Tag: AI

Posted on

AI Won’t Replace Knowledge Management. It Will Finally Force Us to Take It Seriously

(This post was originally shared in my LinkedIn Profile.)

For decades, knowledge management has been the organisational discipline that everyone agreed was important and almost no one invested in properly.

We built SharePoint graveyards. We created knowledge bases nobody searched. We ran “lessons learned” sessions whose lessons were never applied. And when things went wrong — when the expert retired, when the project team disbanded, when the same mistake happened for the third time — we shrugged and called it institutional memory loss, as if it were a natural disaster rather than a preventable one.

Then came AI.

And suddenly, knowledge management isn’t a nice-to-have. It’s the foundation everything else depends on.

AI is only as good as the knowledge it can access

Here’s what the AI vendors won’t tell you in their sales decks: a large language model deployed inside your organisation will reflect the quality of your organisational knowledge. Feed it outdated policies, inconsistent documentation, and tribal knowledge that lives only in people’s heads — and you’ll get confident, fluent, wrong answers at scale.

The organisations rushing to implement AI copilots and knowledge assistants are about to discover something KM practitioners have known for years: garbage in, garbage out is not a technology problem. It’s a knowledge governance problem.

AI doesn’t fix broken knowledge ecosystems. It amplifies them — for better or worse.

The sudden urgency is real — and welcome

I’ve sat in enough boardrooms to know that the argument “we need to capture institutional knowledge before people leave” rarely moved budgets. Neither did “we’re duplicating effort because teams can’t find what exists.”

But “our AI tool is producing unreliable outputs because our knowledge base is a mess”? That one lands differently.

AI is creating the business case for KM that practitioners have struggled to articulate for years. Not because AI is a KM solution — it isn’t — but because AI makes the consequences of poor knowledge management immediately visible and commercially painful.

That’s not a crisis. That’s an opportunity.

What AI actually makes possible

Used thoughtfully, AI tools offer genuine advances for knowledge work:

  • Surfacing what exists — AI-powered search can finally make organisational knowledge findable, cutting through the folder structures and file naming conventions that defeated traditional search.
  • Capturing tacit knowledge — Conversational AI can assist in eliciting and structuring knowledge from subject matter experts in ways that don’t require them to become writers.
  • Reducing the burden of contribution — One of the biggest barriers to knowledge sharing is the effort it takes. AI can lower that friction significantly.
  • Connecting dots across silos — AI can identify relationships between knowledge assets that no human would have the bandwidth to spot.

None of this happens automatically. All of it requires human judgement, governance, and strategy to work.

The human element isn’t optional

Here’s where I’d push back on the more breathless AI narratives: knowledge is not just information. It includes context, relationship, experience, and meaning — the things that make information useful rather than merely available.

AI can retrieve. It cannot always discern. It can summarise. It cannot always judge what matters. It can generate. It cannot replace the wisdom that comes from having lived through something.

The organisations that will succeed with AI-augmented knowledge management are those that treat the human and the machine as partners — where AI handles the retrieval, synthesis, and surface-level generation, and people contribute the judgment, the curation, and the culture that makes knowledge genuinely shared.

The moment KM has been waiting for

I started my career arguing that knowledge management deserved more strategic investment. I’m still making that argument — with AI more executives are “getting it”.

AI has done what years of best practice frameworks and maturity models couldn’t: it has made knowledge management urgent.

That urgency is an opening. The question is whether organisations will use it to build something sustainable — genuine knowledge cultures, sound governance, human-centred practice — or whether they’ll bolt AI onto the same neglected foundations and wonder why the results disappoint.

The technology is ready. The case is made. Now comes the harder work.

Posted on

Keeping Humans in the Loop: A Strategic Approach to AI Implementation

(this post originally appeared on my LinkedIn Profile)

As organisations increasingly adopt artificial intelligence technologies, a critical question emerges: how do we harness AI’s power whilst maintaining essential human oversight and judgment? The answer lies in implementing robust human-in-the-loop (HITL) systems that combine the efficiency of automation with the nuanced decision-making capabilities that only humans possess.

