that.
related to the lifecycle of each type
of document? Source: Gartner 2023
• What GenAI use cases can we envision
and what requirements arise?

© 2024 M-Files. All rights reserved.

13
Generative AI: Getting your people on board
The last dimension to success is people—the Once the scope and technology have been
end-users. They are the ones who experience finalized and built into the first prototype of
both the current situation and the envisioned the future state, the company must engage
future, including all the positives and negatives. end-users in piloting and experiencing the
If they are satisfied with the current state of new normal.
affairs, it will be harder to convince them to
change. If they suffer from information chaos
and demand something better, the change
will be an easier sell.

In both cases, however, the promise of GenAI

is a powerful incentive. If a company can
demonstrate the simplicity and power of By respecting all the feedback
interacting with natural language and can
show a significantly better future situation
from the initial pilot groups
with GenAI, knowledge workers will accept and by carefully addressing
some level of discomfort with a more structured
way of working.
immediate worries, the
company can start to
As with any system, process, or workflow
evolution, a company will need to involve end
build confidence in the
users early—they are the ones invested in the new normal, while also
early analysis to diagnose issues and choose
the right scope. They raise practical issues
growing a group of internal
and question with regards to their current advocates who further sell
way of working. And they can be engaged to
envision and promote a better future.
the solution to the next
wave of users.

© 2024 M-Files. All rights reserved.

14
CONCLUSION
Most organizations use office productivity
suites such as Microsoft 365 to create and
collaborate on content. They also end up
managing the content on these platforms
which can lead to information chaos. As a
result, knowledge workers suffer from poor
productivity, security and compliance are at
risk, and organizations miss the opportunity
to accumulate and reuse organizational
knowledge hidden inside their documents.

Organizations look to GenAI to solve information

chaos but realize that GenAI also needs
well-curated content to produce valuable
results. There’s no shortcut to cleaning up
data, but there are best practices in achieving
that goal. Organizations need to adopt an
approach that prioritizes business impact,
makes technology choices analytically, and
ensures that end users are onboard across
the entire journey.

Organizations who choose to travel this path

will gain a head start in the new information
economy where knowledge workers are
empowered with the entirety of organizational
knowledge and where information technology
is, not a barrier, but a booster.

© 2024 M-Files. All rights reserved.

15
 ABOUT
THE AUTHOR
Ville Somppi has more than 20 years of
experience defining, developing, and
deploying technology solutions for knowledge
intensive industries. Ville has worked with
companies across the globe ranging from
large corporations to small start-ups and
family-owned businesses. He has hands-on
experience deploying, optimizing, and using
Microsoft solutions.

At M-Files, Ville is in charge of understanding

the needs of target industries and improving
product-market fit. Ville holds a masters
degree (MSc) in software engineering from
Tampere University of Technology.

© 2024 M-Files. All rights reserved.

16
M-Files is the leading platform for knowledge work automation.
With the M-Files platform, knowledge workers can find information
faster, work smarter, and achieve more. M-Files features an
innovative metadata-driven architecture, embedded workflow
engine, and advanced artificial intelligence. This enables customers
to eliminate information chaos, improve process efficiency, and
automate security and compliance.

For more information, visit m-files.com

M-Files has offices in eight countries.

To contact one of our regional offices, visit:
m-files.com/en/contact-us

M-Files is a registered trademark of M-Files Corporation.

All other registered trademarks belong to their respective owners.