Protecting your moat

The rise of AI is transforming industries built on professional expertise, from consultancy and finance to law, engineering and architecture. For decades, these firms relied on human labour as the primary engine of value creation. Today, technology is no longer a mere tool (“picks and shovels”) to assist human effort; it is an “excavator” capable of automating the entire process.

This shift forces firms to confront a critical question: who controls the AI stack, and at what layer?

At stake is the future of knowledge work. Organisations that cede control risk funding their own obsolescence. Those that strategically own their AI stack, however, can build defensible moats and retain their competitive edge.

Mapping the AI stack: three strategic entry points

To navigate this transition, organisations must understand the three layers of the AI stack and their trade-offs.

1. High-level goal tools

“Plug-and-play” AI solutions that execute complex tasks (e.g. generating audit reports, analysing contracts, designing structural models, or synthesising market research) with minimal user input.

Pros

• Speed: instant results with no upfront development.
• Pilot-friendly: easy to test with minimal staff involvement.
• Short-term ROI: appeals to firms seeking quick productivity wins.

Cons

• Cannibalisation risk: vendors own the IP, eroding the firm’s unique value proposition.
• Power transfer: clients may eventually bypass the firm to work directly with the software vendor.
• Abstraction trap: solutions are generic and fail to capture the firm’s specific “secret sauce.”

2. Domain engineering

Building proprietary AI tools by codifying an organisation’s intellectual capital, data, prompts, specialised workflows, and proprietary decision trees, into the system.

Pros

• IP ownership: retain control over institutional knowledge and “how” the work gets done.
• Competitive moats: create unique systems that competitors cannot easily replicate.
• Upskilling: professionals evolve into “domain engineers,” blending subject matter expertise with AI fluency.

Cons

• Resource-intensive: requires time to extract and structure human expertise.
• Slower go-to-market: delays initial ROI compared to off-the-shelf tools.

3. Software engineering

Owning the full AI stack from the bottom up by developing the underlying SaaS platform that hosts the domain engineering and high-level goal tools.

Pros

• Total control: zero reliance on third-party infrastructure.
• Cost efficiency: long-term savings as AI-assisted coding drives down development costs.

Cons

• Resource drain: diverts capital and focus away from the core professional service.
• Lag risk: dedicated tech giants will likely outpace in-house development speed.

Is domain engineering the only viable path?

While all three layers have merits, knowledge work organisations face existential risks if they prioritise short-term convenience over long-term strategy.

The high-level goal trap

Investing in vendor-owned AI tools that automate core processes trades immediate gains for permanent vulnerability. Organisations effectively fund the very vendors who will eventually replicate their workflows and marginalise their firm as a middleman.

The software engineering mirage

Building the full technical stack may seem appealing, but most firms lack the technical DNA to compete with well-funded AI startups. The opportunity cost, diverting focus from extracting and codifying their own expertise, is often too high.

The Jylo thesis: domain engineering as a defensible future

We advocate for a hybrid approach that balances control, speed, and sustainability:

1. Start with domain engineering. Extract and codify your firm’s unique IP into AI-driven workflows. Train your experts to become “prompt architects” who refine these systems.
2. Mitigate supply chain risk. Ensure you have an “exit ramp.” Jylo can offer enterprise customers a licence to the source code after a set term, eliminating the risk of vendor lock-in.
3. Avoid the quick fix. High-level goal tools are tempting but erode your differentiation. Domain engineering is where your true value is protected.

Conclusion: own the layer that matters

The AI era demands that knowledge work organisations rethink their value proposition.

By owning the domain engineering layer, firms can preserve their IP as a competitive asset and future-proof their offering against disintermediation.

The alternative, outsourcing the core of your intelligence to vendors, is a Faustian bargain, better known as a deal with the devil.

Firms that choose convenience today … might pay with their relevance tomorrow.