Study

Independent Investment Research

Applied AI Investment Thesis & Diligence

An investment assessment of five Turkish-founded Applied AI companies with venture-scale potential, focused on the evidence behind the thesis rather than company narratives alone.

Companies were screened for Turkish founding roots, Applied / Vertical AI positioning, venture backing, 10+ employees, and demonstrated commercial traction.

Founder-market fit, market opportunity, AI differentiation, commercial traction, key risks, and diligence priorities.

Full research deck

Investment framework

I used four lenses to keep the analysis consistent across companies and to separate an interesting AI product from a potentially investable business.

01 · Team

Evidence to request: Founder track record, domain depth, and customer references.
Decision test: Can this team execute in its chosen customer workflow?

02 · Market

Evidence to request: Customer budgets, reachable buyers, and a bottom-up market model.
Decision test: Is the entry market large enough for a scalable business?

03 · AI product

Evidence to request: Data rights, performance benchmarks, and evidence that the product improves with usage.
Decision test: Does the AI advantage strengthen as deployment grows?

04 · Commercial pull

Evidence to request: Paid contracts, cohort retention, renewals, and customer ROI.
Decision test: Does adoption convert into repeatable revenue?

Company landscape

Five companies, distinct AI applications
Company	Application	Funding / Stage	Key signal
Vispera	Retail intelligence	~$13.1M · Series A	$8.3M estimated 2025 revenue · 70+ countries
Kavaken	Renewable asset intelligence	~$2M · Seed	107 plants / 6,309 MW · 9 countries
Werover	Wind blade monitoring	~$2.1M · Seed	1,000+ turbines · 10 countries
Albert Health	Chronic disease management	$3M+ · Seed	300K+ patients · 30+ business partners
Khenda	Manufacturing video analytics	~$2.2M · Seed	Reported 10–15% OEE gains · Customers include Ford, LG, and Philips

Funding figures are approximate or disclosed. Commercial signals are reported unless explicitly identified as estimates. Different traction metrics represent different forms of adoption and should not be treated as directly comparable.

Selected findings

Vispera

Retail intelligence · Series A

Computer vision expertise and international sales support a global retail intelligence platform.

Shelf-image analysis converts retail execution into measurable store data. The founding team combines academic and industry experience in deep learning and computer vision.

~$2.5B TAM · ~$738M SAM · ~25.4% CAGR, 2025–2033

$8.3M estimated 2025 revenue · 70+ countries, company-reported · ~65% international revenue, source unconfirmed · 50+ reported customers including Coca-Cola, Unilever, Carlsberg, Colgate, L’Oréal, Nestlé, Reckitt, and Carrefour

Key risk. Comparable computer-vision competitors exist, while public evidence of a sustained technical performance advantage remains limited.

Werover

Wind blade monitoring · Seed

Proprietary acoustic data enables continuous blade monitoring for wind operators.

Acoustic sensing and machine learning detect, classify, and track hidden turbine-blade damage before failure. The founders have backgrounds across control engineering, renewable energy, sensing, embedded IoT, and AI.

~$40B wind O&M TAM · ~$1.2B blade inspection SAM

1,000+ turbines / 10 countries, reported · 4.1M hours of blade data, reported · ~£12K MRR, source unconfirmed · 40+ blades saved, company-reported

Key risk. The core diligence questions are whether the proprietary data creates a durable detection advantage and whether deployment economics remain attractive after hardware and support costs.

Khenda

Manufacturing video analytics · Seed

Camera-based analytics offer an entry into factory software without requiring heavy sensor integrations.

Existing cameras and process video are used for time studies, line balancing, downtime analysis, and continuous improvement. The founders bring machine-vision and industrial-automation experience.

~$9.7B TAM, 2026 · ~$5.7B SAM proxy · ~46.5% CAGR to ~$47.9B by 2030

10–15% early OEE gains, company-reported · Up to $200K impact on one CNC station, reported · Customers include Ford Automotive, LG Electronics, Philips, Hengst, and ISAIC

Key risk. The main test is whether ROI repeats across independent factories and whether deployments can scale without becoming bespoke services work.

Evidence quality matters

The analysis deliberately separated the existence of a metric from the quality of evidence behind it. A company-reported number, a customer case, a third-party estimate, and independently verified recurring revenue should not carry the same weight in an investment decision.

Legend

E — Third-party estimate
C — Company claim
K — Customer case
U — Source unconfirmed
— — No data found in the reviewed material

Evidence status
Company	Revenue	Customer ROI / impact	Retention	AI differentiation	International sales
Vispera	E — $8.3M estimate	K — Retail cases	— No cohort data	C — Computer-vision capability	U — ~65% international share
Kavaken	U — $1–1.5M estimate	K — Operator cases	— No cohort data	C — Predictive models	C — 9-country footprint
Werover	U — ~£12K MRR	C — 40+ blades saved	— No cohort data	C — Acoustic models	C — 10-country footprint
Albert Health	— Not disclosed	C — Clinical outcomes	C — 64% at 90 days	C — Condition-specific AI	U — ~5-country footprint
Khenda	— Not disclosed	K — Factory case	— No cohort data	C — Video analytics	— No quantified sales

Customer examples are company-published cases. Clinical outcomes are not equivalent to financial ROI. Geographic footprint does not establish international revenue. “No data” means no supporting data was identified in the material reviewed.

What would change the thesis?

Instead of treating the public research as a final answer, I translated the remaining uncertainty into evidence that could strengthen or weaken each investment thesis.

Vispera

Supports: High renewals and recurring revenue.; Clear wins against competing products.
Weakens: Service-heavy revenue or high churn.; Limited performance advantage.

Kavaken

Supports: Verified savings and expansion of paid MW.; Renewals with attractive cost to serve.
Weakens: Savings fail to repeat across plants.; Service effort absorbs gross profit.

Werover

Supports: Accurate detection with few false alerts.; Healthy margins after hardware costs.
Weakens: False alerts undermine operator trust.; Hardware and support erode margins.

Albert Health

Supports: Durable outcomes in controlled cohorts.; Retention supports viable care economics.
Weakens: Selection bias explains outcome gains.; Weak retention or regulatory barriers.

Khenda

Supports: ROI repeats across independent sites.; Pilots convert into recurring contracts.
Weakens: Gains depend on bespoke projects.; Low conversion or costly deployment.

Across all five companies, valuation, ownership, runway, and round terms remain open diligence items.

Sources

This study was built from publicly available company material, customer cases, funding databases, market research, and third-party sources. Figures can vary across sources, so estimates and company-reported metrics are identified where relevant.

The exercise focused not only on forming an investment thesis, but also on identifying what evidence would be required to increase or reduce conviction.

Full research deck