The Problem With Most AI Companies at Gitex 2025
Walking through Dubai World Trade Centre this October, we counted 247 booths claiming “AI-powered solutions.”
Only 11 could explain their model architecture when asked.
This isn’t a criticism it’s the reality of an industry moving faster than its expertise. Everyone wants to be an AI company. Few understand what that actually means when you’re deploying systems that diagnose medical conditions or manage million-dollar aviation inventories.
We spent three days at Gitex 2025 not just showcasing our work, but having the conversations most companies avoid: What happens when AI fails? How do you validate medical AI without killing your timeline? Why do 73% of automation projects never leave pilot phase?
Here’s what we learned and what we’re building because of it.
What We Actually Showcased (With Real Numbers)
1. EEG-Based Diagnostic AI for Neurological Disorders
The Challenge: Traditional EEG analysis requires 45-90 minutes of specialist time per patient. In Pakistan’s healthcare system, where there are only 600 neurologists for 240 million people, that’s unsustainable.
Our Approach:
- Custom CNN architecture trained on 12,000+ annotated EEG recordings
- Preprocessing pipeline that handles 8 different EEG machine formats
- Validation against gold-standard clinical diagnoses
Current Results:
- Analysis time reduced from 60 minutes to 8 minutes
- 87% concordance with senior neurologist diagnoses for epilepsy detection
- Now processing 200+ scans per month across 3 pilot clinics
What We Don’t Say: This replaces doctors. It doesn’t. It gives them a head start.
The Hard Part Nobody Talks About: Medical AI isn’t about accuracy alone it’s about explaining why the model flagged something. We spent 4 months building interpretability layers that show clinicians exactly which signal patterns triggered alerts. That’s not sexy. It’s necessary.
2. Aviation Parts Marketplace with Intelligent Matching
The Problem: Aviation parts sourcing is a $45B industry running on Excel sheets and email chains. A maintenance manager looking for a specific APU component might contact 30 suppliers and wait 72 hours for quotes.
Our System:
- NLP-powered part number normalization (because “28385-1” and “28385-001” are the same part)
- Real-time inventory matching across 180+ certified suppliers
- Automated compliance verification (FAA, EASA certifications)
Results After 6 Months:
- Average quote time dropped from 3 days to 11 minutes
- 34% cost reduction through competitive bidding automation
- $2.3M in parts transactions processed
What Investors Asked Us: “Can this scale to other industrial supply chains?”
Honest Answer: Yes, but each industry has different compliance requirements. Aviation is hard mode which is why we started there.
3. D2C E-Commerce That Actually Converts (Khaas Honey Case Study)
Most AI marketing is retargeting ads with extra steps. We built something different.
The Setup: Khaas Honey, a premium organic honey brand, was getting traffic but converting at 1.2%—half the industry average.
Our System:
- Behavioral prediction model analyzing 40+ micro-interactions
- Dynamic product bundling based on browsing patterns
- Personalized content sequencing (educational → social proof → offer)
90-Day Results:
- Conversion rate: 1.2% → 3.7%
- Average order value: +42%
- Customer acquisition cost: -28%
- Email open rates: 19% → 34% (personalized subject lines)
The Insight: We weren’t optimizing for clicks. We were optimizing for purchase intent signals—time on ingredient descriptions, comparison behavior, cart hesitation patterns.
The Gitex Moment That Changed Our Perspective
On Day 1, we watched Sheikh Mohammed bin Rashid Al Maktoum walk through the exhibition halls, stopping at startup booths, asking questions, challenging founders.
What struck us wasn’t the ceremony it was the message: Innovation isn’t about technology. It’s about solving problems that matter.
That evening, we rewrote our investor pitch. Removed the buzzwords. Added the failure rates. Focused on impact.
Three investors asked for follow-up meetings. Two specifically mentioned they appreciated the honesty about limitations.
The Three Questions Every Gitex Attendee Asked Us
1. “How is your AI different from [Big Tech Company]?”
Short answer: We build for problems big tech ignores.
Google isn’t optimizing EEG workflows for Pakistani clinics. AWS isn’t customizing aviation supply chains for regional MROs. Our advantage isn’t better algorithms it’s better problem selection and domain expertise.
2. “Can small businesses afford this?”
Current reality: Our healthcare AI costs $800-1,200/month per clinic. That’s real money for a 3-doctor practice.
What we’re building: A usage-based model launching Q2 2026. Pay per scan analyzed, not per month. Makes it accessible to clinics doing 50 scans/month instead of 500.