Understanding Human-in-the-Loop Systems

Human-in-the-loop refers to AI systems designed to incorporate human judgment at critical decision points. Rather than fully automated processes, HITL systems pause at predetermined moments to seek human input, validation, or correction. This approach acknowledges that whilst AI excels at processing vast amounts of data and identifying patterns, humans remain superior at contextual interpretation, ethical reasoning, and handling edge cases.

The concept becomes particularly relevant in Canadian organisations navigating the proposed Artificial Intelligence and Data Act, which emphasises the importance of human oversight in high-impact AI systems.

Why Human Oversight Remains Essential

Contextual Understanding AI systems, despite their sophistication, often struggle with context that humans intuitively grasp. A customer service chatbot might misinterpret sarcasm or fail to recognise when a routine inquiry masks a serious complaint requiring immediate escalation. Human agents can read between the lines and understand the broader context of interactions.

Ethical Decision-Making Complex ethical dilemmas require human judgment that considers values, cultural nuances, and long-term consequences. When AI systems encounter scenarios involving fairness, privacy, or potential harm, human intervention ensures decisions align with organisational values and societal expectations.

Accountability and Trust Maintaining human involvement in AI decision-making processes supports accountability frameworks and builds public trust. Stakeholders are more comfortable with AI systems when they know qualified humans are monitoring and can intervene when necessary.

Continuous Learning and Improvement Human feedback serves as a crucial training mechanism for AI systems. When humans correct AI decisions or provide alternative solutions, this information can be fed back into the system to improve future performance.

Practical Strategies for Implementation

1. Tiered Decision Authority

Implement a hierarchical system where AI handles routine decisions independently but escalates complex cases to human reviewers. For example, in healthcare AI systems used by Canadian hospitals, routine diagnostic suggestions might proceed automatically, whilst unusual cases require physician review.

Implementation Framework:

  • Level 1: AI handles standard cases (80-90% of volume)
  • Level 2: AI flags uncertain cases for human review (5-15%)
  • Level 3: Complex cases requiring specialist human judgment (1-5%)

2. Active Learning Loops

Design systems that continuously learn from human corrections and feedback. This approach is particularly effective in knowledge management systems where subject matter experts can refine AI-generated content recommendations or search results.

3. Confidence Thresholds

Configure AI systems to request human input when their confidence levels fall below predetermined thresholds. This ensures that uncertain decisions receive appropriate human scrutiny whilst maintaining efficiency for high-confidence scenarios.

4. Collaborative Interfaces

Develop user interfaces that facilitate seamless collaboration between humans and AI. Rather than treating AI as a black box, create transparent systems that show their reasoning and allow humans to easily modify or override decisions.

Overcoming Implementation Challenges

Resource Allocation One of the primary challenges in HITL implementation is determining the appropriate level of human involvement. Too much human oversight defeats the purpose of automation, whilst too little risks poor outcomes. Organisations should start conservatively and gradually reduce human intervention as systems prove reliable.

Training and Skill Development Staff need training not just on how to use AI systems, but on how to effectively collaborate with them. This includes understanding AI capabilities and limitations, recognising when to intervene, and providing meaningful feedback for system improvement.

Change Management Some employees may view HITL systems with suspicion, seeing them either as job threats or as additional burdens. Clear communication about the benefits of human-AI collaboration and how it enhances rather than replaces human capabilities is essential.

Sector-Specific Considerations

Healthcare Canadian healthcare organisations implementing AI diagnostics must maintain physician oversight for critical decisions whilst leveraging AI for initial screening and triage. The College of Physicians and Surgeons’ guidelines emphasise that clinical judgment cannot be fully delegated to AI systems.

Financial Services In compliance with Canadian banking regulations, financial institutions using AI for lending decisions must ensure human reviewers can understand and explain automated decisions to customers and regulators.

Government Services Public sector AI implementations require particular attention to transparency and accountability. Citizens have the right to understand how AI systems affect decisions about their benefits, services, or regulatory compliance.

Building Effective Human-AI Teams

The most successful HITL implementations treat humans and AI as complementary team members rather than competitors. This requires:

Clear Role Definition Establish clear boundaries between what AI handles independently and what requires human input. These boundaries should be regularly reviewed and adjusted based on system performance and organisational needs.

