Knowledge management takes center stage in the AI journey

According to the Ark Invest Big Ideas 2025 report, agents will increase enterprise productivity via software. Companies that deploy agents should be able to increase unit volume with the same workforce and optimize their workforce toward higher-value activities. 

Artificial intelligence (AI) will also supercharge knowledge work. Through 2030, Ark expects the amount of software deployed per knowledge worker to grow considerably as businesses invest in productivity solutions. AI agents are poised to accelerate the adoption of digital applications and create an epochal shift in human-computer interaction.

Also: Here’s what AI likely means for traditional BI and analytics tools

According to the 2025 Connectivity Benchmark Report by MuleSoft and Deloitte Digital, 93% of IT leaders report intentions to introduce autonomous AI agents within the next two years, and nearly half have already done so. However, the journey to agentic AI adoption and a digital labor force will not be easy without a deliberate focus on data and knowledge management strategy. 

The Mulesoft and Deloitte research, based on feedback from 1,050 CIOs, reveals that data is trapped across siloed enterprise applications. The average enterprise uses 897 apps, with 45% using 1,000 applications or more — further hindering IT teams’ ability to build a unified experience. Only 29% of enterprise apps are integrated and share information across the business. Most (93%) IT leaders feel that AI will increase developer productivity over the next three years, up seven percentage points since 2023.

Disconnected data remains an overwhelming blocker to legacy modernization for organizations. Today, 83% of enterprises report that integration challenges are a significant barrier to their legacy modernization efforts. And 97% of IT leaders acknowledge that their organizations struggle with integrating end-user experiences. Only 10% of respondents report experiencing no challenges due to data silos, while 74% of organizations find their IT systems to be overly interdependent. 

Poor integration and disconnected data lead to poor knowledge management. Without a strong knowledge management ability, customer experience leaders will be unable to realize the potential of AI, including agents. Deloitte research notes that 25% of enterprises using AI will deploy AI agents by 2025. 

Also: The end of data silos? How SAP is redefining enterprise AI with Joule and Databricks

Accenture research, meanwhile, highlights the importance of knowledge management as a core ‘cognitive digital brain’ capability, needed for businesses to adopt agentic AI solutions. So, why is knowledge management a critical success factor for AI adoption in business, including the ability to use agentic AI to create a digital labor force?

To better understand the impact of knowledge management on the successful adoption of generative AI (Gen AI) and agentic AI in the enterprise, I interviewed one of the world’s leading experts on customer relationship management (CRM), customer experience (CX), knowledge management (KM), and customer service. Michael Maoz is senior vice president of innovation strategy at Salesforce. Before joining Salesforce, Maoz was a research vice president and distinguished analyst at Gartner, serving as the research leader for the customer service and support strategies area. 

Here are the highlights of our conversation.

In my days as a Customer Support leader, knowledge management was a differentiator, though it wasn’t a discipline that would get you promoted. Why is that about to change?

Michael Maoz: Without a strong knowledge management ability, customer experience leaders will be unable to realize the potential of AI. The quality of ‍knowledge management (KM) in the enterprise is receiving new scrutiny as businesses work to scale the use of generative AI and agentic AI. As you like to say, data is the new water — and this is especially true for knowledge articles and content overall: it has to be abundant, clean, and accessible. Support organizations that already have a strong KM competency are scaling up projects like Agentforce (Salesforce’s agentic AI solution).

You have researched why some companies thrive at KM, while others may lag. What patterns do you see, and how do they come about?

Michael Maoz: It will always come down to who owns the KM process. In industries where customer service does not own the process, KM is an immature discipline, through no fault of their own. At the other end of the spectrum are luminaries in industries where knowledge management is a top priority and the customer support team owns, or has the final word on, the knowledge content they use and deliver to customers. These industries, such as technology, aerospace, and medical equipment, adhere to standards, such as the Consortium for Service Innovation’s Knowledge-Centered Service (KCS) or guidelines from the Technology and Services Industry Association (TSIA).

<!–>

If we leave aside any complications that will arise in the KM area now that we want to link it to Gen AI, what makes enterprises great at developing the right content, and making it available in the right format – which could be a long PDF or an image or video – and easily searched for and found on the channel of the customer’s choosing?

Michael Maoz: The great companies have a strong culture around knowledge, and gamify the process. I remember visiting a medical equipment manufacturer in New Hampshire and some of the tech support team had small trophies on their desk for their skill at solving problems. Here is what the best have in common:

• Knowledge management is recognized company-wide as an imperative
• Knowledge creation is centralized and/or coordinated among departments
• Systems are in place to capture knowledge from all relevant sources, including user chat and community forums, phone interactions, and device signals
• Measurement tools monitor content use, and the impact of content on self-service, and their customer surveys reflect how people consistently receive the correct answer regardless of channel (email, messaging, bot, web, or app)
• A strong discipline around content usage, including: knowing how to create content, promoting it to the right channel, and making it available for the right tasks; identifying knowledge gaps where there was a ‘failure to find’ reported by the service representative, customer, or bot; having data that points to knowledge articles or content that needs to be removed because it is outdated or irrelevant
• The KM system is integrated into the CRM system, allowing deep personalization

Companies like Amazon’s Ring division (who sell home doorbells, alarm systems, and security cameras) and Dyson are examples of B2C product-centric companies that understand that good, contextually relevant documentation that’s available across channels and tied to a specific product lowers support costs and delights customers. 

Also: 15 ways AI saved me time at work in 2024 – and how I plan to use it in 2025

Effective companies also recognize the limited number of possible issues that the customer can contact them about. B2B companies in high-tech or medical equipment are usually the most mature in their adoption of KCS-type methodologies. They are subject to product updates, new software releases, and factory recalls, and they sometimes have a third party performing some or all of the support for products.