3. “What’s your biggest failure?”
We deployed a customer service chatbot for a retail client in 2023. Accuracy was 91%. Customer satisfaction dropped 14%.
The lesson: High accuracy doesn’t mean good UX. People don’t want correct answers from robots—they want problems solved by humans who understand context. We pivoted to “AI-assisted human support” where AI handles information retrieval, humans handle conversations. Satisfaction recovered and exceeded baseline.
We now tell every client: AI is a tool for your team, not a replacement.
What Comes After Gitex: Our 2026 Roadmap
Launching in Q1 2026: The Frack Diagnostic API
Making our EEG analysis available as an API for hospital system integrators. Beta partners in Pakistan, UAE, and Kenya.
Pricing: $0.40 per scan analyzed. No minimum commitment.
Expanding Aviation Platform
Adding predictive maintenance modules analyzing aircraft telemetry to forecast part failures before they ground planes.
Target impact: Reduce unscheduled maintenance events by 30%.
Ethics & Transparency Initiative
Publishing our model cards, training data sources, and bias testing results. If healthcare providers trust us with diagnoses, they deserve to know exactly what they’re getting.
The Real Reason Investors Are Paying Attention
It’s not the tech. Every accelerator has AI startups.
It’s the deployment track record:
- 8 production systems currently running
- 3 different industries (with more similarities than differences)
- Real revenue: $340K ARR, growing 31% month-over-month
- Retention: 94% (because our AI actually works)
We’re not a services company pretending to be a product company. We’re not a product company with no customers. We’re proving that specialized AI beats generalized AI when you deeply understand the problem.
Work With Us (But Only If This Applies)
We’re selective about partnerships. Here’s who we work best with:
For Healthcare Clinics & Hospitals
You’re a fit if:
- Processing 100+ diagnostic scans per month
- Frustrated with specialist bottlenecks
- Open to collaborative AI implementation (not plug-and-play)
[Book Technical Demo – Healthcare AI]
For Industrial/Aviation Companies
You’re a fit if:
- Managing complex supply chains with compliance requirements
- Currently using email/phone for procurement
- Ready to invest 90 days in proper implementation
[Request Aviation Platform Walkthrough]
For E-Commerce Brands ($500K+ Annual Revenue)
You’re a fit if:
- Traffic is decent but conversion is stuck
- Willing to share analytics data for model training
- Want a 6-month commitment (AI needs time to learn)
[Get Conversion Optimization Audit]
For Investors
You’re interested if:
- Looking for AI companies with deployed systems, not demos
- Understand that healthcare/industrial moves slower than SaaS
- Value sustainable growth over hockey sticks
[Download Investment Deck + Financial Model]
FAQ: What People Actually Asked at Our Booth
Q: Is this really AI or just smart algorithms?
A: We use supervised learning models (CNNs for EEG, transformers for NLP, ensemble methods for prediction). If you want to debate definitions, we’re happy to talk architecture over coffee.
Q: How long does implementation take?
A: Healthcare: 60-90 days. Aviation: 90-120 days. E-commerce: 30-45 days. The difference is regulatory complexity and integration depth.
Q: What happens if your model makes a wrong diagnosis?
A: Our system is assistive, not autonomous. It flags patterns for clinician review—it never makes final diagnoses. Every output includes confidence scores and reasoning. Legal liability stays with the healthcare provider making clinical decisions.
Q: Can you guarantee ROI?
A: No. We guarantee deployment, training, and support. ROI depends on your team’s adoption and process integration. Our average client sees payback in 5-8 months.
Q: Why should we choose you over [established company]?
A: You probably shouldn’t if you need 24/7 phone support and a brand name your board recognizes. You should if you need systems built for your specific problem, with direct access to the team building them.
What Gitex 2025 Taught Us About the AI Industry
The future isn’t about who has the best algorithm. Google and OpenAI already won that race.
The future is about who can take powerful technology and make it useful for the 99% of businesses that aren’t tech companies.
That’s the gap we’re filling. One clinic, one supply chain, one e-commerce brand at a time.
See you at Gitex 2026 with more systems deployed, more lessons learned, and fewer buzzwords.
Ready to explore how AI can solve your specific problem?
No sales pitch. Just us asking questions about your biggest operational bottleneck and whether we’re the right fit to solve it.
Frack Technologies – Building AI systems that work in the real world, not just in demos.