Understanding the context layer that organizations must develop while building AI agents is the topic I covered in last week’s article. I laid out the architecture of the organization’s AI brain, including knowledge graphs and semantic layers, along with cross-cutting concerns related to AI governance and FinAI (FinOps for AI). Organizations aiming to deploy hundreds to thousands of AI agents will need AI orchestration platforms to coordinate work between agents and people.

That article described a knowledge management architecture – but not the process for creating and maintaining it. Historically, enterprises have struggled with knowledge management initiatives long before AI became a driving force. Some failed to capture knowledge from subject matter experts, while others weren’t effective at getting employees to use it. Knowledge management tools were often disconnected from where the work happened, and organizations struggled with unstructured data governance, especially in managing the lifecycle of documents and other knowledge artifacts. 

Knowledge management’s opportunities and challenges

I will follow up with articles on knowledge management programs. But first, let’s consider the opportunities and challenges facing businesses in healthcare, financial services, construction, manufacturing, and higher education. Look for the patterns: some of these challenges are industry-specific, while others point to common opportunities.

Healthcare: Deliver case-relevant information at the right time

The opportunity: Presenting relevant information to health care professionals for the cases in front of them, improving healthcare decision-making, enhancing the patient experience, and driving  AI ROI.

The challenge: Establishing knowledge management as a fundamental, ongoing business operation and not a technology initiative.

“In industries like healthcare, the challenge isn’t a lack of information but making critical knowledge available when and where people need it,” says Heather Richards, global VP and go-to-market strategy at Verint. “AI can help surface expertise, project history, and operational guidance faster, but the quality of the outcome depends on the quality of the underlying knowledge. Organizations that treat knowledge management as a business discipline, rather than just a technology project, are in a much better position to scale AI successfully.”

Recommendation: Ground and kick off all knowledge management initiatives around change management principles to gain support from both experts and future adopters.

Financial Services: Re-engineer legacy processes with AI capabilities

The opportunity: Enabling a true digital transformation force multiplier by reducing costs and accelerating time-to-decision while improving decision quality.

The challenge: Deciding when to adapt legacy business processes with AI and automation capabilities versus how to re-engineer them, bottom-up, with AI as a foundation.

Daniel Meyer, CTO of Camunda, shared details of an Australian commercial lender during a recent Coffee With Digital Trailblazers on Beyond Legacy Processes: Engineering the High-Velocity Enterprise. “They had many manual handoffs, and once we had that orchestration in place, then we could go in and look at which of these steps actually have to be done by a human. Can we assess risk using AI, present the results to the human for approval or rejection, and then proceed? Can we use AI to research information about the applicant? In the end, the end-to-end application process was 90 times faster,” Meyer said.

Recommendation: Review opportunities to develop both human-augmentation and agentic AI agents, connected with an AI orchestration platform. Look to catalyze a knowledge management program to develop the enterprise’s AI context layer.

Construction: Address challenges with an aging workforce while transforming to AI/digital construction

The opportunity: Architecture, engineering, and construction (AEC) businesses develop competitive capabilities with AI in BIM, digital twins, virtual design and construction (VDC), and other technologies.

The challenge:  Address an aging workforce; capture knowledge without creating job loss fears while attracting entry-level talent to AI/digital construction in operations.

We discussed the AEC industry’s need to invest in knowledge management during the episode, The Cost of Tribal Knowledge: Losing People Can Bring Ops to a Standstill. Bob Salaj, then principal industry advisor at Quickbase, raised two knowledge management challenges:

• Knowledge stored as documents without an information architecture and adoption plan. Salaj said, “Everybody goes in and says, let’s go create a standard operating procedure. In the moment we create that document, it goes into the ether and no one actually ever pulls it out again because no one’s auditing it.”
• Capturing the sixth sense that expertise brings. “Our minds go to the foreman who can walk around and, through smell and touch, know exactly the percent complete of that particular job and where it’s going to be. But let’s also not forget about the innovation that’s occurring extremely rapidly in prefabrication,  VDC, and digital twins. If you lose those people who have tribal knowledge regardless of age, and you’re not documenting in standard operating procedures (SOPs), you’re going to be impacted by it.”

Recommendation: One of the biggest challenges in AEC is treating every construction project independently, with separate tools, information practices, and metrics. AEC businesses should define their project types based on their target growth opportunities and center their knowledge management programs on creating operational standards.

Manufacturing: Empower factory workers to make smarter and faster trade-off decisions

The opportunity: Reduce time-to-decision and empower factory workers collaborating with AI agents to make smarter trade-offs between yield, quality, and other performance indicators.

The challenge: Integrating data collected on the factory floor, back-office supply chain data, compliance requirements, sales forecasts, and other sources to develop the context layer.  

Knowledge management in manufacturing came up during several Coffee With Digital Trailblazers episodes, including the one on Managing AI Agents: New Skills, Operating Models, and Tools.

“A person working in a factory may want to have a higher yield, meaning more products being produced or better quality, but there’s a trade-off, and that trade-off happens a thousand times a day,” said Joanne Friedman, CEO of ReilAI. “You can sacrifice yield for quality but lose a lot of customers because your on-time deliveries fall off.”

Joanne followed up with the story of big automotive manufacturers that had massive recalls costing them billions of dollars because they had overly dialed up yield at the expense of quality.

Recommendation: Friedman reminds executives it’s not enough to know; insight and action must be aligned. Even with connected data, an evolving context layer, and release-ready AI agents, the bigger challenge will be in guiding factory workers to build trust in AI’s recommendations.

Higher Ed: Drive knowledge and AI as a competitive edge 

The opportunity: Turn institutional knowledge and research into a centralized resource for professors, students, and administrators.

The challenge: Establishing a governance structure with a wide range of data owners with different objectives.

Andy MacIsaac, senior strategic solutions manager for education at Laserfiche, says AI in higher education is only as powerful as the institutional knowledge that underpins it. “The real opportunity is to turn decades of information scattered across departments and systems into trusted intelligence that can improve decisions, accelerate student services, and help staff work more effectively. That requires more than an AI tool — it requires a secure, governed foundation that preserves institutional context and allows AI to scale responsibly.”

Recommendation: Leaders will need a top-down, strategic conversation with professors about the opportunity and the urgency of knowledge capture and sharing. Higher education institutions without a centralized knowledge management capability may be on a path to AI disruption.

Knowledge management: The challenge preceding AI’s big opportunities

“AI can accelerate execution, but it can’t replace judgment,” says Rukmini Reddy, SVP of engineering at PagerDuty. “Organizations build critical thinkers by rewarding people who challenge assumptions, connect decisions to the broader system, and stay relentlessly curious about the customer problem they’re solving.”

AI isn’t the only driver for knowledge management initiatives. Enabling creative thinking, curiosity, and challenging the status quo are three people and cultural benefits. These are the antidotes to the tribal knowledge holding back many businesses. 

“If they’re not going to be training these people and they’re going to just let them go, where does that tribal knowledge go?” asked Heather May, founder and president of May Executive Search, during the episode on From Idea to Impact: Shrinking Innovation Cycles With AI. “All the knowledge about your company, the technology, the vision, the mission, the history, is going to just walk out the door. So there has to be an admission by companies: how are you going to maintain knowledge, how are you going to sustain it, and how are you going to value it?”

Joe Puglisi, growth strategist and fractional CIO at 10xnewco, had the mic-dropping moment during the Cost of Tribal Knowledge: Losing People Can Bring Ops to a Standstill episode. “The companies that won’t win are the ones that have the best knowledge management systems. It will be the ones that make tribal knowledge obsolete by making expertise ubiquitous,” Puglisi said.