Why is shining a light on our strong focus on AI, and by that we mean Gen AI and agentic AI, further emphasizing the need for the quality of the knowledge used to address customer issues?

Michael Maoz: How many loud, public displays have we seen in the past year of Gen AI creating content that was fake, incomplete, wrong, or demonstrating bias? Not to mention that it lacked governance around ethical use. No head of customer service or customer experience wants to get that text from the CEO referencing an AI failure. 

Fortunately, there are approaches to KM that accelerate the safe use of your knowledge content by adding a Gen AI component. That is exactly what we are doing with Agentforce. The first is the need to de-risk every Gen AI project. It’s like our Hippocratic Oath: do no harm. That is easy when you are putting Gen AI in front of a closed knowledge repository that has been carefully curated and used to answer only a few questions or write a limited range of texts or emails. 

Also: Why ethics is becoming AI’s biggest challenge

How do you scale that approach to incorporate all of your corporate data, both the structured data in your CRM or ERP system, but also your unstructured data like PDFs and chat logs? For this, you need strong data governance. You need to run the content past the ethical use team to eliminate biases. You need a privacy layer. You need to factor in any governance or data privacy issues. It’s complicated.

What’s the secret of building a trusted KM system that includes all forms of AI?

Michael Maoz: Make friends with marketing. As customer service professionals mature their AI capabilities, they will start to work in a more coordinated way with marketing. Many of the best KM practitioners would find affinity with content creators in marketing. Marketing content teams have an editorial board, a content governance board, and a marketing leadership team that sponsors the content. They work hard to make sure that the brand tone comes through, and in the right style and language, so that it reinforces the brand.

Also: Generative AI is now a must-have tool for technology professionals

That is the future of customer service – empathy at scale, and reinforcing the brand. Marketing understands the different personas that marketing content is targeted at and creates the necessary templates, and they often have a content center of excellence to ensure that writers, content designers, and content strategists work together. Soon marketing and customer service will be highly coordinated. Here are a few KM best practices we are observing:

• Define the single source of truth: The best teams test that the content is up-to-date and compliant. They run tests to avoid hallucinations, which can happen when the AI answers a question using out-of-date information. When there are multiple departments with access to the same information, they make sure that the AI uses the source of the content rather than where it was copied. Our customers even use Agentforce to find duplicate sources of information and check that the sources are harmonized.
• Tag content: Any great technical publications team already does the work of tagging content with metadata so that Gen AI can retrieve it quickly. Now, in advanced use cases, the content is accessed only after first looking at the customer buying history, satisfaction level, influencer score, and propensity to churn. This approach allows for new levels of personalization.
• Survey and feedback: Using KCS practices, teams survey customers to see that the answer meets their expectations and register any changes that need to be made. Working with the AI and data scientists, they can ask for tweaks to the AI model. In the case of Salesforce, we have a reasoning engine that monitors the knowledge ecosystem throughout a process to see if it proceeds as designed. It then works to modify and optimize the process.

What should we expect over the next three years from AI-enabled KM practitioners?

Michael Maoz: The only limit on the use of AI is your imagination. Knowledge management is at the heart of minimizing so many tasks carried out by human agents today. Once the knowledge is trustworthy and retrievable, the possibilities are endless. This is where storytelling comes in – to inspire leaders. For our human service reps, Agentforce can, for example:

• Assist agents with the right advice
• Update multiple, integrated systems to ensure data accuracy
• Act on behalf of the human agent – opening a case, closing a case, retrieving information, sending notifications – to do all the repetitive, dull, but important tasks that AI can perform faster, more accurately, and at scale
• Craft the correct phrase with the right tone
• Suggest what to say or text the customer
• Attach the relevant article, image, or sound file to solve a problem
• Create workflows to onboard new employees
• Summarize a case or phone conversation
• Pull out the most important points in an email or voice recording
• Understand human emotion and respond with empathy or greater patience or by providing more detail

For end customers, KM with Agentforce has amazing benefits. Here too, the only limitation is the imagination. For example, with a customer authenticated on the website or in an app or with conversational AI, the tool can perform actions such as processing an order return or pre-filling a form, checking and updating shipping details, providing personalized advice, negotiating the right discount, or waiving a fee. A conversational AI chatbot using a curated knowledge base can answer all factual questions. 

Also: Your AI transformation depends on these 5 business tactics

That approach leaves the most complex questions – questions for which the answer is “it depends” – to a human agent, in the same way that only a human should handle emergencies or where the customer is agitated, afraid, or anxious. AI can detect these states and direct the interaction away from self-service to a human agent.

What you should do next

There is so much incredible possibility ahead. Who knew that KM could be this exciting? It also reminds me that so much of what Michael outlined will impact people. I can see multiple new skills and recruitment strategies needed, and a degree of change management that some businesses will embrace, while others will struggle. 

According to the World Economic Forum’s Future of Jobs Report 2025, 170 million new jobs will be created by 2030 and AI will play a key role. There will be times when, even with the best knowledge system, businesses will always need to identify when the customer needs and wants to speak with a human, regardless of whether the AI can answer the question. Here are the key takeaways:

• Work on small wins while working out the long game
• Be disciplined with knowledge management because trusted, available, correct information will make or break the AI journey
• Measure as you go to show value
• Work as a cross-organizational team, and use customer insights to evolve to the next level
• Knowledge management, when connected to AI, opens up a whole new world in satisfying customers, lowering costs, and growing the business

This article was co-authored by Michael Maoz, senior vice president of innovation strategy at Salesforce.

Source: Robotics - zdnet.com