Feedback Mechanisms Create structured processes for humans to provide feedback on AI decisions, including both corrections and confirmations of good decisions. This data becomes valuable for system improvement.

Performance Monitoring Implement metrics that measure not just AI accuracy but also the effectiveness of human-AI collaboration. Track how often human interventions improve outcomes and use this data to optimise the balance.

Future-Proofing Human-AI Collaboration

As AI capabilities continue to evolve, HITL systems must be designed for adaptability. What requires human oversight today might be handled autonomously tomorrow, whilst new capabilities may create new needs for human involvement.

Organisations should regularly reassess their HITL configurations, staying informed about advances in AI technology whilst maintaining focus on their core mission and values. The goal isn’t to eliminate human involvement but to ensure it remains meaningful and value-adding.

Conclusion

Keeping humans in the loop isn’t about limiting AI’s potential—it’s about maximising the combined potential of human intelligence and artificial intelligence. By thoughtfully designing HITL systems, organisations can harness AI’s efficiency and analytical power whilst preserving the judgment, creativity, and ethical reasoning that humans bring to complex decisions.

The most successful organisations of the future will be those that master this collaboration, creating systems where humans and AI work together seamlessly to achieve outcomes neither could accomplish alone. In the Canadian context, with increasing regulatory attention to AI governance, human-in-the-loop approaches aren’t just good practice—they’re becoming essential for compliance and public trust.

As knowledge managers and organisational leaders navigate this landscape, the question isn’t whether to keep humans involved in AI systems, but how to do so most effectively. The answer lies in thoughtful design, continuous learning, and a commitment to maintaining the human elements that make organisations truly intelligent.

Posted on

AI-Ready

(this post originally appeared on my LinkedIn profile and has been edited slightly from the original)

I’ve been talking to many people about the importance of keeping humans in the loop and preparing organisations to be AI-ready. With that in mind, I thought I might do a short series on those themes. Today marks the first in a series focused on helping organisations become AI-ready. This has been a theme at the conferences I’ve attended over the last couple of years, so I’m sure many of you have thoughts about this necessity. Feel free to share your thoughts in the comments.

Knowledge managers play a crucial role in preparing organisations for AI adoption. Here are the key areas they need to focus on:

Data Governance and Quality Establish robust data governance frameworks that ensure information is accurate, consistent, and accessible. AI systems are only as good as the data they’re trained on, so knowledge managers must audit existing knowledge repositories, standardise data formats, and implement quality control processes. This includes creating metadata schemas and ensuring compliance with Canadian privacy legislation like PIPEDA.

Knowledge Audit and Documentation Conduct comprehensive audits of organisational knowledge assets – both explicit (documented) and tacit (experiential) knowledge. Map knowledge flows, identify critical knowledge gaps, and document processes that have previously relied on institutional memory. This foundation is essential for training AI systems effectively.

Change Management and Skills Development: Develop change management strategies that address employee concerns about AI, while building digital literacy. Knowledge managers should collaborate with HR and training departments to create programmes that help staff understand how AI will augment rather than replace their work. Focus on developing human skills that complement AI capabilities.

Technology Infrastructure Assessment: Evaluate current knowledge management systems and determine what upgrades or integrations are needed to support AI tools. This might involve migrating to cloud-based platforms, improving search capabilities, or ensuring systems can integrate with AI applications through APIs.

Ethical Guidelines and Bias Mitigation: Establish clear guidelines for ethical AI use within the organisation. Knowledge managers should work with legal teams to develop policies around AI transparency, accountability, and bias prevention. This is particularly important in Canadian organisations operating under evolving AI governance frameworks.

Pilot Programmes and Gradual Implementation Start with small-scale AI implementations in specific knowledge domains before organisation-wide deployment. This allows for testing, learning, and refinement while building internal confidence and expertise.

Cross-Functional Collaboration: Foster partnerships between knowledge management, IT, legal, and business units to ensure AI initiatives align with organisational objectives and compliance requirements.

The key is taking a strategic, human-centred approach that treats AI as a tool to enhance organisational knowledge capabilities rather than simply a technological upgrade.