Tech • AI • Robotics
FDE: The €1M+/year AI job explained
5/10TL;DR
A new AI-enabled advisory model is targeting family-owned companies in the United Arab Emirates by taking a share of measurable profit gains instead of charging software subscriptions or consulting fees.
KEY POINTS
From software to outcomes
The core thesis is that companies increasingly want to buy results rather than tools or processes. In this model, the traditional divide between software, services, consulting and even private equity is collapsing as AI boosts the productivity of service delivery and sharply lowers the cost of building software. One figure underpins that view: for every $1 spent on SaaS, companies still spend about $8 on services.
A performance-fee model tied to EBITDA
The firm approaches clients with a simple proposition: grant access to data, management information and internal operations, and it will work to improve EBITDA. Instead of charging an annual management fee, it takes 15% of measurable performance gains, with 0% management fee up front. The structure mirrors private equity incentives but shifts the emphasis toward operational improvement rather than financial engineering.
Why family-owned Gulf companies are the target
Family-controlled businesses were identified as the easiest entry point because a single shareholder or family authority can impose decisions across the organization. That reduces internal politics compared with fragmented corporate structures. The focus is on Emirati family groups, a market described as large, concentrated and capable of offering rapid referrals once results are proven.
Why the strategy is centered on the UAE
The UAE is seen as unusually suited to this model because of its business culture, relatively light regulatory friction and concentration of wealth. The claim is that only a few markets, including Singapore, the United States and the UAE, have the necessary comfort with performance-based commercial relationships, but the Emirates combine that with simpler operating conditions around data use and corporate execution.
How the data layer is built
The operational process starts by ingesting virtually everything: emails, text messages, Microsoft 365 records, databases, PDFs and internal documents. With consent, software can also be installed on employee computers to capture workflows and screen activity. The objective is to create a context-rich internal data layer rather than rely on accounting categories alone.
Open-source models over frontier AI for ingestion
The economics of the system depend on using open-source models on in-house hardware for large-scale data processing. For one client with 4,000 employees, hardware investment was put at roughly €400,000 to €500,000, or about $100 per employee. Using leading commercial frontier models for the same ingestion work was estimated to cost tens of millions, making that route commercially unrealistic.
Finding hidden costs outside accounting labels
One early example involved marketing waste that did not appear under a marketing line item. Printed brochures sent by post were booked as administrative mail expenses, not as marketing. Once traced back to sales impact, the campaign was judged ineffective and scrapped, producing about €20 million in annual savings and freeing budget that could be reallocated to digital advertising.
Turning employees into internal builders
Rather than positioning AI as a threat to jobs, the model tries to align staff incentives with automation. One proposal offered employees the equivalent of one year of salary if they created internal software that improved EBITDA through a company-run software factory. The premise is that domain experts, not generic autonomous agents, are best placed to identify where code can automate valuable work.
Training over fully autonomous agents
The approach is skeptical of current AI agents operating independently across complex workflows. Instead, it argues that the real value lies in helping employees build deterministic software for their own functions. That has led to plans for workshops and training programs so engineers can use AI tools and integrations tied to software such as Blender, AutoCAD and other technical systems, potentially replacing outside technical consultants that currently cost clients tens of millions.
Only part of the gains are formally billed
Not every productivity gain is tracked. The model assumes that roughly 60% to 70% of results are objective enough to measure directly, such as eliminating a consultant invoice, removing a role, or creating new revenue in a previously inactive business line. The remaining second-order effects are intentionally left unbilled, both to simplify negotiations and to strengthen long-term client relationships.
CONCLUSION
The model reflects a broader shift in enterprise AI from selling tools to selling measurable business outcomes. Its success will depend on whether deep data access, employee buy-in and performance-based pricing can consistently translate into durable EBITDA gains.
Explain this