Why your domain experience is the only moat that AI can’t breach

Our next article may resonate most with leaders in the services sector - legal, accounting, consulting, and other knowledge-intensive fields - where decades of hard-won expertise are the true source of competitive advantage.

Everyone is rushing to build AI into their business. Few talk about the frustration, false starts, and slow grind of turning years of expertise into something machines can actually use.

Eight months ago, I started building a digital twin of our business. What I've learned since then may help anyone trying to protect their competitive advantage in the age of AI. To transform 20 years of accumulated domain experience into an intelligent digital assistant capable of revolutionizing how we serve clients.

What I've learned since then has fundamentally reshaped my understanding of competitive advantage in the AI era.

The mirage of easy integration

Initially, connecting to our SharePoint repository seemed straightforward. After all, we had two decades of meticulously documented award-winning projects, crisis management playbooks, and communication strategies. This treasure trove of domain knowledge would, I believed, seamlessly translate into a smart brain through modern data vectorization techniques.

I was wrong.

Over eight months, I encountered error after error attempting to implement RAG (Retrieval-Augmented Generation) systems. Each vectorization approach promised to unlock our data's potential, yet none could adequately capture the nuanced expertise embedded in our documents. The technical infrastructure - the APIs, the embeddings, the retrieval mechanisms - all functioned perfectly in isolation. But they failed to grasp the contextual richness that makes domain experience valuable.

After weeks of chasing elegant solutions, I found myself back at square one: downloading thousands of files, reading them line by line, and manually deciding what mattered. It felt like failure - until I realized it was the only way to preserve the judgment that makes domain knowledge valuable.

Validation from McKinsey's latest research

My experience aligns remarkably with McKinsey's recent analysis of over 50 agentic AI deployments. These stumbles are a natural evolution of any new technology, and we've seen this pattern before with other innovations. The research reveals six critical lessons that most companies miss when implementing agentic AI:

1. Reimagine workflows, not agents: Focusing on the workflow instead of the agent enabled teams to deploy the right technology at the right point, which is especially important when reengineering complex, multistep workflows. Takeaway: Don't bolt AI onto existing processes - step back and redesign the flow.
2. Match tools to steps: Complex workflows require different technologies at different points - not a one-size-fits-all solution. Takeaway: The right model for one stage may be the wrong one for another. Don't force it.
3. Invest in evaluations: Without proper measurement and feedback loops, AI systems lose effectiveness and trust
4. Track and verify every step: Building observability ensures mistakes surface early and systems improve continuously
5. Reuse components: Companies can develop agents and agent components that can easily be reused across different workflows, and make it simple for developers to access them. This approach can virtually eliminate 30 to 50 percent of the nonessential work typically required
6. Humans still matter: As AI agents continue to proliferate, the question of what role humans will play has generated much anxiety - about job security, on the one hand, and about high expectations for productivity increases, on the other.

This resonates deeply with my journey. I initially chased the flashy agent - the sophisticated technical architecture - without first reimagining our fundamental business process workflows.

The Hard truth about data transformation

Eventually, I accepted a sobering reality: there are no shortcuts to building genuine domain intelligence. The path forward required abandoning my elegant automated solutions for a more laborious approach: downloading data locally, manually reviewing thousands of documents, and meticulously separating relevant issues and crisis management materials from unrelated content.

This ongoing offline process - slicing, dicing, and anonymizing data - feels almost antiquated in our age of instant AI solutions. Yet it's precisely this human-in-the-loop curation that preserves the subtle expertise that makes domain knowledge valuable. Every document I review, every decision about what to include or exclude, adds another layer of human judgment that pure automation cannot replicate.

McKinsey's research validates this approach. An important starting point in redesigning workflows is mapping processes and identifying key user pain points. This step is critical in designing agentic systems that reduce unnecessary work and allow agents and people to collaborate and accomplish business goals more efficiently and effectively.

Why domain experience remains the ultimate moat

This journey has crystallized a crucial insight: while AI capabilities are rapidly commoditizing, deep domain experience remains defensible. Here's why:

1. Context cannot be downloaded: Twenty years of client interactions, crisis responses, and project learnings create a web of contextual understanding that resists simple digitization. Each refined business process carries implicit knowledge that only reveals itself through careful extraction.
2. Judgement requires curation: Knowing which past crisis is relevant to a current client situation, understanding the subtle differences between similar-seeming issues - these judgments require human expertise to encode properly into any AI system. As one legal services provider discovered, Legal reasoning in the company's domain was constantly evolving, with new case law, jurisdictional nuances, and policy interpretations, making it challenging to codify expertise.
3. Trust is built on specificity: Clients don't just want AI-powered insights; they want AI-powered insights grounded in proven, relevant experience. A digital twin that can reference specific, anonymized but real scenarios from two decades of practice offers something generic AI never can.

Building with observability and modularity

Following McKinsey's guidance on building reusable components, I've restructured my approach to create transparent, trackable processes at every step. Rather than building a monolithic AI system, I'm now developing modular components that can be reused across different workflows.

This means matching tools to specific steps rather than forcing a one-size-fits-all solution. Some aspects of our crisis communication expertise require sophisticated language models, while others need simple decision trees. The key is knowing which to apply where - knowledge that comes only from deep domain understanding.

The path forward

As I continue this painstaking process of data preparation, I'm more convinced than ever that our domain experience represents an unassailable moat. While competitors can access the same AI models and implement similar technical architectures, they cannot replicate decades of accumulated wisdom - especially when that wisdom is thoughtfully curated and properly integrated into AI systems.

The future belongs not to those with the best AI, but to those who can best marry AI with irreplaceable domain expertise. Our digital twin, once complete, won't just assist clients with generic advice. It will offer insights drawn from thousands of real scenarios, filtered through years of professional judgment, and delivered with the nuance that only comes from genuine experience.

The lesson for leaders

If you're sitting on years of domain expertise, don't be seduced by the promise of instant AI transformation. The real work - the valuable work - lies in the careful, methodical process of transforming that expertise into AI-ready knowledge. It's about reimagining entire workflows, not just deploying impressive technology.

As McKinsey's research makes clear, successful agentic AI deployment requires investing in evaluation, building observability, and maintaining human oversight. It's slow, it's manual, and it's exactly what will separate the winners from the losers in the AI-augmented future.

Your domain experience isn't just data to be vectorized. It's your moat. Protect it, curate it, and transform it thoughtfully. The competitive advantage you build will be worth every painstaking hour invested.

If you're sitting on years of domain expertise, now is the time to decide: will you let it sit in filing cabinets and SharePoints, or will you transform it into the moat that protects your future? I'd love to hear how you're approaching this - what's worked, what hasn't, and which of McKinsey's six lessons resonates most with your journey.