Customers

Clients are described by industry only. Client names are available on request under an NDA.

Key numbers

$600K+

in Azure credits received by six startups after the AICA Azure Scale Programme, more than $100K each. Microsoft alone decides on Azure credits.

solutions published on Microsoft Marketplace, all running in production for customers.

300+

Microsoft-affiliated startups supported by our founder over his career.

Case studies

AI readiness audit for a freight forwarding company

Up to €200K a year

below-cost shipments found in the company's own data

Business result: Below-cost shipment files of up to about €200,000 a year found in the operational data. A business case of €56,000–105,000 a year in freed capacity and €57,000–110,000 a year in margin, against a first implementation wave of €7,000–21,000. Manual effort of 5.5 to 8.2 full-time equivalents measured in short, repetitive tasks.
Challenge: Leadership wanted to know where automation and AI would pay back, based on evidence from the company's own systems rather than on interviews alone.
What we did: A five-week audit that combined process interviews with direct analysis of the transport management system's database: shipment files, costs, invoices and credit notes. Every figure in the business case is marked as measured, calculated or assumed, and the roadmap is split into waves with decision gates. On-premises, cloud and hybrid options were compared against the client's own criteria.

Automated 24/7 support with Azure SRE Agent

24/7

support with the existing team, about 3× faster response at peak load

Business result: Round-the-clock support became possible without adding people. At peak load, response time fell about three times. Estimate: covering 168 hours a week with engineers on shifts would take about 4–5 additional people (168 hours ÷ about 37.5 working hours per person).
Challenge: An LLM solution used by licensed customers needed round-the-clock support, including fast response to load peaks, from a small engineering team.
What we built: Azure SRE Agent in the client's tenant, connected to Azure DevOps service hooks, Boards and pipelines. The agent detects an issue, diagnoses it and acts within limits approved through hooks; during peak load it raised the tokens-per-minute capacity of the model deployment on its own. Every finding becomes a tracked work item, and every change ships through the same pipelines as engineering work. A dedicated agent skill encodes Microsoft Cloud Adoption Framework guidance, and the agent produces live reports on infrastructure, solution and project health.
Stack: Azure SRE Agent, Azure DevOps (Boards, Pipelines, service hooks), Microsoft Foundry, Bicep, Azure Landing Zones.
Verify: AI Business Partner on Microsoft Marketplace · aibp.io

Investor readiness for second-year AI startups

$600K+

in Azure credits for six startups, more than $100K each

Business result: Six B2B AI startups each received more than $100,000 in Azure credits for their next 12 months, over $600,000 in total. Microsoft alone decides on Azure credits; the programme prepares the startup for that decision and closes the gaps it finds. Each startup also receives an investor-ready findings report and a closing progress report.
Challenge: After their first year of Azure credits, B2B AI startups need to pass investor due diligence and keep scaling on Azure.
What we did: A factual assessment against nine published criteria aligned to the Microsoft Cloud Adoption Framework, across technical and commercial perspectives, each rated Met, Gap or Material gap. An investment committee convened by a member of the Microsoft Affiliated Investor Network reviews the findings. Gap closure follows in four sessions over two months: secure the platform, prove recovery, model the credits, prove traction.
Link: How the Azure Scale Programme works

Terminology assistant in Microsoft 365 Copilot and Teams

Best in Power

winner at Microsoft's global 2025 Powerful Devs Hack Together

Business result: One approved definition for every business and technical term, available where people already work: in Microsoft Teams through Copilot. Teams stop losing time to misunderstandings between business and engineering. The solution was named Best in Power among the eight winners of Microsoft's global 2025 Powerful Devs Hack Together, chosen for innovation, impact, technical usability and alignment with the category.
Challenge: In professional and technical environments the same term often means different things to different teams, which leads to rework and misunderstanding.
What we built: Term Buddy, an AI terminology assistant integrated with Microsoft 365 and Copilot in Teams and grounded in an approved glossary through Azure AI Foundry. Glossary owners manage terms in a Power Apps admin portal. The solution ships with CI/CD pipelines, application lifecycle management practices, Bicep deployment scripts and Azure Key Vault for secrets.
Recognition: Microsoft's judges highlighted the integration of Copilot in Teams, Azure AI Foundry and Power Apps, the admin portal and the strong ALM and CI/CD practices.
Stack: Microsoft 365 Copilot in Teams, Azure AI Foundry, Power Apps, Azure Key Vault, Bicep, CI/CD.
Verify: Microsoft Power Platform Developer Blog: winners of the 2025 Powerful Devs Hack Together

Hybrid AI for a telecom operator

personal data leaving the client's server in clear text

Business result: The operator uses Microsoft Foundry models in Azure on regulated data while personal data never leaves its own server in clear text. A standard CPU server replaces a local GPU estate, and further branches and departments are added from the same standard package rather than built again.
Challenge: Use Azure AI on regulated data while keeping personal data inside the country, without the cost of a local GPU estate.
What we built: AI Compliance Partner on a standard CPU server in the client's data centre. Microsoft Presidio and Foundry Local find and mask personal data before any prompt leaves the server, and restore it in the answer. Recognition models were fine-tuned to raise accuracy for Russian and Kazakh personal data. Larger models on Azure, measured with Foundry Evaluation, set the baseline for answer quality. The deployment is installed through Microsoft Marketplace and Bicep.

AI knowledge consultant for a professional-services firm

3 weeks

from start to a production-grade solution, delivered by one engineer

Business result: Decades of the firm's expertise became available to every employee on demand, in the voices of its three leaders, instead of waiting for a leader's time. Delivered in about three weeks by one engineer working with Azure SRE Agent; the full solution can be rebuilt in a new tenant in 2–3 hours.
Challenge: Expertise was spread across books, 224 podcast episodes, methodology documents and the distinct voices of three company leaders.
What we built: An internal AI consultant on Microsoft Foundry with three leader personas, a GraphRAG knowledge graph of 700+ entities and 1,649 relationships, built-in guardrails for out-of-scope topics and Entra ID sign-in, with DEV, UAT and PROD environments, CI/CD and 22 automated tests.
Stack: Microsoft Foundry, Azure AI Search, Microsoft GraphRAG, Azure Cosmos DB, Azure Functions, Azure Static Web Apps, Azure SRE Agent, Azure DevOps, AI Adoption Portal.

AI-first SDLC for a VC-backed coaching startup

10×

increase in software development speed with the same team

Business result: A 10× increase in the speed of software development processes with the same team. The startup ships its AI-powered corporate coaching product faster and with higher quality, while meeting International Coaching Federation (ICF) standards and Responsible AI requirements from day one.
Challenge: A VC-backed startup reinventing corporate coaching with AI needed to transform its entire software development lifecycle into an AI-first SDLC — from ideation through deployment — without growing the team, and in compliance with ICF professional standards.
What we built: An end-to-end AI-first software development lifecycle built on GitHub Spark for rapid prototyping, Azure DevOps for backlog and CI/CD pipelines, and Azure SRE Agent for automated operations and reliability. The mobile application is built with .NET MAUI for cross-platform delivery. Responsible AI practices are embedded across the pipeline, and the Azure architecture follows the Microsoft Cloud Adoption Framework. All coaching workflows are aligned to International Coaching Federation standards.
Stack: GitHub Spark, Azure DevOps (Boards, Pipelines), Azure SRE Agent, .NET MAUI, Microsoft Foundry, Responsible AI Toolkit, Cloud Adoption Framework, Bicep.

Earlier projects led by our founder

Meeting analytics with Microsoft Viva Insights for an HR services holding

57,600 hours a year

returned to the business, about 32 full-time equivalents

Business result: The weekly two-hour all-hands meeting for 600 people was cancelled after the data showed multitasking during it at five times the company's average. That returns about 57,600 working hours a year, the equivalent of about 32 full-time employees. Estimate: 600 people × 2 hours × 48 working weeks = 57,600 hours; 1,800 working hours per full-time employee.
Challenge: A holding with more than 10,000 employees held a weekly two-hour all-hands meeting for 600 people and wanted evidence on whether it was worth that time.
What we did: Analysed collaboration data with Microsoft Viva Insights Advanced Analytics and compared behaviour during the meeting, including multitasking, against the company's own baseline.
Stack: Microsoft Viva Insights Advanced Analytics, Microsoft 365.

AI assistant for contact-centre operators at a consumer bank

98%

answer accuracy, with 25% shorter query handling time

Business result: 98% answer accuracy, 25% shorter query handling time and more than 1,000 queries a month handled with the assistant's help, with fewer errors and repeat calls.
Challenge: Operators spent too long searching for up-to-date information, which slowed service and caused errors and repeat calls.
What we built: An AI assistant on Azure OpenAI integrated with the bank's infrastructure, with RAG and accuracy evaluation in Azure Machine Learning. The bank's team took part in development and testing and was trained to run the solution.

Exchange-rate forecasting for an export-oriented mining company

1 hour instead of a week

to prepare a forecast

Business result: Forecast preparation went from a week to one hour, about 97% less effort per forecast (estimate: a 40-hour working week to 1 hour), with many more forecast scenarios available for each decision.
Challenge: Most revenue came in US dollars, so currency swings hit profit. Analysts built forecasts by hand with several tools and loaded them into reports manually.
What we built: A regression model in Azure Machine Learning, published as a secure endpoint that Power BI Premium calls directly, on an Azure subscription with MFA and cost budgets. The client's analysts were trained to maintain and develop the model themselves.

Learning on Microsoft Teams for an HR holding

6,000

employees trained on Microsoft Teams

Business result: Training for all 6,000 employees runs on Microsoft Teams. The training chatbot was recognised as one of the strongest implementations of its kind, and the client recorded a video testimonial together with Microsoft.
Challenge: Train a workforce of 6,000 employees and give them quick answers on training topics.
What we built: Corporate training moved onto Microsoft Teams, with an internal training chatbot that answers employees' questions.

Meeting feedback bot for a bank

0 minutes

of organiser time to collect feedback

Business result: Every organiser receives anonymous, structured feedback after each meeting. Participants are taken from the invitation automatically, so collecting feedback takes the organiser no time at all.
Challenge: The bank wanted reliable feedback on its many internal meetings, with questions chosen by each organiser.
What we built: A Power Platform bot in Microsoft Teams that sends an adaptive card with the organiser's questions to all participants after online and hybrid meetings. Responses are stored in SharePoint and stay fully anonymous.

Competition scoring system for a regional gas utility

1,000 submissions

scored within a few hours instead of days of manual work

Business result: All 1,000 competition submissions scored, ranked and reported within a few hours. Manual scoring of the same volume would have taken several working days. Judges received a ready-made breakdown by criteria, removing subjective bias from the ranking.
Challenge: The utility ran a professional competition that attracted about 1,000 submissions. Scoring them by hand was slow, error-prone and inconsistent across judges.
What we built: A Power Platform solution that ingested submissions, applied weighted scoring rubrics configured by the organisers, aggregated judge scores automatically and produced ranked results with per-criterion breakdowns, all within the utility's existing Microsoft 365 environment.
Stack: Power Platform (Power Apps, Power Automate), SharePoint, Microsoft 365.

Want to learn more about a specific case?

Client names and project details are available on request under an NDA.

Request customer details